Machine learning of cerebello-cerebral functional networks for mild cognitive impairment detection

Machine learning of cerebello-cerebral functional networks for mild cognitive impairment detection | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Machine learning of cerebello-cerebral functional networks for mild cognitive impairment detection Qun Yao, Liangcheng Qu, Bo Song, Xixi Wang, Tong Wang, Wenying Ma, and 8 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-2663342/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background: Early identification of degenerative processes in Alzheimer’s disease (AD) is essential. Cerebello-cerebral network changes can be used for early diagnosis of dementia and its stages, namely mild cognitive impairment (MCI) and AD. Methods: Features of cortical thickness (CT) and cerebello-cerebral functional connectivity (FC) extracted from MRI data were used to analyze structural and functional changes, and machine learning for the disease progression classification. Results: CT features have an accuracy of 92.05% for AD vs. HC, 88.64% for MCI vs. HC, and 83.13% for MCI vs. AD. Additionally, combined with convolutional CT and cerebello-cerebral FC features, the accuracy of the classifier reached 94.12% for MCI vs. HC, 90.91% for AD vs. HC, and 89.16% for MCI vs. AD, evaluated using support vector machines. Conclusions: The proposed pipeline offers a promising low-cost alternative for the diagnosis of preclinical AD and can be useful for other degenerative brain disorders. Alzheimer’s disease Mild cognitive impairment Functional magnetic resonance imaging Cerebellum Machine learning Figures Figure 1 Figure 2 Figure 3 Background Alzheimer’s disease (AD), the most frequent cause of dementia, is characterized by abnormal buildup of neurofibrillary tangles and amyloid plaques in the brain, which affects memory, thinking, and behavior [1]. AD is an irreversible and incurable disease, therefore, early detection, such as mild cognitive impairment (MCI) during its preclinical stage, is critical for implementing viable interventions to delay its progression [2]. According to previous studies, a significant proportion of patients with MCI (approximately 10-15% from referral sources such as memory clinics and AD centers) will annually develop AD [3], while healthy control population (i.e., healthy elderly people) converts at a 1-2% rate [4]. The neuropathological progression in AD can be observed several years before clinical symptoms appear [5-8]. Furthermore, its progression is slow and insidious, making the preclinical stage last for several years, even starting decades before the first symptoms of dementia appear. Identification trials targeting individuals in the preclinical stage of AD may be the most effective approach. Based on this reasoning, MCI patients are an ideal population to explore neurophysiological profiles that predict who will develop dementia. In recent years, increasing efforts have shifted the frontline to possible detection in the early stage of MCI, earning more valuable time to lower the high conversion rate. A crucial point is whether the pathophysiology of AD would be detected long before the actual diagnosis of the disease. Consequently, the identification of preclinical AD when severe brain damage associated with AD, such as extensive brain atrophy, has not appeared yet, could significantly improve the benefits of new drugs and vaccines. Resting-state functional magnetic resonance imaging (rs-fMRI) has emerged as a popular modality to investigate the changes of early brain function in patients with MCI. As few visible structural brain changes are present in magnetic resonance imaging (MRI) of MCI patients, rs-fMRI is an in vivo functional imaging technique, which measures blood oxygen level-dependent (BOLD) signals when scanning subjects at natural rest without any explicit task involvement [9]. Functional connectivity (FC) between distant brain regions can be measured based on their time-synchronized rs-fMRI BOLD signals, yielding brain “functional connectomics” or whole-brain functional networks [9]. Previous research has demonstrated that abnormal functional connections and brain networks between brain areas are discovered as early as the MCI stage [10-12]. Furthermore, the minority of pioneering research has focused on the brain networks changes in MCI subjects by using group-level statistic comparisons based on simple regional pair-wise FCs or network properties. Rapid development of neuroimaging techniques has made the integration of large-scale, high-dimensional multimodal neuroimaging data challenging. Subsequently, there has been a surge in interest in computer-aided machine learning (ML) methodologies for integrative analysis. Over the past few years, several MRI biomarkers have been proposed for the classification of AD patients at various disease stages [13-19]. Most of these studies have used simple but very effective classifiers, such as support vector machine (SVM) [20-22], which can be mathematically simplified using optimization techniques [23]. Nevertheless, despite many efforts, the identification of efficient specific AD biomarkers for early diagnosis and disease progression prediction remains a challenging task and requires more research. Previous research has concentrated on the cerebral region, with little emphasis dedicated to the function of the cerebellum in the cognitive regulation of the AD spectrum. The function of the cerebellum in cognitive processing has recently sparked considerable interest in cognitive neuroscience [24-26]. It was revealed that the surface of the cerebellum is folded more tightly than the cerebral cortex, with almost 80% of the neocortex connected to the cognitive network of the brain [27]. Studies have suggested that the role of the cerebellum in the processing of the various cognitive functions such as working memory [28], emotion [29], language [30], and other executive functions [31] can be attributed to its extensive connections to the cortical and subcortical areas (particularly between the Crus I and Crus II, known as the lateral “limbic” cerebellar circuit) [32]. The cerebellum acts as a key part of the adaptive regulator in cognitive processing, supervising the working brain areas, detecting cognitive patterns, changes and errors, and generating responses and distributing feedback to numerous cerebral areas, to increase the effectiveness of brain activity [33]. The previous study from the author’s group showed that the cognitive impairment after cerebellar injury was involved in multiple cognitive domains. The connection efficiency and information integration ability of the brain network were disrupted, indicating that the cerebellum could participate in the integration and regulation of the cognitive network [34]. Additionally, in another work, it was observed that the cerebellum Crus II exhibited increased activity in the preclinical stage of AD, reflecting a compensatory function that could mitigate early AD symptoms [35]. Cerebellar atrophy was shown to be an essential factor in the clinical progression of AD in previous works, but mainly occurred in the late stage of the disease. However, even in the later stages, the shrinkage rate was lower than the average shrinkage in the cerebrum and was independent of the AD moderators [36]. Therefore, the cerebellum survived the preclinical stage of AD and may be resistant to certain neurodegenerative mechanisms [37]. These findings suggest that the cerebellum connectivity may be a potential identification biomarker in the preclinical AD stage. In this paper, it is postulated that the cerebellum serves as the compensatory regulation center of the brain, modulating cerebello-cerebral networks in the early stage of the disease. A novel unified framework is hereby proposed, which utilizes ML to extract deep embedded features from cerebello-cerebral functional networks, without sacrificing the rich spatial information of the images. Preclinical AD pathology has been hypothesized to be detectable by neuroimaging techniques [38]. Traditional approaches might encounter difficulties when processing highly complex spatial-temporal connectomics information, resulting in suboptimal diagnostic accuracy of MCI, partially due to more subtle alterations in brain networks in comparison of MCI vs. healthy controls (HC), and MCI vs. AD groups. Using structural and cerebellar functional features, ML can be very useful in the detection of disease-related abnormal patterns and improving accuracy of individualized diagnosis of MCI. In the present study, the recursive binary method was used to retrieve the dominant features (combined with structural and cerebellar functional features) for the SVM classification of the AD, MCI and HC groups. Methods Participants The present study was approved by the Medical Research Ethical Committee of Nanjing Brain Hospital in Nanjing, China. In total, 43 AD subjects, 40 MCI subjects and 45 HC subjects were recruited between June 2019 and April 2022, following application of inclusion and exclusion criteria. All subjects were right-handed. Written informed consent was obtained from all participants before enrolling in the study. Patient recruitment was performed based on the current diagnostic criteria provided by the National Institute on Aging and the Alzheimer’s Association working group (NIA-AA) in 2018 [ 39 ]. The guidelines are characterized by episodic memory loss in AD pathology supported by the evidence of cerebrospinal fluid (CSF) and imaging biomarkers. Inclusion criteria for patients included age 60–80 years; no deficits in vision or hearing; >8th -grade education; T2-weighted MRI without indication of infection, infarction or focal lesions within 12 months (assessment by two independent expert radiologists). The inclusion criteria for the HC group included no current cognitive issues; no neurological or psychiatric diseases; clinical dementia rating (CDR) score of zero [ 40 ]. Exclusion criteria included patients with other known causes of dementia (i.e., frontotemporal dementia, dementia with Lewy bodies, vascular dementia, severe depression, cerebrovascular disease, poisoning, tumors, metabolic diseases and infections); limitations in MRI scanning, such as claustrophobia or pacemaker implantation; a history of psychiatric or other neurological disorders. All patients underwent a complete clinical investigation, an extensive neuropsychological assessment that explored all cognitive domains, brain MRI scanning and CSF analysis. The eligibility of the patients for this study was confirmed by highly experienced field investigators. Clinical and neuropsychiatric evaluation All participants underwent comprehensive and standard neuropsychological assessments. Global cognition was evaluated using Mini-Mental State Examination (MMSE) [ 41 ], Montreal Cognitive Assessment (MoCA) [ 42 ], CDR and Alzheimer’s Disease Assessment Scale - Cognitive section (ADAS-Cog) [ 43 ]. Episodic memory was assessed using the Rey Auditory-verbal Learning Test (RAVLT) [ 44 ]. Visuospatial abilities were assessed with the Clock Drawing Test (CDT) [ 45 ]. Language function was determined with the Boston Naming Test (BNT) [ 46 ] and the Verbal Fluency Test (VFT) [ 47 ]. Executive function was assessed using parts A and B of the Trail Making Test (TMT) [ 48 ]. Attention was assessed using the Symbol Digit Modalities Test (SDMT) [ 49 ] and the Digit Span Test (DST) [ 50 ]. The emotional condition of the subjects was evaluated using the Hamilton Depression Scale (HAMD) [ 51 ] and the Hamilton Anxiety Scale (HAMA) [ 52 ]. Sleep quality was assessed using the Pittsburgh Sleep Quality Inventory (PSQI) [ 53 ]. Activities of Daily Living (ADL) were used to assess the ability of participants to care for themselves in daily life. Senior neuropsychologists validated all the scales, which were evaluated by experienced clinicians. Cerebrospinal fluid biomarkers A lumbar puncture was performed in all patients to confirm the typical CSF profile of AD pathology with reduced amyloidβ1–42 concentrations and increased phosphorylated- and total- tau levels. Amyloidβ1–42, p-tau, and t-tau levels were detected using the INNOBIA AlzBio3 immunoassay kit-based reagents (Innotest, Fujirebio, Ghent, Belgium). Magnetic resonance imaging data collection All scans were performed using a Siemens 3.0 T scanner (Siemens, Verio, Germany) with an 8-channel radiofrequency coil. Participants were asked to remain as still as possible, close their eyes, remain awake, and not think of anything. Three-dimensional T1 weighted images were collected in a sagittal orientation using a 3D-MPARGE sequence with the following parameters: time repetition (TR) = 1,900 ms, echo time (TE) = 2.48 ms, inversion time (TI) = 900 ms, 176 sagittal slices, 1.0 mm slice thickness, gap = 0.5 mm, matrix = 256 × 256, flip angle (FA) = 9°, field of view (FOV) = 256 × 256 mm, and voxel size = 1 × 1 × 1 mm 3 . Functional images were collected using a gradient-recalled echo-planar imaging pulse sequence with 240-time points; TE = 30 ms; TR = 2,000 ms; number of slices = 36, FOV = 220 × 220 mm 2 ; matrix = 64 × 64; flip angle = 90°; 4.0 mm thickness, and gap = 0 mm. The image acquisition for each subject required about 14 min. Magnetic resonance imaging processing The structure MRI (sMRI) data were processed with the FreeSurfer Image Analysis Suite ( http://surfer.nmr.mgh.harvardedu/ ) which included: automatic Talairach space transformation, image intensity inhomogeneity correction, non-brain tissue removal, intensity normalization, tissue segmentation [ 54 ], automatic topology correction, surface deformation to generate gray/white matter boundaries, fragmentated gray matter/CSF boundary and cerebral cortex. The Desikan-Killiany atlas (34 areas per hemisphere) was used for parcellation [ 55 ]. The rs-fMRI data were preprocessed using the Data Processing and Analysis for Brain Imaging tool (DPABI, http://www.rest.restfmri.net ) [ 56 ]. The first ten volumes of the resting session were discarded for each participant, to account for factors of magnetization equilibrium and environment adaptation. The remaining images were corrected based on slice timing and motion correction (head motion ≤ 2 mm, head motion angle ≤ 2°). Following this, the functional images were co-registered with the high-resolution 3D-T1 structural images. Then, the 3D-T1 images were normalized to the Montreal Neurological Institute (MNI) space using non-linear warping, based on the Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra. After all images were spatially normalized to the T1 space, they were resampled into 3 × 3 × 3 mm 3 voxels and spatially smoothed with a Gaussian filter of 6 mm full-width at half-maximum. The fMRI data were then temporally bandpass-filtered (0.01–0.08 Hz) to remove low-frequency drifts and high-frequency physiological noise. The noise covariates were regressed to further reduce the confounding artifacts of resting head movement and physiological noise (i.e., respiration and cardiac fluctuations), including the Friston 24-motion parameter model, the global mean, the white matter and the cerebrospinal fluid signals. Feature extraction and selection Structural and functional features from each subject were selected for subsequent feature selection (Fig. 1 ). (1) In the structural section, there were 68 MRI cortical thickness (CT) features. (2) In the functional section, there were 180 connectivity features (i.e., FC between the bilateral cerebellum and the brain areas). Two ML experiments were performed in this study. The first experiment was a three-class classification, using only structural features, to predict individual specific diagnosis (AD, MCI, HC). The structural features were combined with the functional features in the second experiment during training to predict the diagnosis for the three groups. Cortical thickness features The atlas used in Desikan-Killiany template included 68 cortical regions (Fig. 1 ). For each cortical region, CT was calculated from MRI features. The CT of each cortical vertex was calculated as the average shortest length between white and pale surfaces. For each participant, this section yielded 68 MRI features. The bilateral frontal, temporal, parietal and occipital CT features were selected to construct eight training sets, respectively. Cerebello-cerebral network features To define nodes of the functional brain network, the brain was segmented into 90 regions using the automatic anatomical labeling (AAL) [ 57 ]. The bilateral cerebellar Crus II areas were extracted as regions of interest (ROI). The FC analysis was voxel-wise performed between each ROI and the entire brain. To establish the reference time series of seed points, the voxels of each seed region from every subject were extracted and averaged. Subsequently, the correlation coefficient between the reference time series and the time series involving all other brain voxels was calculated. The correlation coefficients were transformed into z-values using the Fisher r-to-z transformation for normality enhancement. In this part, 180 bilateral cerebellum connectivity features (90×2 = 180) were obtained for further feature selection. The CT combined with the FC between the bilateral cerebellum Crus II and the brain regions (including the bilateral frontal, bilateral parietal, bilateral temporal and bilateral occipital lobes) were selected as eight training datasets, respectively. All subdivisions within brain regions were combined into feature subsets using the recursive binary method. These feature subsets were used as input to the ML model for training and testing, to assess the performance of subsets in classification tasks. SVM classification The SVM classification model was implemented in MATLAB (The Math Works, Natwick, MA) and LIBSVM ( http://www.csie.ntu.edu.tw/cjlin/libsvm/ ). Binary classification of data was performed in a supervised learning method by projecting low-dimensional linearly indistinguishable problems to high-dimensional to construct optimal decision classification surface [ 58 ]. Let the optimal classification surface established by the SVM during training be \({\omega }\) , as in Eq. 2 − 1, \({\omega }^{T}x+b=0\) , (2 − 1) where x is the feature vector, T is the transpose matrix, and b is a constant. The construction of the optimal decision classification surface in SVM is based on \(\omega\) , by maximizing the subsurface, as in Eq. 2–2, with the linear function as the constraint, \(\text{m}\text{a}\text{x} \frac{1}{‖\omega ‖} s.t. ,{y}_{i}({\omega }^{T}{x}_{i}+b)\ge 1\) (2–2) where y represents the sample category, Eq. 2–2 can be converted to Eq. 2–3 \(.\) \(\text{m}\text{i}\text{n} \frac{1}{2}{‖\omega ‖}^{2} s.t. ,{y}_{i}({\omega }^{T}{x}_{i}+b)\ge 1\) (2–3) The optimization of \({\omega }\) is a convex quadratic programming problem with linear constraints It can be transformed to the dyadic problem domain by adding Lagrange multipliers \({\alpha }\) to the linear constraint function, as in Eq. 2–4. \(\text{L}\left(\omega ,b,\alpha \right)= \frac{1}{2}{‖\omega ‖}^{2}-\sum _{i=1}^{n}{\alpha }_{i}\left({y}_{i}\left({\omega }^{T}{x}_{i}+b\right)-1\right)\) (2–4) Finally, the SMO algorithm can be used to solve for the \({\alpha }\) multiplier. The kernel function used in the SVM classifier was the radial basis kernel function (RBF), where the penalty parameter C and the kernel bandwidth σ in the kernel function ranged from [ \({4}^{-4}\) , \({4}^{4}\) ]. The RBF kernel was defined as follows, \(\text{K}\left({x}_{1},{x}_{2}\right)= \text{e}\text{x}\text{p}(-\frac{‖{x}_{1}-{x}_{2}‖}{2{\sigma }^{2}})\) (2–5) where \({x}_{1}\) , \({x}_{2}\) are two eigenvectors and \({\sigma }\) is the width parameter of the RBF kernel. Because the data area linearly indiscriminable, SVM will first complete the basic computation in the low-dimensional space, and then introduce the kernel function to map the input space to a high-dimensional space to construct the optimal classification surface. Performance evaluation <p Models were evaluated using 4-fold cross-validation. Four models were trained for each experiment, with 75% of the data randomly selected for training and 25% for testing for each one. Thus, this process produced one prediction per sample of the entire dataset. The cross-validation test for each set of feature subsets was repeated four times to assess the variability of the evaluation metrics. Therefore, the resulting evaluation metrics were max over the four cross-validation iterations. Metrics of model performance were accuracy (ACC) (percent correctly classified), sensitivity (SEN) (true positives, correctly identified) and specificity (SPE) (true negatives, correctly identified). Another effective way to evaluate classification results is the receiver operating characteristic (ROC) curve. The ROC curve is the plot of true-positive rate against false-positive rate by changing the discrimination threshold, summarizing the performance of the classifier. The ROC curve results are usually interpreted by the area under the curve (AUC). Statistical analysis All data were tested for normality and variance consistency. Additionally, appropriate statistical tests (i.e., ANOVA, independent sample t-tests or Chi-squared tests) were performed to assess the clinical and demographical differences among the groups (AD, MCI, HC), reported in Table 1 . Statistical analyses were performed with IBM SPSS 25.0 software (SPSS Inc., Chicago, Illinois, USA). A two-sided P < 0.05 was acknowledged as statistically significant. The 95% confidence intervals (CIs) of ACC, SEN, and SPE were calculated by the exact Clopper–Pearson method. The DeLong method was used to calculate the CIs of AUC. Table 1 The table contains the mean values ( ± standard deviation) of demographic and clinical data . Significant group differences were obtained using ANOVA, independent sample t-tests or Chi-squared tests. two-sided P < 0.05 was considered significant. Abbreviations: MMSE: Mini-Mental State Examination; MoCA: Montreal Cognitive Assessment; HAMA: Hamilton Anxiety Scale; HAMD: Hamilton Depression Scale; ADL: Activities of Daily Living; PSQI: Pittsburgh Sleep Quality Inventory; RAVLT: Rey Auditory Verbal Learning Test; CDT: Clock Drawing Test; SDMT: Symbol Digit Modalities Test; DST: Digit Span Test; BNT: Boston Naming Test; VFT: Verbal Fluency Task; TMT: Trail-Making Test; CSF: cerebrospinal fluid; p tau: phosphorylated-tau. AD MCI HC P-value (n = 43) (n = 40) (n = 45) Age (years) 66.93 ± 6.85 66.93 ± 7.82 64.24 ± 6.42 0.123 Gender (male/female) 26/17 19/21 23/22 0.470 Education (years) 9.47 ± 3.37 9.98 ± 3.19 8.87 ± 3.23 0.296 MMSE 15.14 ± 3.73 21.40 ± 2.86 27.13 ± 1.22 < 0.001 MoCA 11.81 ± 3.19 18.83 ± 2.80 25.09 ± 1.08 < 0.001 HAMA 3.55 ± 1.91 3.95 ± 1.84 3.57 ± 1.50 0.521 HAMD 3.72 ± 1.93 3.20 ± 1.68 3.26 ± 1.48 0.313 ADL 27.41 ± 4.56 21.53 ± 1.31 20.24 ± 0.57 < 0.001 PSQI 3.26 ± 1.80 2.65 ± 2.12 2.97 ± 1.60 0.329 RAVLT (memory) 12.53 ± 4.58 17.23 ± 3.48 20.69 ± 2.48 < 0.001 CDT (visual spatial memory) 11.30 ± 4.84 23.38 ± 6.48 26.71 ± 3.12 < 0.001 SDMT (attention) 10.37 ± 4.48 22.05 ± 8.90 32.20 ± 5.88 < 0.001 DST (attention) 5.21 ± 1.95 7.48 ± 1.84 10.27 ± 1.83 < 0.001 BNT (verbal naming ability) 14.58 ± 5.11 19.60 ± 4.17 23.24 ± 2.47 < 0.001 VFT (verbal fluency) 7.60 ± 2.56 10.80 ± 3.11 14.91 ± 2.27 < 0.001 TMT-A (attention) 222.40 ± 112.48 119.58 ± 36.31 67.13 ± 13.10 < 0.001 TMT-B (execution) 362.51 ± 115.94 260.68 ± 94.65 117.91 ± 17.89 < 0.001 CSF amyloidβ1–42, pg/ml 373.54 ± 97.36 414.79 ± 147.66 0.134 CSF total tau, pg/ml 523.61 ± 125.12 469.07 ± 151.65 0.083 CSF p tau, pg/ml 110.78 ± 65.43 91.37 ± 55.81 0.151 Results Clinical characteristics assessment The demographic and clinical characteristics of each participant are summarized in Table 1 . No significant differences were observed with regard to age, gender and education, among the three groups. On the other hand, significant differences were present for each cognitive domain. Overall cognitive levels, episodic memory, visuospatial working memory, attention, language, and execution were significantly lower in the AD group compared to both the MCI and HC groups. The above cognitive domains were significantly lower in the MCI group compared to the HC group (p < 0.001). Feature classification The performance metrics (ACC, SEN and SPE) of sixteen SVM classifiers for separating the three groups (AD, MCI, HC) using subsets of sMRI and/or rs-fMRI features, based on two atlases for each modality, are shown in Table 2 , Table 3 and Fig. 3 . Results show that the optimal classifier was based on sMRI (left temporal CT) features (with ACC of 92.05%, SEN of 95.35% and SPE of 88.89%) for distinguishing between AD and HC (Table 2 ). In terms of separating MCI from HC, the highest ACC of 88.64%, SEN of 85.00% and SPE of 91.11% were achieved by utilizing only the sMRI features. In addition, the 83.13% ACC, 76.74% SEN and 90.00% SPE were obtained for MCI vs. AD, using sMRI features. Table 2 Classification results for the three groups using cortical thickness features Features ACC (%) M [CIs] SEN(%) M [CIs] SPE(%) M [CIs] AUC M [CIs] Right Frontal CT AD vs. HC 86.36 [77.39–92.75] 90.70 [77.86–97.41] 82.22 [67.95-92.00] 0.80 [0.72–0.88] MCI vs. HC 85.88 [76.64–92.49] 85.00 [70.16–94.29] 86.67 [73.21–94.95] 0.78 [0.69–0.87] MCI vs. AD 81.93 [71.95–89.52] 74.42 [58.83–86.48] 90.00 [76.34–97.21] 0.78 [0.69–0.87] Left Frontal CT AD vs. HC 87.50 [78.73–93.59] 90.70 [77.86–97.41] 84.44 [70.54–93.51] 0.83 [0.75–0.91] MCI vs. HC 83.53 [73.91–90.69] 75.00 [58.80-87.31] 91.11 [78.78–97.52] 0.75 [0.66–0.84] MCI vs. AD 83.13 [73.32–90.46] 79.07 [63.96–89.96] 87.50 [73.20-95.81] 0.80 [0.71–0.89] Right Parietal CT AD vs. HC 85.23 [76.06–91.89] 79.07 [63.96–89.96] 91.11 [78.78–97.52] 0.79 [0.70–0.87] MCI vs. HC 81.18 [71.24–88.84] 75.00 [58.80-87.31] 86.67 [73.21–94.95] 0.77 [0.68–0.86] MCI vs. AD 75.90 [65.27–84.62] 69.77 [53.87–82.82] 82.50 [67.22–92.66] 0.69 [0.59–0.79] Left Parietal CT AD vs. HC 82.95 [73.45–90.13] 76.74 [61.37–88.24] 88.89 [75.95–96.29] 0.71 [0.62–0.80] MCI vs. HC 84.71 [75.27–91.60] 75.00 [58.80-87.31] 93.33 [81.73–98.60] 0.75 [0.66–0.84] MCI vs. AD 73.49 [62.66–82.58] 69.77 [53.87–82.82] 77.50 [61.55–89.16) 0.68 [0.58–0.78] Right Temporal CT AD vs. HC 90.91 [82.87–95.99] 93.02 [80.94–98.54] 88.89 [75.95–96.29] 0.86 [0.79–0.93] MCI vs. HC 87.06 [78.02–93.36] 80.00 [64.35–90.95] 93.33 [81.73–98.60] 0.79 [0.70–0.88] MCI vs. AD 80.72 [70.59–88.56] 74.42 [58.83–86.48] 87.50 [73.20-95.81] 0.76 [0.67–0.85] Left Temporal CT AD vs. HC 92.05 [84.30-96.74] 95.35 [84.19–99.43] 88.89 [75.95–96.29] 0.94 [0.89–0.99] MCI vs. HC 88.24 [79.43–94.21] 85.00 [70.16–94.29] 91.11 [78.78–97.52] 0.90 [0.84–0.96] MCI vs. AD 83.13 [73.32–90.46] 76.74 [61.37–88.24] 90.00 [76.34–97.21] 0.78 [0.69–0.87] Right Occipital CT AD vs. HC 73.86 [63.41–82.66] 60.47 [44.41–70.52] 86.67 [73.21–94.95] 0.69 [0.59–0.79] MCI vs. HC 77.65 [67.31–85.97] 65.00 [48.32–79.37] 88.89 [75.95–96.29] 0.68 [0.58–0.78] MCI vs. AD 73.49 [62.66–82.58) 65.12 [49.07–78.99] 82.50 [67.22–92.66] 0.69 [0.59–0.79] Left Occipital CT AD vs. HC 73.86 [63.41–82.66] 65.12 [49.07–78.99] 82.22 [67.95-92.00] 0.67 [57.18–76.82] MCI vs. HC 78.82 [68.61–86.94] 75.00 [58.80-87.31] 82.22 [67.95-92.00] 0.72 [62.45–81.55] MCI vs. AD 72.29 [61.38–81.55] 86.05 [72.07–94.70] 57.50 [40.89–72.96] 0.69 [59.05–78.95] Table 3 Classification results for the three groups using cortical thickness and cerebello-cerebral functional connectivity features Features ACC (%) M (CIs) SEN(%) M (CIs) SPE(%) M (CIs) AUC M (CIs) CT + Right Cerebellar – frontal FC AD vs. HC 90.91 [82.87–95.99] 95.35 [84.19–99.43] 86.67 [73.21–94.95] 0.88 [0.81–0.95] MCI vs. HC 94.12 [86.80-98.06] 97.50 [86.84–99.94] 91.11 [78.78–97.52] 0.95 [0.90–0.99] MCI vs. AD 89.16 [80.41–94.92] 88.37 [74.92–96.11] 90.00 [76.34–97.21] 0.90 [0.84–0.96] CT + Left Cerebellar – frontal FC AD vs. HC 87.50 [78.73–93.59] 95.35 [84.19–99.43] 80.00 [65.40-90.42] 0.75 [0.66–0.84] MCI vs. HC 89.41 [80.85–95.04] 95.00 [83.08–99.39] 84.44 [70.54–93.51] 0.90 [0.90–0.99] MCI vs. AD 87.95 [78.96–94.07] 86.05 [72.07–94.70] 95.00 [83.08–99.39] 0.80 [0.71–0.89) CT + Right Cerebellar – parietal FC AD vs. HC 87.50 [78.73–93.59] 95.35 [84.19–99.43] 80.00 [65.40-90.42] 0.78 [0.69–0.87] MCI vs. HC 90.59 [82.29–95.85] 95.00 [83.08–99.39] 86.67 [73.21–94.95] 0.82 [0.74–0.90] MCI vs. AD 87.95 [78.96–94.07] 83.72 [72.04–94.70] 82.50 [67.22–92.66] 0.82 [0.74–0.90] CT + Left Cerebellar – parietal FC AD vs. HC 86.36 [77.39–92.75] 95.35 [84.19–99.43] 77.78 [62.91–88.80] 0.70 [0.60–0.80] MCI vs. HC 85.88 [76.64–92.49] 92.50 [79.61–98.43] 80.00 [65.40-90.42] 0.72 [0.62–0.82] MCI vs. AD 80.72 [70.59–88.56] 76.74 [61.37–88.24] 85.00 [70.16–94.29] 0.71 [0.61–0.81] CT + Right Cerebellar – temporal FC AD vs. HC 86.63 [77.39–92.75] 93.02 [80.94–98.54] 80.00 [65.40-90.42] 0.73 [0.64–0.82] MCI vs. HC 88.24 [79.43–94.21] 87.50 [73.20-95.81] 88.89 [75.95–96.29] 0.77 [0.68–0.86] MCI vs. AD 87.95 [78.96–94.07] 90.70 [77.86–97.41] 85.00 [70.16–94.29] 0.82 [0.74–0.90] CT + Left Cerebellar – temporal FC AD vs. HC 87.50 [78.73–93.59] 90.70 [77.86–97.41] 82.22 [67.95-92.00] 0.76 [0.67–0.85] MCI vs. HC 85.88 [76.64–92.49] 90.00 [76.34–97.21] 80.00 [65.40-90.42] 0.73 [0.64–0.82] MCI vs. AD 85.54 [76.11–92.30] 88.37 [74.92–96.11] 82.50 [67.22–92.66] 0.77 [0.68–0.86] CT + Right Cerebellar – occipital FC AD vs. HC 78.41 [68.35–86.47] 90.07 [77.86–97.41] 66.67 [51.05-80.00] 0.70 [0.60–0.80] MCI vs. HC 80.00 [69.92–87.90] 85.00 [70.16–94.29] 75.56 [60.46–87.12] 0.68 [0.58–0.78] MCI vs. AD 77.11 [66.58–85.62] 72.09 [56.33–84.67] 82.50 [67.22–92.66] 0.66 [0.56–0.76] CT + Left cerebellar – occipital FC AD vs. HC 73.86 [63.41–82.66] 88.37 [74.92–96.11] 60.00 [44.33–74.30] 0.64 [0.54–0.74] MCI vs. HC 78.18 [71.24–88.84] 87.50 [73.20-95.81] 75.56 [60.46–87.12] 0.70 [0.60–0.80] MCI vs. AD 79.52 [69.24–87.95] 76.74 [61.37–88.24] 82.50 [67.22–92.66] 0.70 [0.60–0.80] Table 3 shows the diagnostic performance of the mixed sMRI (CT) and rs-fMRI (cerebello-cerebral FC) features in the three classification tasks. The classification performance of the combined features of CT with FC from the right cerebellum to the left frontal lobe shows the highest ACC (94.43%), SEN (97.5%) and SPE (91.11%) in MCI vs. HC. AD patients have ACC of 90.91%, SEN of 95.35% and SPE of 86.67% compared with HC. In addition, the model that utilized mixed features reached an ACC of 89.16%, SEN of 88.37% and SPE of 90.00% in MCI vs. AD. Furthermore, the ACC of 88.24%, SEN of 87.50% and SPE of 88.89% were obtained for MCI vs. HC, using combination with mixed features from right cerebellum to bilateral temporal lobes. Finally, the ACC of 90.59%, SEN of 95.00% and SPE of 86.67% were obtained for MCI vs. HC, using combination with mixed features from right cerebellum to bilateral parietal lobes. The highest-ranked features are shown in Table 4 and Table 5. The effectiveness of the classification method combining different measures was further analyzed using the AUCs for the concatenation of all features (Fig. 3 D). Figure 2 shows the ROC curves for individual features and all features (i.e., CT and cerebello-cerebral FC) combinations for each classification group. Discussion Effective and accurate diagnosis of AD is essential to initiate effective treatment. Particularly, early diagnosis of AD is pivotal in therapeutic development and ultimately for effective patient care [ 2 ]. ML methods can be used to identify clinical features and characteristic MR images and patterns for diagnostic predictions [ 59 , 60 ]. This approach can potentially help with dementia predisposition identification in those who develop cognitive complaints [ 61 – 64 ], as this is a common question faced by clinicians. In this study, results showed that the ML model that utilized combined sMRI and rs-fMRI features was more accurate in predicting the diagnostic status of MCI than the ML models using brain morphological features only. In this study, sMRI classification features were first extracted to identify atrophy patterns for the three groups (MCI, AD, HC). The findings were generally consistent with previous research, in which one of the earliest imaging marker was atrophy in the temporal lobe, and specifically in the parahippocampus and entorhinal cortex, which has also been previously found to predict MCI to AD progression [ 65 , 66 ]. The left temporal CT features had the highest classification performance in AD vs. HC groups with an ACC of 92.05%, SEN of 95.35% and SPE of 88.89%. Similarly, for MCI vs. HC groups, the highest atrophy was observed in the left temporal cortex, with diagnostic accuracy of 88.64%, sensitivity of 85.00% and specificity of 91.11%. Finally, for MCI vs. AD groups, in the modeling using sMRI data of the frontal and temporal cortex, an ACC of 83.13% was obtained, which also verifies that the frontal and temporal lobe is involved in the early stage. These areas have been shown to be involved in complex cognitive behavior, decision making and personality expression [ 50 ], showing potential for the early identification of AD progression. Some individuals with MCI present with minimal visible structural brain changes. Some MCI patients have structural atrophy patterns similar to AD, suggesting that MCI may be more similar to AD than HC. The subtle changes between MCI and HC or the similarity between MCI and AD presents the ML models with a higher difficulty challenge to discriminate between the groups, and therefore requires more features. Additionally, clinical experience reflects this challenge as the differentiation between MCI and HC, as well as MCI and AD is based on functional decline, which is occasionally difficult to identify or quantify, causing overlap among groups. Thus, the results identifying features able to distinguish MCI from the other two groups are more clinically meaningful. Combining different types of data could help to further improve the accuracy of the ML model to distinguish between clinically relevant groups. The classification results obtained by combining sMRI and rs-fMRI features in the present study are better than those obtained by the unimodal (sMRI\rs-fMRI) approach, including those of previous research [ 11 , 67 ], for distinguishing MCI from AD and HC (Table 3 ). Most previous studies that constructed brain networks only considered cerebral structural or functional features, while ignoring the cerebellum, and obtained an accuracy lower than that of the present study [ 68 – 72 ]. In this work, the diagnostic performance of combined features of CT with FC from the right cerebellum to the left frontal lobe reached 94.43%. In addition, the mixed-features model reached an accuracy of 90.01% and 89.16% for AD vs. HC and MCI vs. AD, respectively. Previous work from our group revealed that the cerebellum showed increased activity of the frontal and temporal lobes in the pre-dementia stage of AD, reflecting a compensatory function that could mitigate early AD symptoms [ 35 ]. The current study demonstrates that the compensatory regulation of the cerebellum to the frontal lobe is the most significant, which strongly agrees with the findings relating to the enhanced FC between cerebellum and the frontal regions by cerebellum repetitive transcranial magnetic stimulation intervention presented in another work [ 73 ]. Therefore, it is postulated that when early cognitive impairment occurs, the frontal lobe receives increased assistance from the cerebellum. Furthermore, based on the classification model of CT combined with FC from right cerebellum to bilateral temporal and bilateral parietal lobes, respectively, high ACC of 88.24% and 90.59% were obtained for the classification of MCI vs. HC. This result outperformed other biomarkers for early MCI classification, indicating that the cerebellum has a high classification accuracy rate. Consequently, network-derived regulation is a highly effective biomarker. The crucial role of the cerebellum in brain regulation during the development of early cognitive impairment diseases was confirmed in this study. This conclusion may provide help for the diagnosis and treatment of preclinical AD in the future. In this article, the potential of the cerebellum-oriented functional-coefficient network of the brain to be used as a diagnostic biomarker for subjects with early cognitive impairment was explored. The frontal lobe functioned as a “ rich club ” structure with highly interconnected nodes resulting in a large number of connections between networks [ 74 ]. These interconnections are involved in episodic memory, reasoning, and executive functions such as working memory and cognitive flexibility [ 75 , 76 ]. The classification indicates that, as the brain function declines, in the early stage of cognitive impairment, the tightness of the cerebellum and some brain regions will increase. This may be a compensatory regulation mechanism, which only appears in certain brain regions. However, ML studies have focused on only including clinical, cognitive and structural neuroimaging variables. A previous systematic review highlighted the need to explore additional types of data to improve the performance of ML models [ 77 ]. The results of the present study suggest that including cerebello-cerebral FC in ML models has the potential to increase the diagnostic utility and accuracy. These phenomena were hereby proved to be among the most important predictors, along with features of CT and FC from the right cerebellum to the left frontal lobe. Limitations The proposed method effectively improves the accuracy of preclinical diagnosis of AD; however, some limitations need to be highlighted. Future work will focus on the following improvements. To improve the effectiveness of the proposed approach, the data set will be expanded in the following aspects: First, extending the longitudinal data set to better understand the progression of MCI and inclusion of multi-modal data, such as gene and PET data, to investigate different insights into AD characteristics. Second, the parameter acquisition process will be optimized to achieve a higher diagnostic accuracy. Conclusion In conclusion, structure MRI measure cortical thickness (CT) could improve Alzheimer’s disease (AD) diagnosis. Furthermore, the combination of CT and cerebello-cerebral functional connectivity features shows great potential for the early identification of mild cognitive impairment. These results demonstrate the effectiveness of machine learning in preclinical AD diagnosis, based on the resting-state brain functional connectomics with complex and high-dimensional voxel-wise spatio-temporal patterns. This framework provides a new and intuitive approach to fully exploit deeply embedded diagnostic features from sMRI and resting-state functional MRI data to better individualize the diagnosis of various neurological diseases. Nevertheless, more studies are needed to further investigate the impact of cerebellar neuromodulation on the diagnosis of AD. Declarations Acknowledgements The authors gratefully acknowledge the support of Department of Neurology, Nanjing Brain Hospital Nanjing Medical University. Authors’ contributions Jingping Shi, Kuiying Yin, Qun Yao and Liangcheng Qu designed the study. Qun Yao, Liangcheng Qu, Bo Song, Xixi Wang, Tong Wang, Minjie Tian, Wenying Ma, Donglin Zhu, Xingjian Lin recruited patients and collected samples. Wenying Ma, Bo Shen, Zonghong Li performed MRI scans and collected data. Qun Yao, Liangcheng Qu, Bo Song, Xixi Wang, Tong Wang performed the experiments and analyzed the data. Qun Yao and Liangcheng Qu interpreted the data and wrote the manuscript. Funding This work was supported by Nanjing Medical science and Technology Development key project（No. ZKX21034）.  Availability of data and materials The datasets analyzed for the current study are not publicly available but are available from the corresponding author on reasonable request.  Ethical Approval and Consent to participate Individuals provided informed consent prior to participation at each measurement occasion. The Medical Research Ethical Committee of Nanjing Brain Hospital approved the data collection procedures for this study (2022-KY026-01).  Consent for publication Not applicable  Competing interests The authors declare that they have no competing interests.. References Jack, C.R., Jr., D.S. Knopman, W.J. Jagust, R.C. Petersen, M.W. Weiner, P.S. Aisen, et al., Tracking pathophysiological processes in Alzheimer's disease: an updated hypothetical model of dynamic biomarkers. Lancet Neurol, 2013. 12 (2): p. 207-16.http://dx.doi.org/10.1016/s1474-4422(12)70291-0 Hampel, H. and S. Lista, Dementia: The rising global tide of cognitive impairment. Nat Rev Neurol, 2016. 12 (3): p. 131-2.http://dx.doi.org/10.1038/nrneurol.2015.250 Allison, J.R., R.C. Rivers, J.C. Christodoulou, M. Vendruscolo, and C.M. Dobson, A relationship between the transient structure in the monomeric state and the aggregation propensities of α-synuclein and β-synuclein. Biochemistry, 2014. 53 (46): p. 7170-83.http://dx.doi.org/10.1021/bi5009326 Bischkopf, J., A. Busse, and M.C. Angermeyer, Mild cognitive impairment--a review of prevalence, incidence and outcome according to current approaches. Acta Psychiatr Scand, 2002. 106 (6): p. 403-14.http://dx.doi.org/10.1034/j.1600-0447.2002.01417.x Braak, H. and E. Braak, Development of Alzheimer-related neurofibrillary changes in the neocortex inversely recapitulates cortical myelogenesis. Acta Neuropathol, 1996. 92 (2): p. 197-201.http://dx.doi.org/10.1007/s004010050508 Delacourte, A., J.P. David, N. Sergeant, L. Buée, A. Wattez, P. Vermersch, et al., The biochemical pathway of neurofibrillary degeneration in aging and Alzheimer's disease. Neurology, 1999. 52 (6): p. 1158-65.http://dx.doi.org/10.1212/wnl.52.6.1158 Morris, J.C., M. Storandt, D.W. McKeel, Jr., E.H. Rubin, J.L. Price, E.A. Grant, and L. Berg, Cerebral amyloid deposition and diffuse plaques in "normal" aging: Evidence for presymptomatic and very mild Alzheimer's disease. Neurology, 1996. 46 (3): p. 707-19.http://dx.doi.org/10.1212/wnl.46.3.707 Serrano-Pozo, A., M.P. Frosch, E. Masliah, and B.T. Hyman, Neuropathological alterations in Alzheimer disease. Cold Spring Harb Perspect Med, 2011. 1 (1): p. a006189.http://dx.doi.org/10.1101/cshperspect.a006189 Farràs-Permanyer, L., J. Guàrdia-Olmos, and M. Peró-Cebollero, Mild cognitive impairment and fMRI studies of brain functional connectivity: the state of the art. Front Psychol, 2015. 6 : p. 1095.http://dx.doi.org/10.3389/fpsyg.2015.01095 Liu, Z., Y. Zhang, H. Yan, L. Bai, R. Dai, W. Wei, et al., Altered topological patterns of brain networks in mild cognitive impairment and Alzheimer's disease: a resting-state fMRI study. Psychiatry Res, 2012. 202 (2): p. 118-25.http://dx.doi.org/10.1016/j.pscychresns.2012.03.002 Zhang, T., Z. Zhao, C. Zhang, J. Zhang, Z. Jin, and L. Li, Classification of Early and Late Mild Cognitive Impairment Using Functional Brain Network of Resting-State fMRI. Front Psychiatry, 2019. 10 : p. 572.http://dx.doi.org/10.3389/fpsyt.2019.00572 Cai, S., T. Chong, Y. Peng, W. Shen, J. Li, K.M. von Deneen, and L. Huang, Altered functional brain networks in amnestic mild cognitive impairment: a resting-state fMRI study. Brain Imaging Behav, 2017. 11 (3): p. 619-631.http://dx.doi.org/10.1007/s11682-016-9539-0 Fan, Y., N. Batmanghelich, C.M. Clark, and C. Davatzikos, Spatial patterns of brain atrophy in MCI patients, identified via high-dimensional pattern classification, predict subsequent cognitive decline. Neuroimage, 2008. 39 (4): p. 1731-43.http://dx.doi.org/10.1016/j.neuroimage.2007.10.031 Chupin, M., E. Gérardin, R. Cuingnet, C. Boutet, L. Lemieux, S. Lehéricy, et al., Fully automatic hippocampus segmentation and classification in Alzheimer's disease and mild cognitive impairment applied on data from ADNI. Hippocampus, 2009. 19 (6): p. 579-87.http://dx.doi.org/10.1002/hipo.20626 Querbes, O., F. Aubry, J. Pariente, J.A. Lotterie, J.F. Démonet, V. Duret, et al., Early diagnosis of Alzheimer's disease using cortical thickness: impact of cognitive reserve. Brain, 2009. 132 (Pt 8): p. 2036-47.http://dx.doi.org/10.1093/brain/awp105 Westman, E., J.S. Muehlboeck, and A. Simmons, Combining MRI and CSF measures for classification of Alzheimer's disease and prediction of mild cognitive impairment conversion. Neuroimage, 2012. 62 (1): p. 229-38.http://dx.doi.org/10.1016/j.neuroimage.2012.04.056 Cho, Y., J.K. Seong, Y. Jeong, and S.Y. Shin, Individual subject classification for Alzheimer's disease based on incremental learning using a spatial frequency representation of cortical thickness data. Neuroimage, 2012. 59 (3): p. 2217-30.http://dx.doi.org/10.1016/j.neuroimage.2011.09.085 Eskildsen, S.F., P. Coupé, D. García-Lorenzo, V. Fonov, J.C. Pruessner, and D.L. Collins, Prediction of Alzheimer's disease in subjects with mild cognitive impairment from the ADNI cohort using patterns of cortical thinning. Neuroimage, 2013. 65 : p. 511-21.http://dx.doi.org/10.1016/j.neuroimage.2012.09.058 Wang, T., S. Xiao, Y. Liu, Z. Lin, N. Su, X. Li, et al., The efficacy of plasma biomarkers in early diagnosis of Alzheimer's disease. Int J Geriatr Psychiatry, 2014. 29 (7): p. 713-9.http://dx.doi.org/10.1002/gps.4053 Magnin, B., L. Mesrob, S. Kinkingnéhun, M. Pélégrini-Issac, O. Colliot, M. Sarazin, et al., Support vector machine-based classification of Alzheimer's disease from whole-brain anatomical MRI. Neuroradiology, 2009. 51 (2): p. 73-83.http://dx.doi.org/10.1007/s00234-008-0463-x Klöppel, S., C.M. Stonnington, C. Chu, B. Draganski, R.I. Scahill, J.D. Rohrer, et al., Automatic classification of MR scans in Alzheimer's disease. Brain, 2008. 131 (Pt 3): p. 681-9.http://dx.doi.org/10.1093/brain/awm319 Zhang, Y., Z. Dong, P. Phillips, S. Wang, G. Ji, J. Yang, and T.F. Yuan, Detection of subjects and brain regions related to Alzheimer's disease using 3D MRI scans based on eigenbrain and machine learning. Front Comput Neurosci, 2015. 9 : p. 66.http://dx.doi.org/10.3389/fncom.2015.00066 Zhong, X., J. Li, H. Dou, S. Deng, G. Wang, Y. Jiang, et al., Fuzzy nonlinear proximal support vector machine for land extraction based on remote sensing image. PLoS One, 2013. 8 (7): p. e69434.http://dx.doi.org/10.1371/journal.pone.0069434 Rapoport, M., R. van Reekum, and H. Mayberg, The role of the cerebellum in cognition and behavior: a selective review. J Neuropsychiatry Clin Neurosci, 2000. 12 (2): p. 193-8.http://dx.doi.org/10.1176/jnp.12.2.193 Schmahmann, J.D. and D. Caplan, Cognition, emotion and the cerebellum. Brain, 2006. 129 (Pt 2): p. 290-2.http://dx.doi.org/10.1093/brain/awh729 Balsters, J.H., C.D. Whelan, I.H. Robertson, and N. Ramnani, Cerebellum and cognition: evidence for the encoding of higher order rules. Cereb Cortex, 2013. 23 (6): p. 1433-43.http://dx.doi.org/10.1093/cercor/bhs127 Sereno, M.I., J. Diedrichsen, M. Tachrount, G. Testa-Silva, H. d'Arceuil, and C. De Zeeuw, The human cerebellum has almost 80% of the surface area of the neocortex. Proc Natl Acad Sci U S A, 2020. 117 (32): p. 19538-19543.http://dx.doi.org/10.1073/pnas.2002896117 Marvel, C.L. and J.E. Desmond, Functional topography of the cerebellum in verbal working memory. Neuropsychol Rev, 2010. 20 (3): p. 271-9.http://dx.doi.org/10.1007/s11065-010-9137-7 Schmahmann, J.D., The role of the cerebellum in cognition and emotion: personal reflections since 1982 on the dysmetria of thought hypothesis, and its historical evolution from theory to therapy. Neuropsychol Rev, 2010. 20 (3): p. 236-60.http://dx.doi.org/10.1007/s11065-010-9142-x De Smet, H.J., P. Paquier, J. Verhoeven, and P. Mariën, The cerebellum: its role in language and related cognitive and affective functions. Brain Lang, 2013. 127 (3): p. 334-42.http://dx.doi.org/10.1016/j.bandl.2012.11.001 Koziol, L.F., D.E. Budding, and D. Chidekel, From movement to thought: executive function, embodied cognition, and the cerebellum. Cerebellum, 2012. 11 (2): p. 505-25.http://dx.doi.org/10.1007/s12311-011-0321-y Adamaszek, M., F. D'Agata, R. Ferrucci, C. Habas, S. Keulen, K.C. Kirkby, et al., Consensus Paper: Cerebellum and Emotion. Cerebellum, 2017. 16 (2): p. 552-576.http://dx.doi.org/10.1007/s12311-016-0815-8 Raymond, J.L. and J.F. Medina, Computational Principles of Supervised Learning in the Cerebellum. Annu Rev Neurosci, 2018. 41 : p. 233-253.http://dx.doi.org/10.1146/annurev-neuro-080317-061948 Wang, D., Q. Yao, M. Yu, C. Xiao, L. Fan, X. Lin, et al., Topological Disruption of Structural Brain Networks in Patients With Cognitive Impairment Following Cerebellar Infarction. Front Neurol, 2019. 10 : p. 759.http://dx.doi.org/10.3389/fneur.2019.00759 Tang, F., D. Zhu, W. Ma, Q. Yao, Q. Li, and J. Shi, Differences Changes in Cerebellar Functional Connectivity Between Mild Cognitive Impairment and Alzheimer's Disease: A Seed-Based Approach. Front Neurol, 2021. 12 : p. 645171.http://dx.doi.org/10.3389/fneur.2021.645171 Tabatabaei-Jafari, H., E. Walsh, M.E. Shaw, and N. Cherbuin, The cerebellum shrinks faster than normal ageing in Alzheimer's disease but not in mild cognitive impairment. Hum Brain Mapp, 2017. 38 (6): p. 3141-3150.http://dx.doi.org/10.1002/hbm.23580 Liang, K.J. and E.S. Carlson, Resistance, vulnerability and resilience: A review of the cognitive cerebellum in aging and neurodegenerative diseases. Neurobiol Learn Mem, 2020. 170 : p. 106981.http://dx.doi.org/10.1016/j.nlm.2019.01.004 Markesbery, W.R., Neuropathologic alterations in mild cognitive impairment: a review. J Alzheimers Dis, 2010. 19 (1): p. 221-8.http://dx.doi.org/10.3233/jad-2010-1220 Jack, C.R., Jr., D.A. Bennett, K. Blennow, M.C. Carrillo, B. Dunn, S.B. Haeberlein, et al., NIA-AA Research Framework: Toward a biological definition of Alzheimer's disease. Alzheimers Dement, 2018. 14 (4): p. 535-562.http://dx.doi.org/10.1016/j.jalz.2018.02.018 Morris, J.C., The Clinical Dementia Rating (CDR): current version and scoring rules. Neurology, 1993. 43 (11): p. 2412-4.http://dx.doi.org/10.1212/wnl.43.11.2412-a Tombaugh, T.N. and N.J. McIntyre, The mini-mental state examination: a comprehensive review. J Am Geriatr Soc, 1992. 40 (9): p. 922-35.http://dx.doi.org/10.1111/j.1532-5415.1992.tb01992.x Horton, D.K., L.S. Hynan, L.H. Lacritz, H.C. Rossetti, M.F. Weiner, and C.M. Cullum, An Abbreviated Montreal Cognitive Assessment (MoCA) for Dementia Screening. Clin Neuropsychol, 2015. 29 (4): p. 413-25.http://dx.doi.org/10.1080/13854046.2015.1043349 Cano, S.J., H.B. Posner, M.L. Moline, S.W. Hurt, J. Swartz, T. Hsu, and J.C. Hobart, The ADAS-cog in Alzheimer's disease clinical trials: psychometric evaluation of the sum and its parts. J Neurol Neurosurg Psychiatry, 2010. 81 (12): p. 1363-8.http://dx.doi.org/10.1136/jnnp.2009.204008 Vakil, E. and H. Blachstein, Rey Auditory-Verbal Learning Test: structure analysis. J Clin Psychol, 1993. 49 (6): p. 883-90.http://dx.doi.org/10.1002/1097-4679(199311)49:63.0.co;2-6 Mainland, B.J., S. Amodeo, and K.I. Shulman, Multiple clock drawing scoring systems: simpler is better. Int J Geriatr Psychiatry, 2014. 29 (2): p. 127-36.http://dx.doi.org/10.1002/gps.3992 Leite, K.S., E.C. Miotto, R. Nitrini, and M.S. Yassuda, Boston Naming Test (BNT) original, Brazilian adapted version and short forms: normative data for illiterate and low-educated older adults. Int Psychogeriatr, 2017. 29 (5): p. 825-833.http://dx.doi.org/10.1017/s1041610216001952 García-Herranz, S., M.C. Díaz-Mardomingo, C. Venero, and H. Peraita, Accuracy of verbal fluency tests in the discrimination of mild cognitive impairment and probable Alzheimer's disease in older Spanish monolingual individuals. Neuropsychol Dev Cogn B Aging Neuropsychol Cogn, 2020. 27 (6): p. 826-840.http://dx.doi.org/10.1080/13825585.2019.1698710 Hasegawa, C. and N. Hirono, [Trail Making Test (TMT) A and B]. Nihon Rinsho, 2011. 69 Suppl 8 : p. 423-7 Costa, S.L., S. Schwizer Ashkenazi, L.B. Strober, N.D. Chiaravalloti, and E. Vakil, The adapted symbol digit modalities test: Examining the impact of response modality. NeuroRehabilitation, 2021. 49 (2): p. 215-220.http://dx.doi.org/10.3233/nre-218021 Bowden, S.C., V.M. Petrauskas, F.J. Bardenhagen, C.E. Meade, and L.C. Simpson, Exploring the dimensionality of digit span. Assessment, 2013. 20 (2): p. 188-98.http://dx.doi.org/10.1177/1073191112457016 Bagby, R.M., A.G. Ryder, D.R. Schuller, and M.B. Marshall, The Hamilton Depression Rating Scale: has the gold standard become a lead weight? Am J Psychiatry, 2004. 161 (12): p. 2163-77.http://dx.doi.org/10.1176/appi.ajp.161.12.2163 Thompson, E., Hamilton Rating Scale for Anxiety (HAM-A). Occup Med (Lond), 2015. 65 (7): p. 601.http://dx.doi.org/10.1093/occmed/kqv054 Buysse, D.J., C.F. Reynolds, 3rd, T.H. Monk, C.C. Hoch, A.L. Yeager, and D.J. Kupfer, Quantification of subjective sleep quality in healthy elderly men and women using the Pittsburgh Sleep Quality Index (PSQI). Sleep, 1991. 14 (4): p. 331-8 Fischl, B., D.H. Salat, E. Busa, M. Albert, M. Dieterich, C. Haselgrove, et al., Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron, 2002. 33 (3): p. 341-55.http://dx.doi.org/10.1016/s0896-6273(02)00569-x Desikan, R.S., F. Ségonne, B. Fischl, B.T. Quinn, B.C. Dickerson, D. Blacker, et al., An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage, 2006. 31 (3): p. 968-80.http://dx.doi.org/10.1016/j.neuroimage.2006.01.021 Yan, C.G., X.D. Wang, X.N. Zuo, and Y.F. Zang, DPABI: Data Processing & Analysis for (Resting-State) Brain Imaging. Neuroinformatics, 2016. 14 (3): p. 339-51.http://dx.doi.org/10.1007/s12021-016-9299-4 Tzourio-Mazoyer, N., B. Landeau, D. Papathanassiou, F. Crivello, O. Etard, N. Delcroix, et al., Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage, 2002. 15 (1): p. 273-89.http://dx.doi.org/10.1006/nimg.2001.0978 Rajeswari., S. and J.K. Theiva, Support Vector Machine Classification For MRI Images. International Journal of Electronics & Computer Science Engineering, 2012. 1 (3) Skolariki, K., G.M. Terrera, and S. Danso, Multivariate Data Analysis and Machine Learning for Prediction of MCI-to-AD Conversion. Adv Exp Med Biol, 2020. 1194 : p. 81-103.http://dx.doi.org/10.1007/978-3-030-32622-7_8 Bron, E.E., S. Klein, J.M. Papma, L.C. Jiskoot, V. Venkatraghavan, J. Linders, et al., Cross-cohort generalizability of deep and conventional machine learning for MRI-based diagnosis and prediction of Alzheimer's disease. Neuroimage Clin, 2021. 31 : p. 102712.http://dx.doi.org/10.1016/j.nicl.2021.102712 Basaia, S., F. Agosta, L. Wagner, E. Canu, G. Magnani, R. Santangelo, and M. Filippi, Automated classification of Alzheimer's disease and mild cognitive impairment using a single MRI and deep neural networks. Neuroimage Clin, 2019. 21 : p. 101645.http://dx.doi.org/10.1016/j.nicl.2018.101645 Popuri, K., D. Ma, L. Wang, and M.F. Beg, Using machine learning to quantify structural MRI neurodegeneration patterns of Alzheimer's disease into dementia score: Independent validation on 8,834 images from ADNI, AIBL, OASIS, and MIRIAD databases. Hum Brain Mapp, 2020. 41 (14): p. 4127-4147.http://dx.doi.org/10.1002/hbm.25115 Zhou, Y., Z. Song, X. Han, H. Li, and X. Tang, Prediction of Alzheimer's Disease Progression Based on Magnetic Resonance Imaging. ACS Chem Neurosci, 2021. 12 (22): p. 4209-4223.http://dx.doi.org/10.1021/acschemneuro.1c00472 Chagué, P., B. Marro, S. Fadili, M. Houot, A. Morin, J. Samper-González, et al., Radiological classification of dementia from anatomical MRI assisted by machine learning-derived maps. J Neuroradiol, 2021. 48 (6): p. 412-418.http://dx.doi.org/10.1016/j.neurad.2020.04.004 Korolev, I.O., L.L. Symonds, and A.C. Bozoki, Predicting Progression from Mild Cognitive Impairment to Alzheimer's Dementia Using Clinical, MRI, and Plasma Biomarkers via Probabilistic Pattern Classification. PLoS One, 2016. 11 (2): p. e0138866.http://dx.doi.org/10.1371/journal.pone.0138866 Zhang, D. and D. Shen, Predicting future clinical changes of MCI patients using longitudinal and multimodal biomarkers. PLoS One, 2012. 7 (3): p. e33182.http://dx.doi.org/10.1371/journal.pone.0033182 Wei, R., C. Li, N. Fogelson, and L. Li, Prediction of Conversion from Mild Cognitive Impairment to Alzheimer's Disease Using MRI and Structural Network Features. Front Aging Neurosci, 2016. 8 : p. 76.http://dx.doi.org/10.3389/fnagi.2016.00076 Suk, H.I., S.W. Lee, and D. Shen, Subclass-based multi-task learning for Alzheimer's disease diagnosis. Front Aging Neurosci, 2014. 6 : p. 168.http://dx.doi.org/10.3389/fnagi.2014.00168 Moradi, E., A. Pepe, C. Gaser, H. Huttunen, and J. Tohka, Machine learning framework for early MRI-based Alzheimer's conversion prediction in MCI subjects. Neuroimage, 2015. 104 : p. 398-412.http://dx.doi.org/10.1016/j.neuroimage.2014.10.002 Ardekani, B.A., E. Bermudez, A.M. Mubeen, and A.H. Bachman, Prediction of Incipient Alzheimer's Disease Dementia in Patients with Mild Cognitive Impairment. J Alzheimers Dis, 2017. 55 (1): p. 269-281.http://dx.doi.org/10.3233/jad-160594 Suk, H.I., S.W. Lee, and D. Shen, Deep sparse multi-task learning for feature selection in Alzheimer's disease diagnosis. Brain Struct Funct, 2016. 221 (5): p. 2569-87.http://dx.doi.org/10.1007/s00429-015-1059-y Zheng, W., Z. Yao, Y. Li, Y. Zhang, B. Hu, and D. Wu, Brain Connectivity Based Prediction of Alzheimer's Disease in Patients With Mild Cognitive Impairment Based on Multi-Modal Images. Front Hum Neurosci, 2019. 13 : p. 399.http://dx.doi.org/10.3389/fnhum.2019.00399 Yao, Q., F. Tang, Y. Wang, Y. Yan, L. Dong, T. Wang, et al., Effect of cerebellum stimulation on cognitive recovery in patients with Alzheimer disease: A randomized clinical trial. Brain Stimul, 2022. 15 (4): p. 910-920.http://dx.doi.org/10.1016/j.brs.2022.06.004 van den Heuvel, M.P. and O. Sporns, Rich-club organization of the human connectome. J Neurosci, 2011. 31 (44): p. 15775-86.http://dx.doi.org/10.1523/jneurosci.3539-11.2011 Blumenfeld, R.S., C.M. Parks, A.P. Yonelinas, and C. Ranganath, Putting the pieces together: the role of dorsolateral prefrontal cortex in relational memory encoding. J Cogn Neurosci, 2011. 23 (1): p. 257-65.http://dx.doi.org/10.1162/jocn.2010.21459 Blumenfeld, R.S. and C. Ranganath, Prefrontal cortex and long-term memory encoding: an integrative review of findings from neuropsychology and neuroimaging. Neuroscientist, 2007. 13 (3): p. 280-91.http://dx.doi.org/10.1177/1073858407299290 Pellegrini, E., L. Ballerini, M. Hernandez, F.M. Chappell, V. González-Castro, D. Anblagan, et al., Machine learning of neuroimaging for assisted diagnosis of cognitive impairment and dementia: A systematic review. Alzheimers Dement (Amst), 2018. 10 : p. 519-535.http://dx.doi.org/10.1016/j.dadm.2018.07.004 Additional Declarations No competing interests reported. Supplementary Files Supplementalfiles.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-2663342","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":181691760,"identity":"a7757729-8ec9-4b8d-af85-beaeda5843a7","order_by":0,"name":"Qun Yao","email":"","orcid":"","institution":"Affiliated Brain Hospital of Nanjing Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Qun","middleName":"","lastName":"Yao","suffix":""},{"id":181691761,"identity":"b203c8b1-d57f-43bf-ae9b-df027772ca25","order_by":1,"name":"Liangcheng Qu","email":"","orcid":"","institution":"Nanjing Research Institute of Electronic Technology","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Liangcheng","middleName":"","lastName":"Qu","suffix":""},{"id":181691764,"identity":"0c247027-339e-4504-88d0-b58ecfd791a0","order_by":2,"name":"Bo Song","email":"","orcid":"","institution":"Affiliated Brain Hospital of Nanjing Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Bo","middleName":"","lastName":"Song","suffix":""},{"id":181691765,"identity":"4eddd3f0-edb5-4f7c-841a-f1e4bce110c8","order_by":3,"name":"Xixi Wang","email":"","orcid":"","institution":"Affiliated Brain Hospital of Nanjing Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Xixi","middleName":"","lastName":"Wang","suffix":""},{"id":181691768,"identity":"46b48e31-c727-48b3-be84-8c68f5e133b1","order_by":4,"name":"Tong Wang","email":"","orcid":"","institution":"Affiliated Brain Hospital of Nanjing Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Tong","middleName":"","lastName":"Wang","suffix":""},{"id":181691769,"identity":"1fccadb5-4811-42eb-9a67-f77029c88500","order_by":5,"name":"Wenying Ma","email":"","orcid":"","institution":"Affiliated Brain Hospital of Nanjing Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Wenying","middleName":"","lastName":"Ma","suffix":""},{"id":181691772,"identity":"cd9f5981-b601-42cd-83a0-de3336ccf4dc","order_by":6,"name":"Minjie Tian","email":"","orcid":"","institution":"Affiliated Brain Hospital of Nanjing Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Minjie","middleName":"","lastName":"Tian","suffix":""},{"id":181691773,"identity":"ab91940e-5bbc-4144-8ca4-1cae46c6803f","order_by":7,"name":"Bo Shen","email":"","orcid":"","institution":"Affiliated Brain Hospital of Nanjing Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Bo","middleName":"","lastName":"Shen","suffix":""},{"id":181691774,"identity":"4d1c4b3d-d2d4-4eb2-863c-28bf7d7e8bfa","order_by":8,"name":"Xingyang Qi","email":"","orcid":"","institution":"Affiliated Brain Hospital of Nanjing Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Xingyang","middleName":"","lastName":"Qi","suffix":""},{"id":181691775,"identity":"67b3a07c-8e4c-4228-99ff-00a649cf029c","order_by":9,"name":"Donglin Zhu","email":"","orcid":"","institution":"Affiliated Brain Hospital of Nanjing Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Donglin","middleName":"","lastName":"Zhu","suffix":""},{"id":181691776,"identity":"6282d40a-11e0-4be2-93bb-9cfe6e6fd97b","order_by":10,"name":"Xingjian Lin","email":"","orcid":"","institution":"Affiliated Brain Hospital of Nanjing Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Xingjian","middleName":"","lastName":"Lin","suffix":""},{"id":181691777,"identity":"3474f455-524f-41f6-962c-0818ce6fca34","order_by":11,"name":"Zonghong Li","email":"","orcid":"","institution":"Affiliated Brain Hospital of Nanjing Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Zonghong","middleName":"","lastName":"Li","suffix":""},{"id":181691778,"identity":"321f9df9-f932-48cd-aa5e-16fdd3697118","order_by":12,"name":"Jingping Shi","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAw0lEQVRIiWNgGAWjYJCCD4wNEswM7I2NDz8QqYNxBlgLz+FmYwkStAApifQ2AR5i1JtLHz7Y+HOHBbu55MM2BgkGOzndBgJaLPvSEpt5z0gwW85ObHtQwJBsbHaAgBaDMzzmjxnbJJgNbie2G0gwHEjcRoQWw8afIC03D7ZJ8BCrpYEXpOUGI5FaLHvYgH4BarHsSQQGsgERfjHnYQaGWFtdsjn78YcPP1TYyRH2PpRONkDhEqPFjhjFo2AUjIJRMEIBAHnuP9yzU20VAAAAAElFTkSuQmCC","orcid":"","institution":"Affiliated Brain Hospital of Nanjing Medical University","correspondingAuthor":true,"submittingAuthor":false,"prefix":"","firstName":"Jingping","middleName":"","lastName":"Shi","suffix":""},{"id":181691779,"identity":"c700bb99-b662-49ed-8a0c-1ae3f01f9237","order_by":13,"name":"Kuiying Yin","email":"","orcid":"","institution":"Nanjing Research Institute of Electronic Technology","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Kuiying","middleName":"","lastName":"Yin","suffix":""}],"badges":[],"createdAt":"2023-03-07 03:59:15","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-2663342/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-2663342/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":34186810,"identity":"e89285a2-ea12-4176-8982-d1a254e91af7","added_by":"auto","created_at":"2023-03-13 20:05:31","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":351921,"visible":true,"origin":"","legend":"\u003cp\u003eAn illustration of the proposed machinelearning framework based on cortical thickness and functional connectivity of cerebello-cerebral networks extracted from structure magnetic resonance imaging (sMRI) and resting-state functional magnetic resonance imaging (rs-fMRI) for brain disease diagnosis.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-2663342/v1/eade113feb191da29ba57ae8.png"},{"id":34186958,"identity":"58adf40b-7214-431f-bdc4-bcd4332cc1f5","added_by":"auto","created_at":"2023-03-13 20:06:11","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":662628,"visible":true,"origin":"","legend":"\u003cp\u003eReceiver operating characteristic (ROC) curves for discriminating Alzheimer’s disease (AD), mild cognitive impairment (MCI), and healthy controls (HC). ROC curves are plotted for the eight biomarkers separately. (A) ROC curves for AD vs. HC classification using cortical thickness features; (B) for MCI vs. HC classification using cortical thickness features; (C) for MCI vs. AD classification using cortical thickness features; (D) for AD vs. HC classification using cortical thickness and cerebello-cerebral FC features; (E) for MCI vs. HC classification using cortical thickness and cerebello-cerebral FC features; and (F) for MCI vs. AD classification using cortical thickness and cerebello-cerebral FC features with SVM classifiers.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-2663342/v1/79a4ffe0383a08724ca8e9c1.png"},{"id":34186496,"identity":"0b3ed740-bea6-41ad-b836-fa7feffc5885","added_by":"auto","created_at":"2023-03-13 20:04:48","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":272186,"visible":true,"origin":"","legend":"\u003cp\u003eHistogram displaying the distribution of model outputs for accuracy (ACC), sensitivity (SEN), specificity(SPE) and area under the curve (AUC). (A) The ACC of cortical thickness and mixed features for classification of Alzheimer’s disease (AD), mild cognitive impairment (MCI), and healthy controls (HC); (B) the SEN of cortical thickness and mixed featuresfor classification of AD, MCI and HC; (C) the SPE of cortical thickness and mixed features for classification of AD, MCI and HC; and(D) the AUC of cortical thickness and mixed features for classification of AD, MCI and HC.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-2663342/v1/66883af502e82bfd735aae9a.png"},{"id":34828595,"identity":"89819b73-ddf8-4d01-97cd-95b57c320b88","added_by":"auto","created_at":"2023-03-26 17:14:31","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1598511,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-2663342/v1/6e7e27f0-0829-470b-9278-480b633dab69.pdf"},{"id":34186772,"identity":"8a40e934-9d5d-4a82-95c4-842955541ac3","added_by":"auto","created_at":"2023-03-13 20:05:25","extension":"docx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":25518,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementalfiles.docx","url":"https://assets-eu.researchsquare.com/files/rs-2663342/v1/66f062adfcd42bf0f6a2598a.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Machine learning of cerebello-cerebral functional networks for mild cognitive impairment detection","fulltext":[{"header":"Background","content":"\u003cp\u003eAlzheimer\u0026rsquo;s disease (AD),\u0026nbsp;the most frequent cause of dementia, is characterized\u0026nbsp;by abnormal buildup of neurofibrillary tangles and amyloid plaques in the brain, which affects memory, thinking, and behavior\u0026nbsp;[1].\u0026nbsp;AD is an irreversible and incurable disease, therefore, early detection, such as mild cognitive impairment (MCI) during its preclinical stage, is critical for implementing viable interventions to delay its progression\u0026nbsp;[2].\u003c/p\u003e\n\u003cp\u003eAccording to previous studies, a significant proportion of patients with MCI (approximately 10-15% from referral sources such as memory clinics and AD centers) will annually develop AD\u0026nbsp;[3], while healthy control population (i.e., healthy elderly people) converts at a 1-2% rate\u0026nbsp;[4].\u0026nbsp;The neuropathological progression in AD can be observed several years before clinical symptoms appear\u0026nbsp;[5-8]. Furthermore, its progression is slow and insidious, making the preclinical stage last for several years, even starting decades before the first symptoms of dementia appear. Identification trials targeting individuals in the preclinical stage of AD may be the most effective approach. Based on this reasoning, MCI patients are an ideal population to explore neurophysiological profiles that predict who will develop dementia.\u0026nbsp;In recent years, increasing efforts have shifted the frontline to possible detection in the early stage of MCI, earning more valuable time to lower the high conversion rate.\u0026nbsp;A crucial point is whether the pathophysiology of AD would be detected long before the actual diagnosis of the disease. Consequently, the identification of preclinical AD when severe brain damage associated with AD, such as extensive brain atrophy, has not appeared yet, could significantly improve the benefits of new drugs and vaccines.\u003c/p\u003e\n\u003cp\u003eResting-state functional magnetic resonance imaging (rs-fMRI) has emerged as a popular modality to investigate the changes of early brain function in patients with MCI.\u0026nbsp;As few visible structural brain changes are present in magnetic resonance imaging (MRI) of MCI patients, rs-fMRI is an \u003cem\u003ein vivo\u003c/em\u003e functional imaging technique, which measures blood oxygen level-dependent (BOLD) signals when scanning subjects at natural rest without any explicit task involvement\u0026nbsp;[9].\u0026nbsp;Functional connectivity (FC) between distant brain regions can be measured based on their time-synchronized rs-fMRI BOLD signals, yielding brain \u0026ldquo;functional connectomics\u0026rdquo; or whole-brain functional networks\u0026nbsp;[9].\u0026nbsp;Previous research has demonstrated\u0026nbsp;that\u0026nbsp;abnormal\u0026nbsp;functional connections and brain networks\u0026nbsp;between brain areas\u0026nbsp;are discovered as early as the MCI stage\u0026nbsp;[10-12]. Furthermore,\u0026nbsp;the minority of pioneering research has focused on the\u0026nbsp;brain networks\u0026nbsp;changes in MCI subjects by using group-level statistic comparisons based on simple regional pair-wise FCs or network properties.\u0026nbsp;Rapid development of neuroimaging techniques has made the integration of large-scale, high-dimensional multimodal neuroimaging data challenging. Subsequently, there has been a surge in interest in computer-aided machine learning (ML) methodologies for integrative analysis.\u0026nbsp;Over the past few years, several MRI biomarkers have been proposed for the classification of AD patients at various disease stages\u0026nbsp;[13-19].\u0026nbsp;Most of these studies have used simple but very effective classifiers, such as support vector machine (SVM)\u0026nbsp;[20-22],\u0026nbsp;which can be mathematically simplified using optimization techniques\u0026nbsp;[23].\u0026nbsp;Nevertheless, despite many efforts, the identification of efficient specific AD biomarkers for early diagnosis and disease progression prediction remains a challenging task and requires more research.\u003c/p\u003e\n\u003cp\u003ePrevious research has concentrated on the cerebral region, with little emphasis dedicated to the function of the cerebellum in the cognitive regulation of the AD spectrum. The function of the cerebellum in cognitive processing has recently sparked considerable interest in cognitive neuroscience\u0026nbsp;[24-26]. It was revealed that the surface\u0026nbsp;of the cerebellum is folded more tightly than the cerebral cortex, with almost 80% of the neocortex connected to the cognitive network of the brain\u0026nbsp;[27]. Studies have suggested that the role of the cerebellum in the processing of the various cognitive\u0026nbsp;functions such as working memory\u0026nbsp;[28], emotion\u0026nbsp;[29], language\u0026nbsp;[30], and other executive functions\u0026nbsp;[31]\u0026nbsp;can be attributed to\u0026nbsp;its extensive connections to the cortical and subcortical areas\u0026nbsp;(particularly between the Crus I and Crus II, known as the lateral \u0026ldquo;limbic\u0026rdquo; cerebellar circuit)\u0026nbsp;[32].\u0026nbsp;The cerebellum acts as a key part of the adaptive regulator in cognitive processing, supervising the working brain areas, detecting cognitive patterns, changes and errors, and generating responses and distributing feedback to numerous cerebral\u0026nbsp;areas,\u0026nbsp;to increase the effectiveness of\u0026nbsp;brain activity\u0026nbsp;[33].\u0026nbsp;The previous study from the author\u0026rsquo;s group showed that the cognitive impairment after cerebellar injury was involved in multiple cognitive domains. The connection efficiency and information integration ability of the brain network were disrupted, indicating that the cerebellum could participate in the integration and regulation of the cognitive network\u0026nbsp;[34].\u0026nbsp;Additionally, in another work, it was observed that the cerebellum Crus II exhibited increased activity in the\u0026nbsp;preclinical stage of AD, reflecting a compensatory function that could mitigate early AD symptoms\u0026nbsp;[35]. Cerebellar atrophy was shown to be an essential factor in the clinical progression of AD in previous works, but mainly occurred in the late stage of the disease. However, even in the later stages, the shrinkage rate was lower than the average shrinkage in the cerebrum and was independent of the AD moderators\u0026nbsp;[36]. Therefore, the cerebellum survived the preclinical stage of AD and may be resistant to certain neurodegenerative mechanisms\u0026nbsp;[37]. These findings suggest that the cerebellum connectivity may be a potential identification biomarker in the preclinical AD stage.\u003c/p\u003e\n\u003cp\u003eIn this paper, it is postulated that the cerebellum serves as the compensatory regulation center of the brain, modulating cerebello-cerebral networks in the early stage of the disease. A novel unified framework is hereby proposed, which utilizes ML to extract deep embedded features from cerebello-cerebral functional networks, without sacrificing the rich spatial information of the images. Preclinical AD pathology has been hypothesized to be detectable by neuroimaging techniques [38]. Traditional approaches might encounter difficulties when processing highly complex spatial-temporal connectomics information, resulting in suboptimal diagnostic accuracy of MCI, partially due to more subtle alterations in brain networks in comparison of MCI vs. healthy controls (HC), and MCI vs. AD groups. Using structural and cerebellar functional features, ML can be very useful in the detection of disease-related abnormal patterns and improving accuracy of individualized diagnosis of MCI. In the present study, the recursive binary method was used to retrieve the dominant features (combined with structural and cerebellar functional features) for the SVM classification of the AD, MCI and HC groups.\u0026nbsp;\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eParticipants\u003c/h2\u003e \u003cp\u003e The present study was approved by the Medical Research Ethical Committee of Nanjing Brain Hospital in Nanjing, China. In total, 43 AD subjects, 40 MCI subjects and 45 HC subjects were recruited between June 2019 and April 2022, following application of inclusion and exclusion criteria. All subjects were right-handed. Written informed consent was obtained from all participants before enrolling in the study. Patient recruitment was performed based on the current diagnostic criteria provided by the National Institute on Aging and the Alzheimer\u0026rsquo;s Association working group (NIA-AA) in 2018 [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. The guidelines are characterized by episodic memory loss in AD pathology supported by the evidence of cerebrospinal fluid (CSF) and imaging biomarkers. Inclusion criteria for patients included age 60\u0026ndash;80 years; no deficits in vision or hearing; \u0026gt;8th -grade education; T2-weighted MRI without indication of infection, infarction or focal lesions within 12 months (assessment by two independent expert radiologists). The inclusion criteria for the HC group included no current cognitive issues; no neurological or psychiatric diseases; clinical dementia rating (CDR) score of zero [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]. Exclusion criteria included patients with other known causes of dementia (i.e., frontotemporal dementia, dementia with Lewy bodies, vascular dementia, severe depression, cerebrovascular disease, poisoning, tumors, metabolic diseases and infections); limitations in MRI scanning, such as claustrophobia or pacemaker implantation; a history of psychiatric or other neurological disorders. All patients underwent a complete clinical investigation, an extensive neuropsychological assessment that explored all cognitive domains, brain MRI scanning and CSF analysis. The eligibility of the patients for this study was confirmed by highly experienced field investigators.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eClinical and neuropsychiatric evaluation\u003c/h2\u003e \u003cp\u003eAll participants underwent comprehensive and standard neuropsychological assessments. Global cognition was evaluated using Mini-Mental State Examination (MMSE) [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e], Montreal Cognitive Assessment (MoCA) [\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e], CDR and Alzheimer\u0026rsquo;s Disease Assessment Scale\u003cem\u003e-\u003c/em\u003eCognitive section (ADAS-Cog) [\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e]. Episodic memory was assessed using the Rey Auditory-verbal Learning Test (RAVLT) [\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. Visuospatial abilities were assessed with the Clock Drawing Test (CDT) [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]. Language function was determined with the Boston Naming Test (BNT) [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e] and the Verbal Fluency Test (VFT) [\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. Executive function was assessed using parts A and B of the Trail Making Test (TMT) [\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e]. Attention was assessed using the Symbol Digit Modalities Test (SDMT) [\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e] and the Digit Span Test (DST) [\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e]. The emotional condition of the subjects was evaluated using the Hamilton Depression Scale (HAMD) [\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e] and the Hamilton Anxiety Scale (HAMA) [\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e]. Sleep quality was assessed using the Pittsburgh Sleep Quality Inventory (PSQI) [\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e]. Activities of Daily Living (ADL) were used to assess the ability of participants to care for themselves in daily life. Senior neuropsychologists validated all the scales, which were evaluated by experienced clinicians.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eCerebrospinal fluid biomarkers\u003c/h2\u003e \u003cp\u003eA lumbar puncture was performed in all patients to confirm the typical CSF profile of AD pathology with reduced amyloidβ1\u0026ndash;42 concentrations and increased phosphorylated- and total- tau levels. Amyloidβ1\u0026ndash;42, p-tau, and t-tau levels were detected using the INNOBIA AlzBio3 immunoassay kit-based reagents (Innotest, Fujirebio, Ghent, Belgium).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eMagnetic resonance imaging data collection\u003c/h2\u003e \u003cp\u003eAll scans were performed using a Siemens 3.0 T scanner (Siemens, Verio, Germany) with an 8-channel radiofrequency coil. Participants were asked to remain as still as possible, close their eyes, remain awake, and not think of anything. Three-dimensional T1 weighted images were collected in a sagittal orientation using a 3D-MPARGE sequence with the following parameters: time repetition (TR)\u0026thinsp;=\u0026thinsp;1,900 ms, echo time (TE)\u0026thinsp;=\u0026thinsp;2.48 ms, inversion time (TI)\u0026thinsp;=\u0026thinsp;900 ms, 176 sagittal slices, 1.0 mm slice thickness, gap\u0026thinsp;=\u0026thinsp;0.5 mm, matrix\u0026thinsp;=\u0026thinsp;256 \u0026times; 256, flip angle (FA)\u0026thinsp;=\u0026thinsp;9\u0026deg;, field of view (FOV)\u0026thinsp;=\u0026thinsp;256 \u0026times; 256 mm, and voxel size\u0026thinsp;=\u0026thinsp;1 \u0026times; 1 \u0026times; 1 mm\u003csup\u003e3\u003c/sup\u003e. Functional images were collected using a gradient-recalled echo-planar imaging pulse sequence with 240-time points; TE\u0026thinsp;=\u0026thinsp;30 ms; TR\u0026thinsp;=\u0026thinsp;2,000 ms; number of slices\u0026thinsp;=\u0026thinsp;36, FOV\u0026thinsp;=\u0026thinsp;220 \u0026times; 220 mm\u003csup\u003e2\u003c/sup\u003e; matrix\u0026thinsp;=\u0026thinsp;64 \u0026times; 64; flip angle\u0026thinsp;=\u0026thinsp;90\u0026deg;; 4.0 mm thickness, and gap\u0026thinsp;=\u0026thinsp;0 mm. The image acquisition for each subject required about 14 min.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eMagnetic resonance imaging processing\u003c/h2\u003e \u003cp\u003eThe structure MRI (sMRI) data were processed with the FreeSurfer Image Analysis Suite (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://surfer.nmr.mgh.harvardedu/\u003c/span\u003e\u003cspan address=\"http://surfer.nmr.mgh.harvardedu/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) which included: automatic Talairach space transformation, image intensity inhomogeneity correction, non-brain tissue removal, intensity normalization, tissue segmentation [\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e], automatic topology correction, surface deformation to generate gray/white matter boundaries, fragmentated gray matter/CSF boundary and cerebral cortex. The Desikan-Killiany atlas (34 areas per hemisphere) was used for parcellation [\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e]. The rs-fMRI data were preprocessed using the Data Processing and Analysis for Brain Imaging tool (DPABI, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.rest.restfmri.net\u003c/span\u003e\u003cspan address=\"http://www.rest.restfmri.net\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) [\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e]. The first ten volumes of the resting session were discarded for each participant, to account for factors of magnetization equilibrium and environment adaptation. The remaining images were corrected based on slice timing and motion correction (head motion\u0026thinsp;\u0026le;\u0026thinsp;2 mm, head motion angle\u0026thinsp;\u0026le;\u0026thinsp;2\u0026deg;). Following this, the functional images were co-registered with the high-resolution 3D-T1 structural images. Then, the 3D-T1 images were normalized to the Montreal Neurological Institute (MNI) space using non-linear warping, based on the Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra. After all images were spatially normalized to the T1 space, they were resampled into 3 \u0026times; 3 \u0026times; 3 mm\u003csup\u003e3\u003c/sup\u003e voxels and spatially smoothed with a Gaussian filter of 6 mm full-width at half-maximum. The fMRI data were then temporally bandpass-filtered (0.01\u0026ndash;0.08 Hz) to remove low-frequency drifts and high-frequency physiological noise. The noise covariates were regressed to further reduce the confounding artifacts of resting head movement and physiological noise (i.e., respiration and cardiac fluctuations), including the Friston 24-motion parameter model, the global mean, the white matter and the cerebrospinal fluid signals.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eFeature extraction and selection\u003c/h2\u003e \u003cp\u003eStructural and functional features from each subject were selected for subsequent feature selection (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). (1) In the structural section, there were 68 MRI cortical thickness (CT) features. (2) In the functional section, there were 180 connectivity features (i.e., FC between the bilateral cerebellum and the brain areas). Two ML experiments were performed in this study. The first experiment was a three-class classification, using only structural features, to predict individual specific diagnosis (AD, MCI, HC). The structural features were combined with the functional features in the second experiment during training to predict the diagnosis for the three groups.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eCortical thickness features\u003c/h2\u003e \u003cp\u003eThe atlas used in Desikan-Killiany template included 68 cortical regions (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). For each cortical region, CT was calculated from MRI features. The CT of each cortical vertex was calculated as the average shortest length between white and pale surfaces. For each participant, this section yielded 68 MRI features. The bilateral frontal, temporal, parietal and occipital CT features were selected to construct eight training sets, respectively.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eCerebello-cerebral network features\u003c/h2\u003e \u003cp\u003eTo define nodes of the functional brain network, the brain was segmented into 90 regions using the automatic anatomical labeling (AAL) [\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e]. The bilateral cerebellar Crus II areas were extracted as regions of interest (ROI). The FC analysis was voxel-wise performed between each ROI and the entire brain. To establish the reference time series of seed points, the voxels of each seed region from every subject were extracted and averaged. Subsequently, the correlation coefficient between the reference time series and the time series involving all other brain voxels was calculated. The correlation coefficients were transformed into z-values using the Fisher r-to-z transformation for normality enhancement. In this part, 180 bilateral cerebellum connectivity features (90\u0026times;2\u0026thinsp;=\u0026thinsp;180) were obtained for further feature selection. The CT combined with the FC between the bilateral cerebellum Crus II and the brain regions (including the bilateral frontal, bilateral parietal, bilateral temporal and bilateral occipital lobes) were selected as eight training datasets, respectively.\u003c/p\u003e \u003cp\u003eAll subdivisions within brain regions were combined into feature subsets using the recursive binary method. These feature subsets were used as input to the ML model for training and testing, to assess the performance of subsets in classification tasks.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eSVM classification\u003c/h2\u003e \u003cp\u003eThe SVM classification model was implemented in MATLAB (The Math Works, Natwick, MA) and LIBSVM (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.csie.ntu.edu.tw/cjlin/libsvm/\u003c/span\u003e\u003cspan address=\"http://www.csie.ntu.edu.tw/cjlin/libsvm/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Binary classification of data was performed in a supervised learning method by projecting low-dimensional linearly indistinguishable problems to high-dimensional to construct optimal decision classification surface [\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eLet the optimal classification surface established by the SVM during training be\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({\\omega }\\)\u003c/span\u003e\u003c/span\u003e, as in Eq.\u0026nbsp;2\u0026thinsp;\u0026minus;\u0026thinsp;1,\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Taba\" border=\"1\"\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({\\omega }^{T}x+b=0\\)\u003c/span\u003e\u003c/span\u003e,\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(2\u0026thinsp;\u0026minus;\u0026thinsp;1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"2\"\u003ewhere \u003cem\u003ex\u003c/em\u003e is the feature vector, \u003csup\u003e\u003cem\u003eT\u003c/em\u003e\u003c/sup\u003e is the transpose matrix, and \u003cem\u003eb\u003c/em\u003e is a constant. The construction of the optimal decision classification surface in SVM is based on \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\omega\\)\u003c/span\u003e\u003c/span\u003e, by maximizing the subsurface, as in Eq.\u0026nbsp;2\u0026ndash;2, with the linear function as the constraint,\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Tabb\" border=\"1\"\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\text{m}\\text{a}\\text{x} \\frac{1}{‖\\omega ‖} s.t. ,{y}_{i}({\\omega }^{T}{x}_{i}+b)\\ge 1\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(2\u0026ndash;2)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003ewhere \u003cem\u003ey\u003c/em\u003e represents the sample category, Eq.\u0026nbsp;2\u0026ndash;2 can be converted to Eq.\u0026nbsp;2\u0026ndash;3\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(.\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Tabc\" border=\"1\"\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\text{m}\\text{i}\\text{n} \\frac{1}{2}{‖\\omega ‖}^{2} s.t. ,{y}_{i}({\\omega }^{T}{x}_{i}+b)\\ge 1\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(2\u0026ndash;3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe optimization of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({\\omega }\\)\u003c/span\u003e\u003c/span\u003e is a convex quadratic programming problem with linear constraints It can be transformed to the dyadic problem domain by adding Lagrange multipliers \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({\\alpha }\\)\u003c/span\u003e\u003c/span\u003e to the linear constraint function, as in Eq.\u0026nbsp;2\u0026ndash;4.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Tabd\" border=\"1\"\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\text{L}\\left(\\omega ,b,\\alpha \\right)= \\frac{1}{2}{‖\\omega ‖}^{2}-\\sum _{i=1}^{n}{\\alpha }_{i}\\left({y}_{i}\\left({\\omega }^{T}{x}_{i}+b\\right)-1\\right)\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(2\u0026ndash;4)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eFinally, the SMO algorithm can be used to solve for the \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({\\alpha }\\)\u003c/span\u003e\u003c/span\u003e multiplier. The kernel function used in the SVM classifier was the radial basis kernel function (RBF), where the penalty parameter \u003cem\u003eC\u003c/em\u003e and the kernel bandwidth \u003cem\u003eσ\u003c/em\u003e in the kernel function ranged from [\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({4}^{-4}\\)\u003c/span\u003e\u003c/span\u003e,\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({4}^{4}\\)\u003c/span\u003e\u003c/span\u003e]. The RBF kernel was defined as follows,\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Tabe\" border=\"1\"\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\text{K}\\left({x}_{1},{x}_{2}\\right)= \\text{e}\\text{x}\\text{p}(-\\frac{‖{x}_{1}-{x}_{2}‖}{2{\\sigma }^{2}})\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(2\u0026ndash;5)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({x}_{1}\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({x}_{2}\\)\u003c/span\u003e\u003c/span\u003e are two eigenvectors and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({\\sigma }\\)\u003c/span\u003e\u003c/span\u003e is the width parameter of the RBF kernel.\u003c/p\u003e \u003cp\u003eBecause the data area linearly indiscriminable, SVM will first complete the basic computation in the low-dimensional space, and then introduce the kernel function to map the input space to a high-dimensional space to construct the optimal classification surface.\u003c/p\u003e \u003cdiv id=\"Sec12\" class=\"Section3\"\u003e \u003ch2\u003ePerformance evaluation\u003c/h2\u003e \u003cp Models were evaluated using 4-fold cross-validation. Four models were trained for each experiment, with 75% of the data randomly selected for training and 25% for testing for each one. Thus, this process produced one prediction per sample of the entire dataset. The cross-validation test for each set of feature subsets was repeated four times to assess the variability of the evaluation metrics. Therefore, the resulting evaluation metrics were max over the four cross-validation iterations. Metrics of model performance were accuracy (ACC) (percent correctly classified), sensitivity (SEN) (true positives, correctly identified) and specificity (SPE) (true negatives, correctly identified). Another effective way to evaluate classification results is the receiver operating characteristic (ROC) curve. The ROC curve is the plot of true-positive rate against false-positive rate by changing the discrimination threshold, summarizing the performance of the classifier. The ROC curve results are usually interpreted by the area under the curve (AUC).\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eAll data were tested for normality and variance consistency. Additionally, appropriate statistical tests (i.e., ANOVA, independent sample t-tests or Chi-squared tests) were performed to assess the clinical and demographical differences among the groups (AD, MCI, HC), reported in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Statistical analyses were performed with IBM SPSS 25.0 software (SPSS Inc., Chicago, Illinois, USA). A two-sided P\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was acknowledged as statistically significant. The 95% confidence intervals (CIs) of ACC, SEN, and SPE were calculated by the exact Clopper\u0026ndash;Pearson method. The DeLong method was used to calculate the CIs of AUC.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u003cem\u003eThe table contains the mean values (\u003c/em\u003e\u0026plusmn;\u0026thinsp;\u003cem\u003estandard deviation) of demographic and clinical data\u003c/em\u003e. Significant group differences were obtained using ANOVA, independent sample t-tests or Chi-squared tests. two-sided P\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered significant. Abbreviations: MMSE: Mini-Mental State Examination; MoCA: Montreal Cognitive Assessment; HAMA: Hamilton Anxiety Scale; HAMD: Hamilton Depression Scale; ADL: \u003cem\u003eActivities of Daily Living;\u003c/em\u003e PSQI: Pittsburgh Sleep Quality Inventory; RAVLT: Rey Auditory Verbal Learning Test; CDT: Clock Drawing Test; SDMT: Symbol Digit Modalities Test; DST: Digit Span Test; BNT: Boston Naming Test; VFT: Verbal Fluency Task; TMT: Trail-Making Test; CSF: cerebrospinal fluid; p tau: phosphorylated-tau.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAD\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMCI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eHC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eP-value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;43)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;40)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;45)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge (years)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e66.93\u0026thinsp;\u0026plusmn;\u0026thinsp;6.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e66.93\u0026thinsp;\u0026plusmn;\u0026thinsp;7.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e64.24\u0026thinsp;\u0026plusmn;\u0026thinsp;6.42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.123\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGender (male/female)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e26/17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e19/21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e23/22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.470\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEducation (years)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e9.47\u0026thinsp;\u0026plusmn;\u0026thinsp;3.37\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9.98\u0026thinsp;\u0026plusmn;\u0026thinsp;3.19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e8.87\u0026thinsp;\u0026plusmn;\u0026thinsp;3.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.296\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMMSE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e15.14\u0026thinsp;\u0026plusmn;\u0026thinsp;3.73\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e21.40\u0026thinsp;\u0026plusmn;\u0026thinsp;2.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e27.13\u0026thinsp;\u0026plusmn;\u0026thinsp;1.22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMoCA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e11.81\u0026thinsp;\u0026plusmn;\u0026thinsp;3.19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e18.83\u0026thinsp;\u0026plusmn;\u0026thinsp;2.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e25.09\u0026thinsp;\u0026plusmn;\u0026thinsp;1.08\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHAMA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.55\u0026thinsp;\u0026plusmn;\u0026thinsp;1.91\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.95\u0026thinsp;\u0026plusmn;\u0026thinsp;1.84\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3.57\u0026thinsp;\u0026plusmn;\u0026thinsp;1.50\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.521\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHAMD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.72\u0026thinsp;\u0026plusmn;\u0026thinsp;1.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.20\u0026thinsp;\u0026plusmn;\u0026thinsp;1.68\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3.26\u0026thinsp;\u0026plusmn;\u0026thinsp;1.48\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.313\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eADL\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e27.41\u0026thinsp;\u0026plusmn;\u0026thinsp;4.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e21.53\u0026thinsp;\u0026plusmn;\u0026thinsp;1.31\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e20.24\u0026thinsp;\u0026plusmn;\u0026thinsp;0.57\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePSQI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.26\u0026thinsp;\u0026plusmn;\u0026thinsp;1.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.65\u0026thinsp;\u0026plusmn;\u0026thinsp;2.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.97\u0026thinsp;\u0026plusmn;\u0026thinsp;1.60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.329\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRAVLT (memory)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e12.53\u0026thinsp;\u0026plusmn;\u0026thinsp;4.58\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e17.23\u0026thinsp;\u0026plusmn;\u0026thinsp;3.48\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e20.69\u0026thinsp;\u0026plusmn;\u0026thinsp;2.48\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCDT (visual spatial memory)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e11.30\u0026thinsp;\u0026plusmn;\u0026thinsp;4.84\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e23.38\u0026thinsp;\u0026plusmn;\u0026thinsp;6.48\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e26.71\u0026thinsp;\u0026plusmn;\u0026thinsp;3.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSDMT (attention)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e10.37\u0026thinsp;\u0026plusmn;\u0026thinsp;4.48\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e22.05\u0026thinsp;\u0026plusmn;\u0026thinsp;8.90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e32.20\u0026thinsp;\u0026plusmn;\u0026thinsp;5.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDST (attention)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5.21\u0026thinsp;\u0026plusmn;\u0026thinsp;1.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7.48\u0026thinsp;\u0026plusmn;\u0026thinsp;1.84\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10.27\u0026thinsp;\u0026plusmn;\u0026thinsp;1.83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBNT (verbal naming ability)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e14.58\u0026thinsp;\u0026plusmn;\u0026thinsp;5.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e19.60\u0026thinsp;\u0026plusmn;\u0026thinsp;4.17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e23.24\u0026thinsp;\u0026plusmn;\u0026thinsp;2.47\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVFT (verbal fluency)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7.60\u0026thinsp;\u0026plusmn;\u0026thinsp;2.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e10.80\u0026thinsp;\u0026plusmn;\u0026thinsp;3.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e14.91\u0026thinsp;\u0026plusmn;\u0026thinsp;2.27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTMT-A (attention)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e222.40\u0026thinsp;\u0026plusmn;\u0026thinsp;112.48\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e119.58\u0026thinsp;\u0026plusmn;\u0026thinsp;36.31\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e67.13\u0026thinsp;\u0026plusmn;\u0026thinsp;13.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTMT-B (execution)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e362.51\u0026thinsp;\u0026plusmn;\u0026thinsp;115.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e260.68\u0026thinsp;\u0026plusmn;\u0026thinsp;94.65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e117.91\u0026thinsp;\u0026plusmn;\u0026thinsp;17.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCSF amyloidβ1\u0026ndash;42, pg/ml\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e373.54\u0026thinsp;\u0026plusmn;\u0026thinsp;97.36\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e414.79\u0026thinsp;\u0026plusmn;\u0026thinsp;147.66\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.134\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCSF total tau, pg/ml\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e523.61\u0026thinsp;\u0026plusmn;\u0026thinsp;125.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e469.07\u0026thinsp;\u0026plusmn;\u0026thinsp;151.65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.083\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCSF p tau, pg/ml\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e110.78\u0026thinsp;\u0026plusmn;\u0026thinsp;65.43\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e91.37\u0026thinsp;\u0026plusmn;\u0026thinsp;55.81\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.151\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eClinical characteristics assessment\u003c/h2\u003e \u003cp\u003eThe demographic and clinical characteristics of each participant are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. No significant differences were observed with regard to age, gender and education, among the three groups. On the other hand, significant differences were present for each cognitive domain. Overall cognitive levels, episodic memory, visuospatial working memory, attention, language, and execution were significantly lower in the AD group compared to both the MCI and HC groups. The above cognitive domains were significantly lower in the MCI group compared to the HC group (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eFeature classification\u003c/h2\u003e \u003cp\u003eThe performance metrics (ACC, SEN and SPE) of sixteen SVM classifiers for separating the three groups (AD, MCI, HC) using subsets of sMRI and/or rs-fMRI features, based on two atlases for each modality, are shown in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e3\u003c/span\u003e. Results show that the optimal classifier was based on sMRI (left temporal CT) features (with ACC of 92.05%, SEN of 95.35% and SPE of 88.89%) for distinguishing between AD and HC (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). In terms of separating MCI from HC, the highest ACC of 88.64%, SEN of 85.00% and SPE of 91.11% were achieved by utilizing only the sMRI features. In addition, the 83.13% ACC, 76.74% SEN and 90.00% SPE were obtained for MCI vs. AD, using sMRI features.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eClassification results for the three groups using cortical thickness features\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFeatures\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eACC (%)\u003c/p\u003e \u003cp\u003eM [CIs]\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSEN(%)\u003c/p\u003e \u003cp\u003eM [CIs]\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSPE(%)\u003c/p\u003e \u003cp\u003eM [CIs]\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003cp\u003eM [CIs]\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003eRight Frontal CT\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e86.36 [77.39\u0026ndash;92.75]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e90.70 [77.86\u0026ndash;97.41]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e82.22 [67.95-92.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.80 [0.72\u0026ndash;0.88]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e85.88 [76.64\u0026ndash;92.49]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e85.00 [70.16\u0026ndash;94.29]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e86.67 [73.21\u0026ndash;94.95]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.78 [0.69\u0026ndash;0.87]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e81.93 [71.95\u0026ndash;89.52]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e74.42 [58.83\u0026ndash;86.48]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e90.00 [76.34\u0026ndash;97.21]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.78 [0.69\u0026ndash;0.87]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eLeft Frontal CT\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e87.50 [78.73\u0026ndash;93.59]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e90.70 [77.86\u0026ndash;97.41]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e84.44 [70.54\u0026ndash;93.51]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.83 [0.75\u0026ndash;0.91]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e83.53 [73.91\u0026ndash;90.69]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e75.00 [58.80-87.31]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e91.11 [78.78\u0026ndash;97.52]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.75 [0.66\u0026ndash;0.84]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e83.13 [73.32\u0026ndash;90.46]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e79.07 [63.96\u0026ndash;89.96]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e87.50 [73.20-95.81]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.80 [0.71\u0026ndash;0.89]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRight Parietal CT\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e85.23 [76.06\u0026ndash;91.89]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e79.07 [63.96\u0026ndash;89.96]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e91.11 [78.78\u0026ndash;97.52]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.79 [0.70\u0026ndash;0.87]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e81.18 [71.24\u0026ndash;88.84]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e75.00 [58.80-87.31]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e86.67 [73.21\u0026ndash;94.95]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.77 [0.68\u0026ndash;0.86]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e75.90 [65.27\u0026ndash;84.62]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e69.77 [53.87\u0026ndash;82.82]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e82.50 [67.22\u0026ndash;92.66]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.69 [0.59\u0026ndash;0.79]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eLeft Parietal CT\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e82.95 [73.45\u0026ndash;90.13]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e76.74 [61.37\u0026ndash;88.24]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e88.89 [75.95\u0026ndash;96.29]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.71 [0.62\u0026ndash;0.80]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e84.71 [75.27\u0026ndash;91.60]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e75.00 [58.80-87.31]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e93.33 [81.73\u0026ndash;98.60]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.75 [0.66\u0026ndash;0.84]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e73.49 [62.66\u0026ndash;82.58]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e69.77 [53.87\u0026ndash;82.82]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e77.50 [61.55\u0026ndash;89.16)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.68 [0.58\u0026ndash;0.78]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRight Temporal CT\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e90.91 [82.87\u0026ndash;95.99]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e93.02 [80.94\u0026ndash;98.54]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e88.89 [75.95\u0026ndash;96.29]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.86 [0.79\u0026ndash;0.93]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e87.06 [78.02\u0026ndash;93.36]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e80.00 [64.35\u0026ndash;90.95]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e93.33 [81.73\u0026ndash;98.60]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.79 [0.70\u0026ndash;0.88]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e80.72 [70.59\u0026ndash;88.56]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e74.42 [58.83\u0026ndash;86.48]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e87.50 [73.20-95.81]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.76 [0.67\u0026ndash;0.85]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eLeft Temporal CT\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e92.05 [84.30-96.74]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e95.35 [84.19\u0026ndash;99.43]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e88.89 [75.95\u0026ndash;96.29]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.94 [0.89\u0026ndash;0.99]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e88.24 [79.43\u0026ndash;94.21]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e85.00 [70.16\u0026ndash;94.29]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e91.11 [78.78\u0026ndash;97.52]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.90 [0.84\u0026ndash;0.96]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e83.13 [73.32\u0026ndash;90.46]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e76.74 [61.37\u0026ndash;88.24]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e90.00 [76.34\u0026ndash;97.21]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.78 [0.69\u0026ndash;0.87]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRight Occipital CT\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e73.86 [63.41\u0026ndash;82.66]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e60.47 [44.41\u0026ndash;70.52]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e86.67 [73.21\u0026ndash;94.95]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.69 [0.59\u0026ndash;0.79]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e77.65 [67.31\u0026ndash;85.97]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e65.00 [48.32\u0026ndash;79.37]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e88.89 [75.95\u0026ndash;96.29]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.68 [0.58\u0026ndash;0.78]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e73.49 [62.66\u0026ndash;82.58)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e65.12 [49.07\u0026ndash;78.99]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e82.50 [67.22\u0026ndash;92.66]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.69 [0.59\u0026ndash;0.79]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eLeft Occipital CT\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e73.86 [63.41\u0026ndash;82.66]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e65.12 [49.07\u0026ndash;78.99]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e82.22 [67.95-92.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.67 [57.18\u0026ndash;76.82]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e78.82 [68.61\u0026ndash;86.94]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e75.00 [58.80-87.31]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e82.22 [67.95-92.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.72 [62.45\u0026ndash;81.55]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e72.29 [61.38\u0026ndash;81.55]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e86.05 [72.07\u0026ndash;94.70]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e57.50 [40.89\u0026ndash;72.96]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.69 [59.05\u0026ndash;78.95]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eClassification results for the three groups using cortical thickness and cerebello-cerebral functional connectivity features\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFeatures\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eACC (%)\u003c/p\u003e \u003cp\u003eM (CIs)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSEN(%)\u003c/p\u003e \u003cp\u003eM (CIs)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSPE(%)\u003c/p\u003e \u003cp\u003eM (CIs)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003cp\u003eM (CIs)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003eCT\u0026thinsp;+\u0026thinsp;Right Cerebellar \u0026ndash; frontal FC\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e90.91 [82.87\u0026ndash;95.99]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e95.35 [84.19\u0026ndash;99.43]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e86.67 [73.21\u0026ndash;94.95]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.88 [0.81\u0026ndash;0.95]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e94.12 [86.80-98.06]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e97.50 [86.84\u0026ndash;99.94]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e91.11 [78.78\u0026ndash;97.52]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.95 [0.90\u0026ndash;0.99]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e89.16 [80.41\u0026ndash;94.92]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e88.37 [74.92\u0026ndash;96.11]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e90.00 [76.34\u0026ndash;97.21]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.90 [0.84\u0026ndash;0.96]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCT\u0026thinsp;+\u0026thinsp;Left Cerebellar \u0026ndash; frontal FC\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e87.50 [78.73\u0026ndash;93.59]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e95.35 [84.19\u0026ndash;99.43]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e80.00 [65.40-90.42]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.75 [0.66\u0026ndash;0.84]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e89.41 [80.85\u0026ndash;95.04]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e95.00 [83.08\u0026ndash;99.39]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e84.44 [70.54\u0026ndash;93.51]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.90 [0.90\u0026ndash;0.99]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e87.95 [78.96\u0026ndash;94.07]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e86.05 [72.07\u0026ndash;94.70]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e95.00 [83.08\u0026ndash;99.39]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.80 [0.71\u0026ndash;0.89)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCT\u0026thinsp;+\u0026thinsp;Right Cerebellar \u0026ndash; parietal FC\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e87.50 [78.73\u0026ndash;93.59]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e95.35 [84.19\u0026ndash;99.43]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e80.00 [65.40-90.42]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.78 [0.69\u0026ndash;0.87]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e90.59 [82.29\u0026ndash;95.85]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e95.00 [83.08\u0026ndash;99.39]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e86.67 [73.21\u0026ndash;94.95]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.82 [0.74\u0026ndash;0.90]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e87.95 [78.96\u0026ndash;94.07]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e83.72 [72.04\u0026ndash;94.70]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e82.50 [67.22\u0026ndash;92.66]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.82 [0.74\u0026ndash;0.90]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCT\u0026thinsp;+\u0026thinsp;Left Cerebellar \u0026ndash; parietal FC\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e86.36 [77.39\u0026ndash;92.75]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e95.35 [84.19\u0026ndash;99.43]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e77.78 [62.91\u0026ndash;88.80]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.70 [0.60\u0026ndash;0.80]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e85.88 [76.64\u0026ndash;92.49]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e92.50 [79.61\u0026ndash;98.43]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e80.00 [65.40-90.42]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.72 [0.62\u0026ndash;0.82]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e80.72 [70.59\u0026ndash;88.56]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e76.74 [61.37\u0026ndash;88.24]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e85.00 [70.16\u0026ndash;94.29]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.71 [0.61\u0026ndash;0.81]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCT\u0026thinsp;+\u0026thinsp;Right Cerebellar \u0026ndash; temporal FC\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e86.63 [77.39\u0026ndash;92.75]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e93.02 [80.94\u0026ndash;98.54]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e80.00 [65.40-90.42]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.73 [0.64\u0026ndash;0.82]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e88.24 [79.43\u0026ndash;94.21]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e87.50 [73.20-95.81]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e88.89 [75.95\u0026ndash;96.29]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.77 [0.68\u0026ndash;0.86]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e87.95 [78.96\u0026ndash;94.07]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e90.70 [77.86\u0026ndash;97.41]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e85.00 [70.16\u0026ndash;94.29]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.82 [0.74\u0026ndash;0.90]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCT\u0026thinsp;+\u0026thinsp;Left Cerebellar \u0026ndash; temporal FC\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e87.50 [78.73\u0026ndash;93.59]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e90.70 [77.86\u0026ndash;97.41]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e82.22 [67.95-92.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.76 [0.67\u0026ndash;0.85]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e85.88 [76.64\u0026ndash;92.49]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e90.00 [76.34\u0026ndash;97.21]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e80.00 [65.40-90.42]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.73 [0.64\u0026ndash;0.82]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e85.54 [76.11\u0026ndash;92.30]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e88.37 [74.92\u0026ndash;96.11]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e82.50 [67.22\u0026ndash;92.66]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.77 [0.68\u0026ndash;0.86]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCT\u0026thinsp;+\u0026thinsp;Right Cerebellar \u0026ndash; occipital FC\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e78.41 [68.35\u0026ndash;86.47]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e90.07 [77.86\u0026ndash;97.41]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e66.67 [51.05-80.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.70 [0.60\u0026ndash;0.80]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e80.00 [69.92\u0026ndash;87.90]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e85.00 [70.16\u0026ndash;94.29]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e75.56 [60.46\u0026ndash;87.12]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.68 [0.58\u0026ndash;0.78]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e77.11 [66.58\u0026ndash;85.62]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e72.09 [56.33\u0026ndash;84.67]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e82.50 [67.22\u0026ndash;92.66]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.66 [0.56\u0026ndash;0.76]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCT\u0026thinsp;+\u0026thinsp;Left cerebellar \u0026ndash; occipital FC\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAD vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e73.86 [63.41\u0026ndash;82.66]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e88.37 [74.92\u0026ndash;96.11]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e60.00 [44.33\u0026ndash;74.30]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.64 [0.54\u0026ndash;0.74]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. HC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e78.18 [71.24\u0026ndash;88.84]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e87.50 [73.20-95.81]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e75.56 [60.46\u0026ndash;87.12]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.70 [0.60\u0026ndash;0.80]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI vs. AD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e79.52 [69.24\u0026ndash;87.95]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e76.74 [61.37\u0026ndash;88.24]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e82.50 [67.22\u0026ndash;92.66]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.70 [0.60\u0026ndash;0.80]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e shows the diagnostic performance of the mixed sMRI (CT) and rs-fMRI (cerebello-cerebral FC) features in the three classification tasks. The classification performance of the combined features of CT with FC from the right cerebellum to the left frontal lobe shows the highest ACC (94.43%), SEN (97.5%) and SPE (91.11%) in MCI vs. HC. AD patients have ACC of 90.91%, SEN of 95.35% and SPE of 86.67% compared with HC. In addition, the model that utilized mixed features reached an ACC of 89.16%, SEN of 88.37% and SPE of 90.00% in MCI vs. AD. Furthermore, the ACC of 88.24%, SEN of 87.50% and SPE of 88.89% were obtained for MCI vs. HC, using combination with mixed features from right cerebellum to bilateral temporal lobes. Finally, the ACC of 90.59%, SEN of 95.00% and SPE of 86.67% were obtained for MCI vs. HC, using combination with mixed features from right cerebellum to bilateral parietal lobes. The highest-ranked features are shown in Table\u0026nbsp;4 and Table\u0026nbsp;5.\u003c/p\u003e \u003cp\u003eThe effectiveness of the classification method combining different measures was further analyzed using the AUCs for the concatenation of all features (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e3\u003c/span\u003eD). Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows the ROC curves for individual features and all features (i.e., CT and cerebello-cerebral FC) combinations for each classification group.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eEffective and accurate diagnosis of AD is essential to initiate effective treatment. Particularly, early diagnosis of AD is pivotal in therapeutic development and ultimately for effective patient care [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. ML methods can be used to identify clinical features and characteristic MR images and patterns for diagnostic predictions [\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e]. This approach can potentially help with dementia predisposition identification in those who develop cognitive complaints [\u003cspan additionalcitationids=\"CR62 CR63\" citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e], as this is a common question faced by clinicians. In this study, results showed that the ML model that utilized combined sMRI and rs-fMRI features was more accurate in predicting the diagnostic status of MCI than the ML models using brain morphological features only.\u003c/p\u003e \u003cp\u003eIn this study, sMRI classification features were first extracted to identify atrophy patterns for the three groups (MCI, AD, HC). The findings were generally consistent with previous research, in which one of the earliest imaging marker was atrophy in the temporal lobe, and specifically in the parahippocampus and entorhinal cortex, which has also been previously found to predict MCI to AD progression [\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e, \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e]. The left temporal CT features had the highest classification performance in AD vs. HC groups with an ACC of 92.05%, SEN of 95.35% and SPE of 88.89%. Similarly, for MCI vs. HC groups, the highest atrophy was observed in the left temporal cortex, with diagnostic accuracy of 88.64%, sensitivity of 85.00% and specificity of 91.11%. Finally, for MCI vs. AD groups, in the modeling using sMRI data of the frontal and temporal cortex, an ACC of 83.13% was obtained, which also verifies that the frontal and temporal lobe is involved in the early stage. These areas have been shown to be involved in complex cognitive behavior, decision making and personality expression [\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e], showing potential for the early identification of AD progression. Some individuals with MCI present with minimal visible structural brain changes. Some MCI patients have structural atrophy patterns similar to AD, suggesting that MCI may be more similar to AD than HC. The subtle changes between MCI and HC or the similarity between MCI and AD presents the ML models with a higher difficulty challenge to discriminate between the groups, and therefore requires more features. Additionally, clinical experience reflects this challenge as the differentiation between MCI and HC, as well as MCI and AD is based on functional decline, which is occasionally difficult to identify or quantify, causing overlap among groups. Thus, the results identifying features able to distinguish MCI from the other two groups are more clinically meaningful. Combining different types of data could help to further improve the accuracy of the ML model to distinguish between clinically relevant groups.\u003c/p\u003e \u003cp\u003eThe classification results obtained by combining sMRI and rs-fMRI features in the present study are better than those obtained by the unimodal (sMRI\\rs-fMRI) approach, including those of previous research [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e], for distinguishing MCI from AD and HC (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Most previous studies that constructed brain networks only considered cerebral structural or functional features, while ignoring the cerebellum, and obtained an accuracy lower than that of the present study [\u003cspan additionalcitationids=\"CR69 CR70 CR71\" citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e]. In this work, the diagnostic performance of combined features of CT with FC from the right cerebellum to the left frontal lobe reached 94.43%. In addition, the mixed-features model reached an accuracy of 90.01% and 89.16% for AD vs. HC and MCI vs. AD, respectively. Previous work from our group revealed that the cerebellum showed increased activity of the frontal and temporal lobes in the pre-dementia stage of AD, reflecting a compensatory function that could mitigate early AD symptoms [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. The current study demonstrates that the compensatory regulation of the cerebellum to the frontal lobe is the most significant, which strongly agrees with the findings relating to the enhanced FC between cerebellum and the frontal regions by cerebellum repetitive transcranial magnetic stimulation intervention presented in another work [\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e]. Therefore, it is postulated that when early cognitive impairment occurs, the frontal lobe receives increased assistance from the cerebellum. Furthermore, based on the classification model of CT combined with FC from right cerebellum to bilateral temporal and bilateral parietal lobes, respectively, high ACC of 88.24% and 90.59% were obtained for the classification of MCI vs. HC. This result outperformed other biomarkers for early MCI classification, indicating that the cerebellum has a high classification accuracy rate. Consequently, network-derived regulation is a highly effective biomarker. The crucial role of the cerebellum in brain regulation during the development of early cognitive impairment diseases was confirmed in this study. This conclusion may provide help for the diagnosis and treatment of preclinical AD in the future.\u003c/p\u003e \u003cp\u003eIn this article, the potential of the cerebellum-oriented functional-coefficient network of the brain to be used as a diagnostic biomarker for subjects with early cognitive impairment was explored. The frontal lobe functioned as a \u003cem\u003e\u0026ldquo;\u003c/em\u003erich club\u003cem\u003e\u0026rdquo;\u003c/em\u003e structure with highly interconnected nodes resulting in a large number of connections between networks [\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e]. These interconnections are involved in episodic memory, reasoning, and executive functions such as working memory and cognitive flexibility [\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e, \u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e]. The classification indicates that, as the brain function declines, in the early stage of cognitive impairment, the tightness of the cerebellum and some brain regions will increase. This may be a compensatory regulation mechanism, which only appears in certain brain regions.\u003c/p\u003e \u003cp\u003e However, ML studies have focused on only including clinical, cognitive and structural neuroimaging variables. A previous systematic review highlighted the need to explore additional types of data to improve the performance of ML models [\u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e77\u003c/span\u003e]. The results of the present study suggest that including cerebello-cerebral FC in ML models has the potential to increase the diagnostic utility and accuracy. These phenomena were hereby proved to be among the most important predictors, along with features of CT and FC from the right cerebellum to the left frontal lobe.\u003c/p\u003e \u003cp\u003e\u003cb\u003eLimitations\u003c/b\u003e\u003c/p\u003e \u003cp\u003eThe proposed method effectively improves the accuracy of preclinical diagnosis of AD; however, some limitations need to be highlighted. Future work will focus on the following improvements. To improve the effectiveness of the proposed approach, the data set will be expanded in the following aspects: First, extending the longitudinal data set to better understand the progression of MCI and inclusion of multi-modal data, such as gene and PET data, to investigate different insights into AD characteristics. Second, the parameter acquisition process will be optimized to achieve a higher diagnostic accuracy.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn conclusion, structure MRI measure cortical thickness (CT) could improve Alzheimer\u0026rsquo;s disease (AD) diagnosis. Furthermore, the combination of CT and cerebello-cerebral functional connectivity features shows great potential for the early identification of mild cognitive impairment. These results demonstrate the effectiveness of machine learning in preclinical AD diagnosis, based on the resting-state brain functional connectomics with complex and high-dimensional voxel-wise spatio-temporal patterns. This framework provides a new and intuitive approach to fully exploit deeply embedded diagnostic features from sMRI and resting-state functional MRI data to better individualize the diagnosis of various neurological diseases. Nevertheless, more studies are needed to further investigate the impact of cerebellar neuromodulation on the diagnosis of AD.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors gratefully acknowledge the support of Department of Neurology, Nanjing Brain Hospital Nanjing Medical University.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJingping Shi, Kuiying Yin, Qun Yao and Liangcheng Qu designed the study. Qun Yao, Liangcheng Qu, Bo Song, Xixi Wang, Tong Wang, Minjie Tian, Wenying Ma, Donglin Zhu, Xingjian Lin recruited patients and collected samples. Wenying Ma, Bo Shen, Zonghong Li performed MRI scans and collected data. Qun Yao, Liangcheng Qu, Bo Song, Xixi Wang, Tong Wang performed the experiments and analyzed the data. Qun Yao and Liangcheng Qu interpreted the data and wrote the manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by Nanjing Medical science and Technology Development key project（No. ZKX21034）.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets analyzed for the current study are not publicly available but are available from the corresponding author on reasonable request.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthical Approval and Consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIndividuals provided informed consent prior to participation at each measurement occasion. The Medical Research Ethical Committee of Nanjing Brain Hospital\u0026nbsp;approved the data collection procedures for this study (2022-KY026-01).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests..\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eJack, C.R., Jr., D.S. Knopman, W.J. Jagust, R.C. Petersen, M.W. Weiner, P.S. Aisen, et al., Tracking pathophysiological processes in Alzheimer's disease: an updated hypothetical model of dynamic biomarkers. Lancet Neurol, 2013. \u003cstrong\u003e12\u003c/strong\u003e(2): p. 207-16.http://dx.doi.org/10.1016/s1474-4422(12)70291-0\u003c/li\u003e\n\u003cli\u003eHampel, H. and S. Lista, Dementia: The rising global tide of cognitive impairment. Nat Rev Neurol, 2016. \u003cstrong\u003e12\u003c/strong\u003e(3): p. 131-2.http://dx.doi.org/10.1038/nrneurol.2015.250\u003c/li\u003e\n\u003cli\u003eAllison, J.R., R.C. Rivers, J.C. Christodoulou, M. Vendruscolo, and C.M. Dobson, A relationship between the transient structure in the monomeric state and the aggregation propensities of \u0026alpha;-synuclein and \u0026beta;-synuclein. Biochemistry, 2014. \u003cstrong\u003e53\u003c/strong\u003e(46): p. 7170-83.http://dx.doi.org/10.1021/bi5009326\u003c/li\u003e\n\u003cli\u003eBischkopf, J., A. Busse, and M.C. Angermeyer, Mild cognitive impairment--a review of prevalence, incidence and outcome according to current approaches. Acta Psychiatr Scand, 2002. \u003cstrong\u003e106\u003c/strong\u003e(6): p. 403-14.http://dx.doi.org/10.1034/j.1600-0447.2002.01417.x\u003c/li\u003e\n\u003cli\u003eBraak, H. and E. Braak, Development of Alzheimer-related neurofibrillary changes in the neocortex inversely recapitulates cortical myelogenesis. Acta Neuropathol, 1996. \u003cstrong\u003e92\u003c/strong\u003e(2): p. 197-201.http://dx.doi.org/10.1007/s004010050508\u003c/li\u003e\n\u003cli\u003eDelacourte, A., J.P. David, N. Sergeant, L. Bu\u0026eacute;e, A. Wattez, P. Vermersch, et al., The biochemical pathway of neurofibrillary degeneration in aging and Alzheimer's disease. Neurology, 1999. \u003cstrong\u003e52\u003c/strong\u003e(6): p. 1158-65.http://dx.doi.org/10.1212/wnl.52.6.1158\u003c/li\u003e\n\u003cli\u003eMorris, J.C., M. Storandt, D.W. McKeel, Jr., E.H. Rubin, J.L. Price, E.A. Grant, and L. Berg, Cerebral amyloid deposition and diffuse plaques in \"normal\" aging: Evidence for presymptomatic and very mild Alzheimer's disease. Neurology, 1996. \u003cstrong\u003e46\u003c/strong\u003e(3): p. 707-19.http://dx.doi.org/10.1212/wnl.46.3.707\u003c/li\u003e\n\u003cli\u003eSerrano-Pozo, A., M.P. Frosch, E. Masliah, and B.T. Hyman, Neuropathological alterations in Alzheimer disease. Cold Spring Harb Perspect Med, 2011. \u003cstrong\u003e1\u003c/strong\u003e(1): p. a006189.http://dx.doi.org/10.1101/cshperspect.a006189\u003c/li\u003e\n\u003cli\u003eFarr\u0026agrave;s-Permanyer, L., J. Gu\u0026agrave;rdia-Olmos, and M. Per\u0026oacute;-Cebollero, Mild cognitive impairment and fMRI studies of brain functional connectivity: the state of the art. Front Psychol, 2015. \u003cstrong\u003e6\u003c/strong\u003e: p. 1095.http://dx.doi.org/10.3389/fpsyg.2015.01095\u003c/li\u003e\n\u003cli\u003eLiu, Z., Y. Zhang, H. Yan, L. Bai, R. Dai, W. Wei, et al., Altered topological patterns of brain networks in mild cognitive impairment and Alzheimer's disease: a resting-state fMRI study. Psychiatry Res, 2012. \u003cstrong\u003e202\u003c/strong\u003e(2): p. 118-25.http://dx.doi.org/10.1016/j.pscychresns.2012.03.002\u003c/li\u003e\n\u003cli\u003eZhang, T., Z. Zhao, C. Zhang, J. Zhang, Z. Jin, and L. Li, Classification of Early and Late Mild Cognitive Impairment Using Functional Brain Network of Resting-State fMRI. Front Psychiatry, 2019. \u003cstrong\u003e10\u003c/strong\u003e: p. 572.http://dx.doi.org/10.3389/fpsyt.2019.00572\u003c/li\u003e\n\u003cli\u003eCai, S., T. Chong, Y. Peng, W. Shen, J. Li, K.M. von Deneen, and L. Huang, Altered functional brain networks in amnestic mild cognitive impairment: a resting-state fMRI study. Brain Imaging Behav, 2017. \u003cstrong\u003e11\u003c/strong\u003e(3): p. 619-631.http://dx.doi.org/10.1007/s11682-016-9539-0\u003c/li\u003e\n\u003cli\u003eFan, Y., N. Batmanghelich, C.M. Clark, and C. Davatzikos, Spatial patterns of brain atrophy in MCI patients, identified via high-dimensional pattern classification, predict subsequent cognitive decline. Neuroimage, 2008. \u003cstrong\u003e39\u003c/strong\u003e(4): p. 1731-43.http://dx.doi.org/10.1016/j.neuroimage.2007.10.031\u003c/li\u003e\n\u003cli\u003eChupin, M., E. G\u0026eacute;rardin, R. Cuingnet, C. Boutet, L. Lemieux, S. Leh\u0026eacute;ricy, et al., Fully automatic hippocampus segmentation and classification in Alzheimer's disease and mild cognitive impairment applied on data from ADNI. Hippocampus, 2009. \u003cstrong\u003e19\u003c/strong\u003e(6): p. 579-87.http://dx.doi.org/10.1002/hipo.20626\u003c/li\u003e\n\u003cli\u003eQuerbes, O., F. Aubry, J. Pariente, J.A. Lotterie, J.F. D\u0026eacute;monet, V. Duret, et al., Early diagnosis of Alzheimer's disease using cortical thickness: impact of cognitive reserve. Brain, 2009. \u003cstrong\u003e132\u003c/strong\u003e(Pt 8): p. 2036-47.http://dx.doi.org/10.1093/brain/awp105\u003c/li\u003e\n\u003cli\u003eWestman, E., J.S. Muehlboeck, and A. Simmons, Combining MRI and CSF measures for classification of Alzheimer's disease and prediction of mild cognitive impairment conversion. Neuroimage, 2012. \u003cstrong\u003e62\u003c/strong\u003e(1): p. 229-38.http://dx.doi.org/10.1016/j.neuroimage.2012.04.056\u003c/li\u003e\n\u003cli\u003eCho, Y., J.K. Seong, Y. Jeong, and S.Y. Shin, Individual subject classification for Alzheimer's disease based on incremental learning using a spatial frequency representation of cortical thickness data. Neuroimage, 2012. \u003cstrong\u003e59\u003c/strong\u003e(3): p. 2217-30.http://dx.doi.org/10.1016/j.neuroimage.2011.09.085\u003c/li\u003e\n\u003cli\u003eEskildsen, S.F., P. Coup\u0026eacute;, D. Garc\u0026iacute;a-Lorenzo, V. Fonov, J.C. Pruessner, and D.L. Collins, Prediction of Alzheimer's disease in subjects with mild cognitive impairment from the ADNI cohort using patterns of cortical thinning. Neuroimage, 2013. \u003cstrong\u003e65\u003c/strong\u003e: p. 511-21.http://dx.doi.org/10.1016/j.neuroimage.2012.09.058\u003c/li\u003e\n\u003cli\u003eWang, T., S. Xiao, Y. Liu, Z. Lin, N. Su, X. Li, et al., The efficacy of plasma biomarkers in early diagnosis of Alzheimer's disease. Int J Geriatr Psychiatry, 2014. \u003cstrong\u003e29\u003c/strong\u003e(7): p. 713-9.http://dx.doi.org/10.1002/gps.4053\u003c/li\u003e\n\u003cli\u003eMagnin, B., L. Mesrob, S. Kinkingn\u0026eacute;hun, M. P\u0026eacute;l\u0026eacute;grini-Issac, O. Colliot, M. Sarazin, et al., Support vector machine-based classification of Alzheimer's disease from whole-brain anatomical MRI. Neuroradiology, 2009. \u003cstrong\u003e51\u003c/strong\u003e(2): p. 73-83.http://dx.doi.org/10.1007/s00234-008-0463-x\u003c/li\u003e\n\u003cli\u003eKl\u0026ouml;ppel, S., C.M. Stonnington, C. Chu, B. Draganski, R.I. Scahill, J.D. Rohrer, et al., Automatic classification of MR scans in Alzheimer's disease. Brain, 2008. \u003cstrong\u003e131\u003c/strong\u003e(Pt 3): p. 681-9.http://dx.doi.org/10.1093/brain/awm319\u003c/li\u003e\n\u003cli\u003eZhang, Y., Z. Dong, P. Phillips, S. Wang, G. Ji, J. Yang, and T.F. Yuan, Detection of subjects and brain regions related to Alzheimer's disease using 3D MRI scans based on eigenbrain and machine learning. Front Comput Neurosci, 2015. \u003cstrong\u003e9\u003c/strong\u003e: p. 66.http://dx.doi.org/10.3389/fncom.2015.00066\u003c/li\u003e\n\u003cli\u003eZhong, X., J. Li, H. Dou, S. Deng, G. Wang, Y. Jiang, et al., Fuzzy nonlinear proximal support vector machine for land extraction based on remote sensing image. PLoS One, 2013. \u003cstrong\u003e8\u003c/strong\u003e(7): p. e69434.http://dx.doi.org/10.1371/journal.pone.0069434\u003c/li\u003e\n\u003cli\u003eRapoport, M., R. van Reekum, and H. Mayberg, The role of the cerebellum in cognition and behavior: a selective review. J Neuropsychiatry Clin Neurosci, 2000. \u003cstrong\u003e12\u003c/strong\u003e(2): p. 193-8.http://dx.doi.org/10.1176/jnp.12.2.193\u003c/li\u003e\n\u003cli\u003eSchmahmann, J.D. and D. Caplan, Cognition, emotion and the cerebellum. Brain, 2006. \u003cstrong\u003e129\u003c/strong\u003e(Pt 2): p. 290-2.http://dx.doi.org/10.1093/brain/awh729\u003c/li\u003e\n\u003cli\u003eBalsters, J.H., C.D. Whelan, I.H. Robertson, and N. Ramnani, Cerebellum and cognition: evidence for the encoding of higher order rules. Cereb Cortex, 2013. \u003cstrong\u003e23\u003c/strong\u003e(6): p. 1433-43.http://dx.doi.org/10.1093/cercor/bhs127\u003c/li\u003e\n\u003cli\u003eSereno, M.I., J. Diedrichsen, M. Tachrount, G. Testa-Silva, H. d'Arceuil, and C. De Zeeuw, The human cerebellum has almost 80% of the surface area of the neocortex. Proc Natl Acad Sci U S A, 2020. \u003cstrong\u003e117\u003c/strong\u003e(32): p. 19538-19543.http://dx.doi.org/10.1073/pnas.2002896117\u003c/li\u003e\n\u003cli\u003eMarvel, C.L. and J.E. Desmond, Functional topography of the cerebellum in verbal working memory. Neuropsychol Rev, 2010. \u003cstrong\u003e20\u003c/strong\u003e(3): p. 271-9.http://dx.doi.org/10.1007/s11065-010-9137-7\u003c/li\u003e\n\u003cli\u003eSchmahmann, J.D., The role of the cerebellum in cognition and emotion: personal reflections since 1982 on the dysmetria of thought hypothesis, and its historical evolution from theory to therapy. Neuropsychol Rev, 2010. \u003cstrong\u003e20\u003c/strong\u003e(3): p. 236-60.http://dx.doi.org/10.1007/s11065-010-9142-x\u003c/li\u003e\n\u003cli\u003eDe Smet, H.J., P. Paquier, J. Verhoeven, and P. Mari\u0026euml;n, The cerebellum: its role in language and related cognitive and affective functions. Brain Lang, 2013. \u003cstrong\u003e127\u003c/strong\u003e(3): p. 334-42.http://dx.doi.org/10.1016/j.bandl.2012.11.001\u003c/li\u003e\n\u003cli\u003eKoziol, L.F., D.E. Budding, and D. Chidekel, From movement to thought: executive function, embodied cognition, and the cerebellum. Cerebellum, 2012. \u003cstrong\u003e11\u003c/strong\u003e(2): p. 505-25.http://dx.doi.org/10.1007/s12311-011-0321-y\u003c/li\u003e\n\u003cli\u003eAdamaszek, M., F. D'Agata, R. Ferrucci, C. Habas, S. Keulen, K.C. Kirkby, et al., Consensus Paper: Cerebellum and Emotion. Cerebellum, 2017. \u003cstrong\u003e16\u003c/strong\u003e(2): p. 552-576.http://dx.doi.org/10.1007/s12311-016-0815-8\u003c/li\u003e\n\u003cli\u003eRaymond, J.L. and J.F. Medina, Computational Principles of Supervised Learning in the Cerebellum. Annu Rev Neurosci, 2018. \u003cstrong\u003e41\u003c/strong\u003e: p. 233-253.http://dx.doi.org/10.1146/annurev-neuro-080317-061948\u003c/li\u003e\n\u003cli\u003eWang, D., Q. Yao, M. Yu, C. Xiao, L. Fan, X. Lin, et al., Topological Disruption of Structural Brain Networks in Patients With Cognitive Impairment Following Cerebellar Infarction. Front Neurol, 2019. \u003cstrong\u003e10\u003c/strong\u003e: p. 759.http://dx.doi.org/10.3389/fneur.2019.00759\u003c/li\u003e\n\u003cli\u003eTang, F., D. Zhu, W. Ma, Q. Yao, Q. Li, and J. Shi, Differences Changes in Cerebellar Functional Connectivity Between Mild Cognitive Impairment and Alzheimer's Disease: A Seed-Based Approach. Front Neurol, 2021. \u003cstrong\u003e12\u003c/strong\u003e: p. 645171.http://dx.doi.org/10.3389/fneur.2021.645171\u003c/li\u003e\n\u003cli\u003eTabatabaei-Jafari, H., E. Walsh, M.E. Shaw, and N. Cherbuin, The cerebellum shrinks faster than normal ageing in Alzheimer's disease but not in mild cognitive impairment. Hum Brain Mapp, 2017. \u003cstrong\u003e38\u003c/strong\u003e(6): p. 3141-3150.http://dx.doi.org/10.1002/hbm.23580\u003c/li\u003e\n\u003cli\u003eLiang, K.J. and E.S. Carlson, Resistance, vulnerability and resilience: A review of the cognitive cerebellum in aging and neurodegenerative diseases. Neurobiol Learn Mem, 2020. \u003cstrong\u003e170\u003c/strong\u003e: p. 106981.http://dx.doi.org/10.1016/j.nlm.2019.01.004\u003c/li\u003e\n\u003cli\u003eMarkesbery, W.R., Neuropathologic alterations in mild cognitive impairment: a review. J Alzheimers Dis, 2010. \u003cstrong\u003e19\u003c/strong\u003e(1): p. 221-8.http://dx.doi.org/10.3233/jad-2010-1220\u003c/li\u003e\n\u003cli\u003eJack, C.R., Jr., D.A. Bennett, K. Blennow, M.C. Carrillo, B. Dunn, S.B. Haeberlein, et al., NIA-AA Research Framework: Toward a biological definition of Alzheimer's disease. Alzheimers Dement, 2018. \u003cstrong\u003e14\u003c/strong\u003e(4): p. 535-562.http://dx.doi.org/10.1016/j.jalz.2018.02.018\u003c/li\u003e\n\u003cli\u003eMorris, J.C., The Clinical Dementia Rating (CDR): current version and scoring rules. Neurology, 1993. \u003cstrong\u003e43\u003c/strong\u003e(11): p. 2412-4.http://dx.doi.org/10.1212/wnl.43.11.2412-a\u003c/li\u003e\n\u003cli\u003eTombaugh, T.N. and N.J. McIntyre, The mini-mental state examination: a comprehensive review. J Am Geriatr Soc, 1992. \u003cstrong\u003e40\u003c/strong\u003e(9): p. 922-35.http://dx.doi.org/10.1111/j.1532-5415.1992.tb01992.x\u003c/li\u003e\n\u003cli\u003eHorton, D.K., L.S. Hynan, L.H. Lacritz, H.C. Rossetti, M.F. Weiner, and C.M. Cullum, An Abbreviated Montreal Cognitive Assessment (MoCA) for Dementia Screening. Clin Neuropsychol, 2015. \u003cstrong\u003e29\u003c/strong\u003e(4): p. 413-25.http://dx.doi.org/10.1080/13854046.2015.1043349\u003c/li\u003e\n\u003cli\u003eCano, S.J., H.B. Posner, M.L. Moline, S.W. Hurt, J. Swartz, T. Hsu, and J.C. Hobart, The ADAS-cog in Alzheimer's disease clinical trials: psychometric evaluation of the sum and its parts. J Neurol Neurosurg Psychiatry, 2010. \u003cstrong\u003e81\u003c/strong\u003e(12): p. 1363-8.http://dx.doi.org/10.1136/jnnp.2009.204008\u003c/li\u003e\n\u003cli\u003eVakil, E. and H. Blachstein, Rey Auditory-Verbal Learning Test: structure analysis. J Clin Psychol, 1993. \u003cstrong\u003e49\u003c/strong\u003e(6): p. 883-90.http://dx.doi.org/10.1002/1097-4679(199311)49:6\u0026lt;883::aid-jclp2270490616\u0026gt;3.0.co;2-6\u003c/li\u003e\n\u003cli\u003eMainland, B.J., S. Amodeo, and K.I. Shulman, Multiple clock drawing scoring systems: simpler is better. Int J Geriatr Psychiatry, 2014. \u003cstrong\u003e29\u003c/strong\u003e(2): p. 127-36.http://dx.doi.org/10.1002/gps.3992\u003c/li\u003e\n\u003cli\u003eLeite, K.S., E.C. Miotto, R. Nitrini, and M.S. Yassuda, Boston Naming Test (BNT) original, Brazilian adapted version and short forms: normative data for illiterate and low-educated older adults. Int Psychogeriatr, 2017. \u003cstrong\u003e29\u003c/strong\u003e(5): p. 825-833.http://dx.doi.org/10.1017/s1041610216001952\u003c/li\u003e\n\u003cli\u003eGarc\u0026iacute;a-Herranz, S., M.C. D\u0026iacute;az-Mardomingo, C. Venero, and H. Peraita, Accuracy of verbal fluency tests in the discrimination of mild cognitive impairment and probable Alzheimer's disease in older Spanish monolingual individuals. Neuropsychol Dev Cogn B Aging Neuropsychol Cogn, 2020. \u003cstrong\u003e27\u003c/strong\u003e(6): p. 826-840.http://dx.doi.org/10.1080/13825585.2019.1698710\u003c/li\u003e\n\u003cli\u003eHasegawa, C. and N. Hirono, [Trail Making Test (TMT) A and B]. Nihon Rinsho, 2011. \u003cstrong\u003e69 Suppl 8\u003c/strong\u003e: p. 423-7\u003c/li\u003e\n\u003cli\u003eCosta, S.L., S. Schwizer Ashkenazi, L.B. Strober, N.D. Chiaravalloti, and E. Vakil, The adapted symbol digit modalities test: Examining the impact of response modality. NeuroRehabilitation, 2021. \u003cstrong\u003e49\u003c/strong\u003e(2): p. 215-220.http://dx.doi.org/10.3233/nre-218021\u003c/li\u003e\n\u003cli\u003eBowden, S.C., V.M. Petrauskas, F.J. Bardenhagen, C.E. Meade, and L.C. Simpson, Exploring the dimensionality of digit span. Assessment, 2013. \u003cstrong\u003e20\u003c/strong\u003e(2): p. 188-98.http://dx.doi.org/10.1177/1073191112457016\u003c/li\u003e\n\u003cli\u003eBagby, R.M., A.G. Ryder, D.R. Schuller, and M.B. Marshall, The Hamilton Depression Rating Scale: has the gold standard become a lead weight? Am J Psychiatry, 2004. \u003cstrong\u003e161\u003c/strong\u003e(12): p. 2163-77.http://dx.doi.org/10.1176/appi.ajp.161.12.2163\u003c/li\u003e\n\u003cli\u003eThompson, E., Hamilton Rating Scale for Anxiety (HAM-A). Occup Med (Lond), 2015. \u003cstrong\u003e65\u003c/strong\u003e(7): p. 601.http://dx.doi.org/10.1093/occmed/kqv054\u003c/li\u003e\n\u003cli\u003eBuysse, D.J., C.F. Reynolds, 3rd, T.H. Monk, C.C. Hoch, A.L. Yeager, and D.J. Kupfer, Quantification of subjective sleep quality in healthy elderly men and women using the Pittsburgh Sleep Quality Index (PSQI). Sleep, 1991. \u003cstrong\u003e14\u003c/strong\u003e(4): p. 331-8\u003c/li\u003e\n\u003cli\u003eFischl, B., D.H. Salat, E. Busa, M. Albert, M. Dieterich, C. Haselgrove, et al., Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron, 2002. \u003cstrong\u003e33\u003c/strong\u003e(3): p. 341-55.http://dx.doi.org/10.1016/s0896-6273(02)00569-x\u003c/li\u003e\n\u003cli\u003eDesikan, R.S., F. S\u0026eacute;gonne, B. Fischl, B.T. Quinn, B.C. Dickerson, D. Blacker, et al., An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage, 2006. \u003cstrong\u003e31\u003c/strong\u003e(3): p. 968-80.http://dx.doi.org/10.1016/j.neuroimage.2006.01.021\u003c/li\u003e\n\u003cli\u003eYan, C.G., X.D. Wang, X.N. Zuo, and Y.F. Zang, DPABI: Data Processing \u0026amp; Analysis for (Resting-State) Brain Imaging. Neuroinformatics, 2016. \u003cstrong\u003e14\u003c/strong\u003e(3): p. 339-51.http://dx.doi.org/10.1007/s12021-016-9299-4\u003c/li\u003e\n\u003cli\u003eTzourio-Mazoyer, N., B. Landeau, D. Papathanassiou, F. Crivello, O. Etard, N. Delcroix, et al., Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage, 2002. \u003cstrong\u003e15\u003c/strong\u003e(1): p. 273-89.http://dx.doi.org/10.1006/nimg.2001.0978\u003c/li\u003e\n\u003cli\u003eRajeswari., S. and J.K. Theiva, Support Vector Machine Classification For MRI Images. International Journal of Electronics \u0026amp; Computer Science Engineering, 2012. \u003cstrong\u003e1\u003c/strong\u003e(3)\u003c/li\u003e\n\u003cli\u003eSkolariki, K., G.M. Terrera, and S. Danso, Multivariate Data Analysis and Machine Learning for Prediction of MCI-to-AD Conversion. Adv Exp Med Biol, 2020. \u003cstrong\u003e1194\u003c/strong\u003e: p. 81-103.http://dx.doi.org/10.1007/978-3-030-32622-7_8\u003c/li\u003e\n\u003cli\u003eBron, E.E., S. Klein, J.M. Papma, L.C. Jiskoot, V. Venkatraghavan, J. Linders, et al., Cross-cohort generalizability of deep and conventional machine learning for MRI-based diagnosis and prediction of Alzheimer's disease. Neuroimage Clin, 2021. \u003cstrong\u003e31\u003c/strong\u003e: p. 102712.http://dx.doi.org/10.1016/j.nicl.2021.102712\u003c/li\u003e\n\u003cli\u003eBasaia, S., F. Agosta, L. Wagner, E. Canu, G. Magnani, R. Santangelo, and M. Filippi, Automated classification of Alzheimer's disease and mild cognitive impairment using a single MRI and deep neural networks. Neuroimage Clin, 2019. \u003cstrong\u003e21\u003c/strong\u003e: p. 101645.http://dx.doi.org/10.1016/j.nicl.2018.101645\u003c/li\u003e\n\u003cli\u003ePopuri, K., D. Ma, L. Wang, and M.F. Beg, Using machine learning to quantify structural MRI neurodegeneration patterns of Alzheimer's disease into dementia score: Independent validation on 8,834 images from ADNI, AIBL, OASIS, and MIRIAD databases. Hum Brain Mapp, 2020. \u003cstrong\u003e41\u003c/strong\u003e(14): p. 4127-4147.http://dx.doi.org/10.1002/hbm.25115\u003c/li\u003e\n\u003cli\u003eZhou, Y., Z. Song, X. Han, H. Li, and X. Tang, Prediction of Alzheimer's Disease Progression Based on Magnetic Resonance Imaging. ACS Chem Neurosci, 2021. \u003cstrong\u003e12\u003c/strong\u003e(22): p. 4209-4223.http://dx.doi.org/10.1021/acschemneuro.1c00472\u003c/li\u003e\n\u003cli\u003eChagu\u0026eacute;, P., B. Marro, S. Fadili, M. Houot, A. Morin, J. Samper-Gonz\u0026aacute;lez, et al., Radiological classification of dementia from anatomical MRI assisted by machine learning-derived maps. J Neuroradiol, 2021. \u003cstrong\u003e48\u003c/strong\u003e(6): p. 412-418.http://dx.doi.org/10.1016/j.neurad.2020.04.004\u003c/li\u003e\n\u003cli\u003eKorolev, I.O., L.L. Symonds, and A.C. Bozoki, Predicting Progression from Mild Cognitive Impairment to Alzheimer's Dementia Using Clinical, MRI, and Plasma Biomarkers via Probabilistic Pattern Classification. PLoS One, 2016. \u003cstrong\u003e11\u003c/strong\u003e(2): p. e0138866.http://dx.doi.org/10.1371/journal.pone.0138866\u003c/li\u003e\n\u003cli\u003eZhang, D. and D. Shen, Predicting future clinical changes of MCI patients using longitudinal and multimodal biomarkers. PLoS One, 2012. \u003cstrong\u003e7\u003c/strong\u003e(3): p. e33182.http://dx.doi.org/10.1371/journal.pone.0033182\u003c/li\u003e\n\u003cli\u003eWei, R., C. Li, N. Fogelson, and L. Li, Prediction of Conversion from Mild Cognitive Impairment to Alzheimer's Disease Using MRI and Structural Network Features. Front Aging Neurosci, 2016. \u003cstrong\u003e8\u003c/strong\u003e: p. 76.http://dx.doi.org/10.3389/fnagi.2016.00076\u003c/li\u003e\n\u003cli\u003eSuk, H.I., S.W. Lee, and D. Shen, Subclass-based multi-task learning for Alzheimer's disease diagnosis. Front Aging Neurosci, 2014. \u003cstrong\u003e6\u003c/strong\u003e: p. 168.http://dx.doi.org/10.3389/fnagi.2014.00168\u003c/li\u003e\n\u003cli\u003eMoradi, E., A. Pepe, C. Gaser, H. Huttunen, and J. Tohka, Machine learning framework for early MRI-based Alzheimer's conversion prediction in MCI subjects. Neuroimage, 2015. \u003cstrong\u003e104\u003c/strong\u003e: p. 398-412.http://dx.doi.org/10.1016/j.neuroimage.2014.10.002\u003c/li\u003e\n\u003cli\u003eArdekani, B.A., E. Bermudez, A.M. Mubeen, and A.H. Bachman, Prediction of Incipient Alzheimer's Disease Dementia in Patients with Mild Cognitive Impairment. J Alzheimers Dis, 2017. \u003cstrong\u003e55\u003c/strong\u003e(1): p. 269-281.http://dx.doi.org/10.3233/jad-160594\u003c/li\u003e\n\u003cli\u003eSuk, H.I., S.W. Lee, and D. Shen, Deep sparse multi-task learning for feature selection in Alzheimer's disease diagnosis. Brain Struct Funct, 2016. \u003cstrong\u003e221\u003c/strong\u003e(5): p. 2569-87.http://dx.doi.org/10.1007/s00429-015-1059-y\u003c/li\u003e\n\u003cli\u003eZheng, W., Z. Yao, Y. Li, Y. Zhang, B. Hu, and D. Wu, Brain Connectivity Based Prediction of Alzheimer's Disease in Patients With Mild Cognitive Impairment Based on Multi-Modal Images. Front Hum Neurosci, 2019. \u003cstrong\u003e13\u003c/strong\u003e: p. 399.http://dx.doi.org/10.3389/fnhum.2019.00399\u003c/li\u003e\n\u003cli\u003eYao, Q., F. Tang, Y. Wang, Y. Yan, L. Dong, T. Wang, et al., Effect of cerebellum stimulation on cognitive recovery in patients with Alzheimer disease: A randomized clinical trial. Brain Stimul, 2022. \u003cstrong\u003e15\u003c/strong\u003e(4): p. 910-920.http://dx.doi.org/10.1016/j.brs.2022.06.004\u003c/li\u003e\n\u003cli\u003evan den Heuvel, M.P. and O. Sporns, Rich-club organization of the human connectome. J Neurosci, 2011. \u003cstrong\u003e31\u003c/strong\u003e(44): p. 15775-86.http://dx.doi.org/10.1523/jneurosci.3539-11.2011\u003c/li\u003e\n\u003cli\u003eBlumenfeld, R.S., C.M. Parks, A.P. Yonelinas, and C. Ranganath, Putting the pieces together: the role of dorsolateral prefrontal cortex in relational memory encoding. J Cogn Neurosci, 2011. \u003cstrong\u003e23\u003c/strong\u003e(1): p. 257-65.http://dx.doi.org/10.1162/jocn.2010.21459\u003c/li\u003e\n\u003cli\u003eBlumenfeld, R.S. and C. Ranganath, Prefrontal cortex and long-term memory encoding: an integrative review of findings from neuropsychology and neuroimaging. Neuroscientist, 2007. \u003cstrong\u003e13\u003c/strong\u003e(3): p. 280-91.http://dx.doi.org/10.1177/1073858407299290\u003c/li\u003e\n\u003cli\u003ePellegrini, E., L. Ballerini, M. Hernandez, F.M. Chappell, V. Gonz\u0026aacute;lez-Castro, D. Anblagan, et al., Machine learning of neuroimaging for assisted diagnosis of cognitive impairment and dementia: A systematic review. Alzheimers Dement (Amst), 2018. \u003cstrong\u003e10\u003c/strong\u003e: p. 519-535.http://dx.doi.org/10.1016/j.dadm.2018.07.004\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Alzheimer’s disease, Mild cognitive impairment, Functional magnetic resonance imaging, Cerebellum, Machine learning","lastPublishedDoi":"10.21203/rs.3.rs-2663342/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-2663342/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEarly identification of degenerative processes in Alzheimer’s disease (AD) is essential. Cerebello-cerebral network changes can be used for early diagnosis of dementia and its stages, namely mild cognitive impairment (MCI) and AD.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFeatures of cortical thickness (CT) and cerebello-cerebral functional connectivity (FC) extracted from MRI data were used to analyze structural and functional changes, and machine learning for the disease progression classification.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCT features have an accuracy of 92.05% for AD vs. HC, 88.64% for MCI vs. HC, and 83.13% for MCI vs. AD. Additionally, combined with convolutional CT and cerebello-cerebral FC features, the accuracy of the classifier reached 94.12% for MCI vs. HC, 90.91% for AD vs. HC, and 89.16% for MCI vs. AD, evaluated using support vector machines.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe proposed pipeline offers a promising low-cost alternative for the diagnosis of preclinical AD and can be useful for other degenerative brain disorders.\u003c/p\u003e","manuscriptTitle":"Machine learning of cerebello-cerebral functional networks for mild cognitive impairment detection","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2023-03-13 19:09:34","doi":"10.21203/rs.3.rs-2663342/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"e274de6c-01fb-4f48-9763-1e83ee185c3a","owner":[],"postedDate":"March 13th, 2023","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2023-03-26T17:14:23+00:00","versionOfRecord":[],"versionCreatedAt":"2023-03-13 19:09:34","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-2663342","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-2663342","identity":"rs-2663342","version":["v1"]},"buildId":"7rjqhiLT3MXkJMwkYKINL","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}