18F-FDG PET radiomics-based machine learning model for differentiating pathological subtypes in locally advanced cervical cancer

18F-FDG PET radiomics-based machine learning model for differentiating pathological subtypes in locally advanced cervical cancer | 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 18F-FDG PET radiomics-based machine learning model for differentiating pathological subtypes in locally advanced cervical cancer Huiling Liu, Mi Lao, Cheng Chang, Yalin Zhang, Yong Yin, Ruozheng Wang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3197925/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 Purpose To determine diagnostic performance of 18 F-fluorodeoxyglucose (FDG) positron emission tomography (PET) and computed tomography (CT) radiomics-based machine learning (ML) for classification of cervical adenocarcinoma (AC) and squamous cell carcinoma (SCC). Methods A total of 195 patients with locally advanced cervical cancer were enrolled in this study, and randomly allocated to training cohort (n = 136) and validation cohort (n = 59) in a ratio of 7:3. Radiomics features were extracted from pretreatment 18 F-FDG PET/CT and selected by the Pearson correlation coefficient and the least absolute shrinkage and selection operator regression analysis. Six ML classifiers were trained and validated, and the best-performing classifier was selected based on accuracy, sensitivity, specificity, and area under the curve (AUC). The performance of different models was assessed and compared using the DeLong test. Results Five PET and one CT radiomics features were selected and incorporated into the ML classifiers. The PET radiomics model constructed based on the lightGBM algorithm had an accuracy of 0.915 and an AUC of 0.851 (95% CI, 0.715–0.986) in the validation cohort, which were higher than that of the CT radiomics model (accuracy: 0.661; AUC: 0.513 [95% CI, 0.339–0.688]). The DeLong test revealed no significant difference in AUC between the combined radiomics model and the PET radiomics model in both the training cohort ( P = 0.347) and the validation cohort ( P = 0.776). Conclusions The 18 F-FDG PET radiomics model can be used as a clinically applicable tool for differentiating pathological subtypes in patients with locally advanced cervical cancer. Locally advanced cervical cancer positron emission tomography radiomics adenocarcinoma squamous cell carcinoma Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Based on the data from GLOBOCAN 2020, cervical cancer ranks as the fourth most prevalent cancer among women globally (Saleh et al. 2020 ). In 2020, there were approximately 604,000 new cases and 342,000 deaths from cervical cancer globally, accounting for approximately 3.1% and 3.4% of the total new cases and deaths from female cancers respectively (Sung et al. 2021 ). According to the latest cancer data in China, in 2015, there were approximately 111,000 reported new cases of cervical cancer and approximately 34,000 deaths, accounting for 6.24% and 3.94% of new cancer cases and deaths in Chinese women, respectively. Notably, cervical cancer has exhibited a trend of affecting younger individuals, with the incidence rate significantly increasing after the age of 25 and peaking between the ages of 50 and 54 (Zheng et al. 2022 ). Squamous cell carcinoma (SCC) is the most common pathological subtype of cervical cancer, accounting for about 70–75% of cases, followed by adenocarcinoma (AC) which represents approximately 10–25% of cases. The incidence of AC has been observed to increase in recent decades (Gadducci et al. 2019 ; Smith et al. 2023 ). Although SCC and AC share many risk factors such as HPV infection, multiple sexual partners, and long-term use of oral contraceptives, they have different histological origins (Gallardo-Alvarado et al. 2022 ). However, despite these differences, SCC and AC are still considered as a whole (Williams et al. 2015 ). According to international and country-specific guidelines for locally advanced cervical cancer (LACC), there are no distinct differences in treatment strategies for AC and SCC (Pujade-Lauraine et al. 2022 ). Patients with LACC who receive radiation therapy or concurrent chemoradiotherapy have a worse prognosis for AC compared to SCC, highlighting the need for alternative treatment options specifically for AC cases (Hu et al. 2018 ). Therefore, there is a critical need to advance research on early-stage differentiation of pathological subtypes of cervical cancer. It is essential to uncover the differences between AC and SCC from multiple perspectives, explore the underlying reasons for these differences, and develop personalized treatment strategies and plans, as this holds significant importance (Fujiwara et al. 2014 ). Biopsy serves as the standard for qualitative diagnosis of cervical cancer. However, biopsy is an invasive procedure associated with risks of bleeding and infection. Point-to-point biopsy performed on larger tumors only evaluate a small portion of the sample, resulting in sampling bias and an inability to comprehensively assess tumor heterogeneity (Fan et al. 2017 ; Kim et al. 2020 ; Nassiri et al. 2022 ). Previous study investigated the value of apparent diffusion coefficient (ADC) in evaluating the pathological subtypes and grading of cervical cancer and found that the ADC value of AC was higher than that of SCC (Karunya et al. 2017 ). However, Winfield et al. discovered that ADC cannot differentiate between SCC and AC (Winfield et al. 2017 ). Shao et al. found that only the parameter β derived from fractional order calculus models based on Diffusion-weighted imaging (DWI) can distinguish the pathological subtypes of cervical cancer, with SCC having significantly lower β values than ADC (Shao et al. 2022 ). Radiomics is a rapidly growing field of research that utilizes medical images to extract quantitative features, converting them into high-dimensional data for analysis and exploration. This technique enhances our understanding of diseases and provides valuable support for clinical decision-making (McCague et al. 2023 ; Litvin et al. 2021 ). Malignant tumors exhibit significant spatial variation within the tumor at morphological and histopathological levels, including cellularity, vascularization, extracellular matrix, and necrotic components (Davnall et al. 2012 ; Patkulkar et al. 2023 ). As a non-invasive tool, radiomics quantifies intra-tumoral heterogeneity and is widely used in diagnosis, treatment response evaluation, and survival prediction (Liu et al. 2019 ). Recently, magnetic resonance imaging (MRI) radiomics has been employed to differentiate between SCC and AC (Wang et al. 2022 ). To the best of our knowledge, there is currently no research investigating the use of radiomics based on PET/CT images, utilizing diverse machine learning (ML) algorithms, to differentiate between these pathological subtypes. Therefore, the aim of this study was to develop and validate an optimal ML model based on pretherapeutic PET/CT for differentiating between SCC and AC in cervical cancer. Materials and methods Patients This single-center retrospective study was conducted in accordance with the Declaration of Helsinki. Ethical approval (No. SDTHEC2023006030) was obtained from the Institutional Review Board of the Affiliated Cancer Hospital of Shandong First Medical University, and the requirement for written informed consent was waived. The study included patients with a diagnosis of cervical cancer between September 2015 and February 2022. Inclusion criteria were as follows: (1) pathologically confirmed cervical cancer, (2) underwent 18 F-FDG PET/CT, and (3) complete clinical data retrieved from the electronic medical records. Exclusion criteria included: (1) a history of any previous anticancer treatment, (2) pathological types other than squamous cell carcinoma (SCC) and adenocarcinoma (ADC), (3) patients with a diagnosis of other unrelated malignant tumors, (4) presence of extensive abdominal metastasis, (5) poor PET/CT image quality, or (6) primary maximal tumor diameter less than 1.0 cm. Ultimately, a total of 195 patients were enrolled and randomly divided into a training cohort (n = 136) and a validation cohort (n = 59) in a 7:3 ratio. Figure 1 illustrates a flow chart outlining the process of patient selection. The clinical information of the patients, including age, pathology, maximal tumor diameter (MTD) on PET/CT images, menopausal status, lymph node metastasis (LNM), red blood cell count, and others, was collected from electronic medical records. PET/CT examination All patients underwent 18 F-FDG PET/CT using the Philips Gemini TF PET/CT scanner (Phillips Medical Systems, Netherlands) with a standardized scan setup and parameters. The 18 F-FDG was generated by the MINItrace cyclotron from GE Healthcare, ensuring a radiochemical purity of over 95%. Prior to scanning, patients were required to fast for at least 6 hours, and their fasting blood glucose levels were monitored to ensure levels below 140 mg/dL (7.8mmol/L). An intravenous injection of 4.4 MBq/kg of 18 F-FDG was administered, followed by a resting period of approximately 1 hour before a whole-body PET/CT scan was performed. Patients were positioned in a supine position with their arms raised above their head, and the scan covered from the top of the head to the mid-thigh. During the scan, patients were instructed to breathe freely, and respiratory gating techniques were utilized to acquire the images. CT data was used for attenuation correction, and the images were reconstructed using the ordered subset expectation maximization method, resulting in fused images in the transverse, coronal, and sagittal planes. Tumor Segmentation PET images were attenuated, corrected, reconstructed in multiple layers, and then fused with CT images. The resulting images were imported into MIM Maestro version 7.1.7 (MIM Software Inc., Cleveland, OH, USA). The regions of interest (ROIs) were delineated using a fixed threshold value at 42% of the maximum standardized uptake value (SUVmax) of the primary tumor. Regions corresponding to the bladder were manually excluded from the analysis. For the obtained ROIs, various parameters such as metabolic active tumor volume (MTV), mean standardized uptake value (SUVmean), total lesion glycolysis (TLG), and SUVmax were calculated using MIM Software. The contoured ROIs were then transferred to PET and CT images using rigid registration. Another experienced oncologist carefully reviewed and modified the transferred results on a slice-by-slice basis. Feature extraction and normalization A total of 1409 CT radiomics features and 1409 PET radiomics features were extracted from each segmented ROI using the AccuContour software version 3.2 (Manteia Medical Technologies Co. Ltd., Xiamen, China), which is a commercial software that allows for standardized preprocessing of medical imaging data. The radiomics features based on the original images include shape features, first-order intensity histogram features, gray-level co-occurrence matrix (GLCM) features, gray-level run-length matrix (GLRLM) features, gray-level size zone matrix (GLSZM) features, neighboring gray-tone difference matrices (NGTDM), and gray-level dependence matrix (GLDM) features. Feature selection and model development All features were standardized to Z scores with the mean and standard deviation. The Pearson correlation coefficient (PCC) for each feature pair was calculated to evaluate their similarity (29), and if the PCC value exceeded 0.9, one of the features was randomly eliminated. After this process, the dimension of the feature space was reduced, and each feature became independent of one another. Then, the least absolute shrinkage and selection operator (LASSO) regression analysis with 10-fold cross-validation was employed to select the effective radiomics features. Clinical features were selected using logistic regression analysis. Separate models with good prediction performance were built to differentiate pathological subtypes in locally advanced cervical cancer. Figure 2 illustrates the radiomics workflow of this study. Statistical analysis Quantitative data that follow a normal distribution were presented as mean ± standard deviation (s), while qualitative data were expressed as frequency (percentage). The patient characteristics between the training and validation cohorts were compared using various statistical tests, such as the Pearson chi-square test, Fisher's exact test, Student's t-test, and Mann-Whitney U test. Clinical features were selected using univariate and multivariate logistic regression analysis. Six ML classifiers, including logistic regression (LR), Naive Bayes (NB), support vector machine (SVM), k-nearest neighbors (KNN), light gradient boosting machine (lightGBM), and multilayer perceptron neural network (MLP), were used to build a model to differentiate pathological subtypes. The optimal ML model was selected based on its area under the curve (AUC), accuracy (ACC), sensitivity (SEN), and specificity (SPE). The AUC values were compared between different models using the DeLong test. The data analyses were performed using SPSS software (Version 25.0, IBM Corp., Armonk, NY, USA) and R software (Version 3.4.0, R Foundation for Statistical Computing, Vienna, Austria). A two-sided p < 0.05 was considered statistically significant. Results Clinical characteristics and PET metabolic parameters Table 1 presents the clinical characteristics and PET metabolic parameters of 195 patients with LACC. The results of the univariate logistic regression analysis are provided in Table 2 . None of the clinical features or PET metabolic parameters showed significant differentiation ability for the pathological subtypes. Table 1 Clinical characteristics and PET metabolic parameters in the training cohort and validation cohort. Training cohort (N = 136) Validation cohort (N = 59) t /χ 2 / Z P- value Age (years) 54.29 ± 9.97 53.29 ± 11.55 0.612 a 0.541 Pathology 1.919 b 0.166 SCC 115(84.6%) 45(76.3%) AC 21(15.4%) 14(23.7%) Abortion 1.389 b 0.239 NO 77(56.6%) 28(47.5%) YES 59(43.3%) 31(52.5%) MTD (cm) 4.96 ± 1.62 5.22 ± 1.67 -1.062 a 0.290 LNM 0.174 b 0.676 NO 48(35.3%) 19(32.20%) YES 88(64.7%) 40(67.80%) Para-aortic LNM 0.289 b 0.591 NO 106(77.9%) 48(81.4%) YES 30(22.1%) 11(18.6%) Menopause 0.05 b 0.823 NO 53(39.0%) 24(40.7%) YES 83(61.0%) 35(59.3%) SUVmax (SUVbw) 15.72 ± 6.23 16.20 ± 5.82 -0.503 a 0.615 MTV (ml) 29.43(15.53,54.09) 33.42(21.42,59.93) -1.202 c 0.230 SUVmean (SUVbw) 9.25 ± 3.66 9.59 ± 3.44 -0.613 a 0.540 TLG (SUVbw*ml) 253.07(119.97,552.82) 331.91(154.36,662.90) -1.061 c 0.289 WBC count 6.81 ± 2.68 7.24 ± 2.99 0.236 a 0.316 RBC count 4.11 ± 0.49 117.61 ± 23.73 -1.378 a 0.170 Plt count 290.82 ± 99.88 323.24 ± 109.59 -1.934 a 0.056 lymphocyte count 1.66 ± 0.54 1.63 ± 0.55 0.275 a 0.783 neutrophile count 4.51 ± 2.42 4.92 ± 2.41 -1.069 a 0.286 Hb count 120.18 ± 16.51 117.61 ± 23.73 0.755 a 0.452 SCC, squamous cell carcinoma; AC, adenocarcinoma; MTD, maximal tumor diameter; LNM, lymph node metastasis; SUV, standardized uptake value; MTV, metabolic active tumor volume; TLG, total lesion glycolysis; WBC, white blood cell; RBC, red blood cell; WBC, white blood cell; Plt, blood platelet; Hb, hemoglobin; a t value; b χ 2 value ; c Z value; Table 2 Univariate logistic regression analysis to differentiate pathological subtypes in the training cohort. Univariate logistic analysis OR 95% CI P -value Age (years) 1.022 0.986–1.060 0.228 Abortion 0.915 0.647–1.295 0.617 MTD (cm) 0.859 0.680–1.087 0.206 LNM 0.642 0.304–1.355 0.245 Para-aortic LNM 0.739 0.284–1.921 0.535 Menopause 0.974 0.462–2.056 0.945 SUVmax (SUVbw) 0.968 0.909–1.031 0.316 MTV (ml) 0.996 0.986–1.006 0.414 SUVmean (SUVbw) 0.933 0.837–1.040 0.211 TLG (SUVbw*ml) 1.000 0.999–1.001 0.385 WBC count 0.911 0.778–1.067 0.249 RBC count 0.995 0.496–1.997 0.989 Plt count 1.000 0.996–1.003 0.806 lymphocyte count 0.780 0.386–1.575 0.488 neutrophile count 0.929 0.780–1.107 0.410 Hb count 0.991 0.972–1.010 0.755 MTD, maximal tumor diameter; LNM, lymph node metastasis; MTV, metabolic active tumor volume; SUV, standardized uptake value; TLG, total lesion glycolysis; WBC, white blood cell; RBC, red blood cell; WBC, white blood cell; Plt, blood platelet; Hb, hemoglobin; Radiomics features and models development There were 1409 radiomics features extracted from the ROI of CT and PET images, respectively. Among them, a total of 391 and 242 radiomics features were selected from the CT and PET images, respectively, based on the PCC. Subsequently, the LASSO regression analysis was performed to select one CT radiomics feature (Fig. 3 A, 3 C) and five PET radiomics features (Fig. 3 B, 3 D, 4 ). Radiomics models performance Table 3 presents a summary of the prediction performance in distinguishing between AC and SCC using various ML classifiers in the training and validation cohorts. The LightGBM model exhibited superior performance in terms of AUC, ACC, SEN, SPE compared to the other ML models, and was consequently employed as the ML algorithm for differentiating the described pathological subtypes. Table 3 Performance of machine learning classifiers for differentiating pathological subtypes in training and validation cohort. ML DS PET radiomics model CT radiomics model AUC 95% CI ACC SEN SPE AUC 95% CI ACC SEN SPE LR T 0.916 0.852–0.979 0.919 0.714 0.957 0.597 0.441–0.753 0.779 0.429 0.843 V 0.779 0.631–0.928 0.814 0.571 0.889 0.521 0.330–0.711 0.746 0.286 0.909 NB T 0.848 0.739–0.957 0.919 0.667 0.965 0.684 0.549–0.820 0.603 0.762 0.574 V 0.719 0.517–0.921 0.847 0.643 0.911 0.524 0.334–0.712 0.746 0.286 0.909 SVM T 0.941 0.885–0.998 0.941 0.857 0.957 0.612 0.465–0.760 0.632 0.619 0.635 V 0.811 0.647–0.975 0.864 0.786 0.889 0.484 0.287–0.681 0.780 0.214 0.977 KNN T 0.96 0.931–0.989 0.824 1.000 0.791 0.802 0.735–0.870 0.559 1.000 0.478 V 0.700 0.535–0.865 0.847 0.357 1.000 0.417 0.253–0.582 0.763 0.071 1.000 LightGBM T 0.955 0.922–0.988 0.868 0.952 0.852 0.752 0.642–0.862 0.713 0.667 0.761 V 0.851 0.715–0.986 0.915 0.643 1.000 0.513 0.339–0.688 0.661 0.286 0.814 MLP T 0.930 0.877–0.984 0.809 0.905 0.791 0.597 0.440–0.753 0.779 0.429 0.843 V 0.816 0.667–0.965 0.847 0.643 0.911 0.521 0.330–0.711 0.746 0.286 0.909 ML, machine learning; DS, data set; CI, confidence interval; ACC, Accuracy; SEN, Sensitivity; SPE, Specificity; LR, logistic regression; T, training cohort; V, validation cohort; NB, Naive Bayes; SVM, support vector machine; KNN, k-nearest neighbors; lightGBM, light gradient boosting machine; MLP, multilayer perceptron neural network. Figure 5 illustrates the ROC curves of the CT radiomics model, PET radiomics model, and combined model. In the training cohort, the combined radiomics model demonstrated the best differentiation performance (AUC = 0.968), followed by the PET radiomics model (AUC = 0.955), while the differentiation performance of the CT radiomics model was average (AUC = 0.752). The DeLong test indicated that there was no statistically significant difference between the combined radiomics model and the PET radiomics model ( P = 0.347). Nevertheless, both the combined radiomics model and the PET radiomics model outperformed the CT radiomics model significantly ( P < 0.001). In the validation cohort, the PET radiomics model had the best differentiation effectiveness (AUC = 0.851), followed by the combined radiomics model (AUC = 0.842), while the differentiation performance of the CT radiomics model was poor (AUC = 0.513). The DeLong test showed no statistically significant difference between the combined radiomics model and the PET radiomics model ( P = 0.776). However, both the combined radiomics model and the PET radiomics model were significantly better than the CT radiomics model ( P = 0.005 and P = 0.007, respectively). We evaluated the clinical utility of the three radiomics models by plotting decision curve analysis (Fig. 6 ), which revealed that the PET radiomics model outperformed the other models in terms of accuracy and effectiveness. Discussion Pathological diagnosis is considered the gold standard for the detection of cervical cancer, with cervical cytology and cervical biopsy being the primary recommended methods (Rajaram and Gupta 2021 ). Cervical cytology examines abnormal cells obtained from the cervical transformation zone. However, AC may sometimes result in cytological false negatives (Sasieni et al. 2009 ). Moreover, point-to-point biopsies are unable to comprehensively evaluate tumor heterogeneity. Conventional imaging examinations also struggle to differentiate between different pathological subtypes of cervical cancer and analyze tumor heterogeneity. In this study, we did not find clinical features and PET metabolic parameters that could be used to distinguish pathological subtypes. However, we successfully developed six ML models based on PET and CT images respectively, among which the lightGBM model based on PET radiomics features performed excellent in distinguishing AC and SCC. Previous literature has indicated that CT radiomics features exhibit better predictive performance than PET radiomics features in predicting survival, and CT radiomics features are also more abundant than PET features (Liu et al. 2022 ; Kirienko et al. 2018 ). However, this study is contrary to that, as the selected PET radiomics features are significantly more than CT radiomics features, and the performance of the PET radiomics model in distinguishing SCC and AC is notably superior to that of the CT radiomics model. Furthermore, the Delong test show that although there is a slight improvement in performance when combining PET radiomics features with CT radiomics features, the increase in AUC value does not reach statistical significance (the p-values of the training and validation cohort are 0.347 and 0.776, respectively). This finding highlights the importance of functional imaging-based radiomics research in differentiating tumor pathological types, as it suggests that the PET radiomics model can effectively discriminate between SCC and AC without assistance. MRI is widely utilized for primary tumor evaluation in cervical cancer. Technological advancements have introduced various functional MR sequences, including DWI, dynamic contrast-enhanced (DCE) imaging, and perfusion-weighted imaging (Matani et al. 2022 ). The DWI sequence enables differentiation of cellular structures based on water molecule diffusion differences, quantified by the ADC map. DCE imaging is extensively studied and indirectly characterizes tumor perfusion by examining tumor microvasculature. Perfusion-weighted imaging, using a contrast agent such as gadolinium, assesses enhancement and other pharmacokinetic characteristics on T1 or T2-weighted sequences (Matani et al. 2022 ). Wang et al. also achieved good differentiation between SCC and AC using a multi-parameter MRI radiomics model based on ADC, enhanced T1-weighted imaging, and other anatomical and functional sequences (Wang et al. 2022 ). However, the differentiation performance of the multiparametric MRI-based radiomics model (AUC = 0.89) was lower than that of the purely PET radiomics model in the present study (AUC = 0.955). These findings demonstrate the advantages of PET radiomics features over multiparametric MRI radiomics features to a certain extent. However, it is important to consider that the observed differences can be attributed not only to the tumor's heterogeneity, specifically related to molecular mechanisms, cell arrangement, and tissue morphology in the two pathological types, but also to variations in tumor cell metabolism and their close association with different imaging techniques. Campos-Parra et al. found that compared to AC, SCC exhibits higher activation levels of key cancer pathways, such as IL-17, JAK/STAT, and Ras signaling (Campos-Parra et al. 2022 ). High-risk human papilloma virus (HPV)-16 infection is more common in SCC, while HPV-18 and HPV-45 are more frequently observed in AC (Campos-Parra et al. 2022 ; Wang et al. 2019 ). Priego-Hernández et al. discovered that cervical cancer and HPV-16 positive cell lines have increased expression of HIF-1αand glucose metabolism-related genes (GLUT1, LDHA, CAIX, MCT4, and BSG genes) (Priego-Hernández et al. 2022 ). Furthermore, there are significant variations in the expression of glucose metabolism-related genes between SCC and AC (Martinez-Morales et al. 2021 ). Choi et al. demonstrated that tumor FDG uptake is associated with glucose transporters (Glut-1 and Glut-3), with SCC exhibiting higher expression intensity and proportion of Glut-1 compared to AC. Consequently, SCC demonstrates higher SUVmax and stronger FDG uptake capacity (Choi et al. 2015 ). Therefore, the differential expression of pathogenic molecular mechanisms, especially glucose metabolism genes, determines the metabolic differences of tumor cells, while cell arrangement and tissue morphology determine the spatial heterogeneity of tumor cells. The tumor heterogeneity revealed by PET images manifests these metabolic differences and spatial heterogeneity of tumor cells. The tumor heterogeneity revealed by PET images manifests these metabolic differences and spatial heterogeneity of tumor cells. In this study, we employed six ML algorithms to develop models for distinguishing SCC and AC. Among the algorithms, the radiomics model constructed by the LightGBM algorithm exhibited excellent differentiation performance, accuracy, sensitivity, and specificity, with a relatively balanced performance. This finding is consistent with a similar study conducted by Lam et al., who investigated the correlation between radiomics features and tumor mutation burden in glioma based on MRI images using LR, SVM, and six other ML algorithms (Lam et al. 2022 ). They found that the radiomics model constructed by the LightGBM algorithm also demonstrated the best discriminative performance with relatively balanced sensitivity and specificity. Furthermore, researchers have successfully achieved good discriminative performance in distinguishing low-grade and high-grade meningiomas using the LightGBM algorithm for both radiomics and deep learning models (Yang et al. 2022 ). Similarly, Chang et al. constructed LightGBM and convolutional neural network (CNN) models based on non-contrast CT and enhanced images to differentiate thymic epithelial tumors from other anterior mediastinal tumors (Chang et al. 2023 ). The results demonstrated that the LightGBM model outperformed the CNN model in both the non-contrast CT dataset and the enhanced CT dataset. The LightGBM algorithm, which is based on the gradient boosting decision tree (GBDT) model, optimizes the search for optimal split points and the tree growth process. It supports efficient parallel training and possesses advantages such as faster training speed, lower memory consumption, better accuracy, and quick processing of massive data, making it widely applicable. Therefore, ML can better handle complex nonlinear relationships in large-scale datasets and hold great potential for clinical applications (Luo 2021 ). However, it is important to acknowledge that ML models and algorithms also have limitations, including overfitting and lack of interpretability. Overfitting can undermine predictive performance, while the lack of interpretability can hinder the use of ML (Luo et al. 2020 ). Hence, it is essential to prioritize the future optimization of ML algorithms and conduct independent validations to verify their performance. There were several limitations in this study. Firstly, it was a retrospective and preliminary study, which introduced a potential selection bias despite the use of strict inclusion and exclusion criteria. Secondly, HPV status was not available for some patients when retrieving the electronic medical record system. Lastly, this research was conducted in a single-center with a relatively small sample size. To improve the generalizability of the model, it is necessary to investigate a larger sample size from multiple centers in future research. Conclusion In conclusion, the PET-based radiomics model constructed in the present study can be used as an effective tool for differentiating pathological subtypes in patients with locally advanced cervical cancer. Declarations Author contributions RW and YY: contributions to conception and design; HL: contributions to acquisition, analysis, and interpretation of data, and drafting the manuscript; ML: revised critically for important intellectual content; CC and YZ: participated in acquisition of data. All authors have reviewed and approved this version of the article, and due care has been taken to ensure the integrity of the work. Funding This work was supported by the Special Funds Project of Central Guidance on Local Science and Technology Development (ZYYD2022B18), the State Key Laboratory of Pathogenesis, Prevention and Treatment of High Incidence Diseases in Central Asian Fund (SKL-HIDCA-2020-GJ4) and the Key Research and Development Program of Xinjiang Uygur Autonomous Region of China (2022B03019-5). Data availability The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request. Conflict of interest The authors declare that they have no conflict of interest. Informed consent Informed patient consent was not applicable. References Saleh M, Virarkar M, Javadi S et al (2020) Cervical Cancer: 2018 Revised International Federation of Gynecology and Obstetrics Staging System and the Role of Imaging. AJR Am J Roentgenol 214(5):1182-1195. doi:10.2214/ajr.19.21819 Sung H, Ferlay J, Siegel RL et al (2021) Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 71(3):209-249. doi:10.3322/caac.21660 Zheng R, Zhang S, Zeng H et al (2022) Cancer incidence and mortality in China, 2016. Journal of the National Cancer Center 2(1):1-9. doi:10.1016/j.jncc.2022.02.002 Gadducci A, Guerrieri ME, Cosio S (2019) Adenocarcinoma of the uterine cervix: Pathologic features, treatment options, clinical outcome and prognostic variables. Crit Rev Oncol Hematol 135(103-114. doi:10.1016/j.critrevonc.2019.01.006 Smith AJB, Beavis AL, Rositch AF et al (2023) Disparities in Diagnosis and Treatment of Cervical Adenocarcinoma Compared With Squamous Cell Carcinoma: An Analysis of the National Cancer Database, 2004-2017. J Low Genit Tract Dis 27(1):29-34. doi:10.1097/lgt.0000000000000702 Gallardo-Alvarado L, Cantú-de León D, Ramirez-Morales R et al (2022) Tumor histology is an independent prognostic factor in locally advanced cervical carcinoma: A retrospective study. BMC Cancer 22(1):401. doi:10.1186/s12885-022-09506-3 Williams NL, Werner TL, Jarboe EA et al (2015) Adenocarcinoma of the cervix: should we treat it differently? Curr Oncol Rep 17(4):17. doi:10.1007/s11912-015-0440-6 Pujade-Lauraine E, Tan DSP, Leary A et al (2022) Comparison of global treatment guidelines for locally advanced cervical cancer to optimize best care practices: A systematic and scoping review. Gynecol Oncol. doi:10.1016/j.ygyno.2022.08.013 Hu K, Wang W, Liu X et al (2018) Comparison of treatment outcomes between squamous cell carcinoma and adenocarcinoma of cervix after definitive radiotherapy or concurrent chemoradiotherapy. Radiat Oncol 13(1):249. doi:10.1186/s13014-018-1197-5 Fujiwara K, Monk B, Devouassoux-Shisheboran M (2014) Adenocarcinoma of the uterine cervix: why is it different? Curr Oncol Rep 16(12):416. doi:10.1007/s11912-014-0416-y Fan A, Zhang L, Wang C et al (2017) Analysis of clinical factors correlated with the accuracy of colposcopically directed biopsy. Arch Gynecol Obstet 296(5):965-972. doi:10.1007/s00404-017-4500-z Kim SI, Kim SJ, Suh DH et al (2020) Pathologic discrepancies between colposcopy-directed biopsy and loop electrosurgical excision procedure of the uterine cervix in women with cytologic high-grade squamous intraepithelial lesions. J Gynecol Oncol 31(2):e13. doi:10.3802/jgo.2020.31.e13 Nassiri S, Aminimoghaddam S, Sadaghian MR et al (2022) Evaluation of the diagnostic accuracy of the cervical biopsy under colposcopic vision. Eur J Transl Myol 32(4). doi:10.4081/ejtm.2022.10670 Karunya RJ, Tharani P, John S et al (2017) Role of Functional Magnetic Resonance Imaging Derived Parameters as Imaging Biomarkers and Correlation with Clinicopathological Features in Carcinoma of Uterine Cervix. J Clin Diagn Res 11(8):Xc06-xc11. doi:10.7860/jcdr/2017/29165.10426 Winfield JM, Orton MR, Collins DJ et al (2017) Separation of type and grade in cervical tumours using non-mono-exponential models of diffusion-weighted MRI. Eur Radiol 27(2):627-636. doi:10.1007/s00330-016-4417-0 Shao X, An L, Liu H et al (2022) Cervical Carcinoma: Evaluation Using Diffusion MRI With a Fractional Order Calculus Model and its Correlation With Histopathologic Findings. Front Oncol 12:851677. doi:10.3389/fonc.2022.851677 McCague C, Ramlee S, Reinius M et al (2023) Introduction to radiomics for a clinical audience. Clin Radiol 78(2):83-98. doi:10.1016/j.crad.2022.08.149 Litvin AA, Burkin DA, Kropinov AA et al (2021) Radiomics and Digital Image Texture Analysis in Oncology (Review). Sovrem Tekhnologii Med 13(2):97-104. doi:10.17691/stm2021.13.2.11 Davnall F, Yip CS, Ljungqvist G et al (2012) Assessment of tumor heterogeneity: an emerging imaging tool for clinical practice? Insights Imaging 3(6):573-89. doi:10.1007/s13244-012-0196-6 Patkulkar PA, Subbalakshmi AR, Jolly MK et al (2023) Mapping Spatiotemporal Heterogeneity in Tumor Profiles by Integrating High-Throughput Imaging and Omics Analysis. ACS Omega 8(7):6126-6138. doi:10.1021/acsomega.2c06659 Liu Z, Wang S, Dong D et al (2019) The Applications of Radiomics in Precision Diagnosis and Treatment of Oncology: Opportunities and Challenges. Theranostics 9(5):1303-1322. doi:10.7150/thno.30309 Wang W, Jiao Y, Zhang L et al (2022) Multiparametric MRI-based radiomics analysis: differentiation of subtypes of cervical cancer in the early stage. Acta Radiol 63(6):847-856. doi:10.1177/02841851211014188 Rajaram S, Gupta B (2021) Screening for cervical cancer: Choices & dilemmas. Indian J Med Res 154(2):210-220. doi:10.4103/ijmr.IJMR_857_20 Sasieni P, Castanon A, Cuzick J (2009) Screening and adenocarcinoma of the cervix. Int J Cancer 125(3):525-9. doi:10.1002/ijc.24410 Liu S, Li R, Liu Q et al (2022) Radiomics model of 18F-FDG PET/CT imaging for predicting disease-free survival of early-stage uterine cervical squamous cancer. Cancer Biomark 33(2):249-259. doi:10.3233/cbm-210201 Kirienko M, Cozzi L, Antunovic L et al (2018) Prediction of disease-free survival by the PET/CT radiomic signature in non-small cell lung cancer patients undergoing surgery. Eur J Nucl Med Mol Imaging 45(2):207-217. doi:10.1007/s00259-017-3837-7 Matani H, Patel AK, Horne ZD et al (2022) Utilization of functional MRI in the diagnosis and management of cervical cancer. Front Oncol 12:1030967. doi:10.3389/fonc.2022.1030967 Campos-Parra AD, Pérez-Quintanilla M, Martínez-Gutierrez AD et al (2022) Molecular Differences between Squamous Cell Carcinoma and Adenocarcinoma Cervical Cancer Subtypes: Potential Prognostic Biomarkers. Curr Oncol 29(7):4689-4702. doi:10.3390/curroncol29070372 Wang WP, An JS, Yao HW et al (2019) Prevalence and attribution of high-risk HPV in different histological types of cervical cancer. Zhonghua Fu Chan Ke Za Zhi 54(5):293-300. doi:10.3760/cma.j.issn.0529-567x.2019.05.002 Priego-Hernández VD, Arizmendi-Izazaga A, Soto-Flores DG et al (2022) Expression of HIF-1α and Genes Involved in Glucose Metabolism Is Increased in Cervical Cancer and HPV-16-Positive Cell Lines. Pathogens 12(1). doi:10.3390/pathogens12010033 Martinez-Morales P, Morán Cruz I, Roa-de la Cruz L et al (2021) Hallmarks of glycogene expression and glycosylation pathways in squamous and adenocarcinoma cervical cancer. PeerJ 9:e12081. doi:10.7717/peerj.12081 Choi WH, Yoo Ie R, O JH et al (2015) Is the Glut expression related to FDG uptake in PET/CT of non-small cell lung cancer patients? Technol Health Care 23 Suppl 2:S311-8. doi:10.3233/thc-150967 Lam LHT, Chu NT, Tran TO et al (2022) A Radiomics-Based Machine Learning Model for Prediction of Tumor Mutational Burden in Lower-Grade Gliomas. Cancers (Basel) 14(14). doi:10.3390/cancers14143492 Yang L, Xu P, Zhang Y et al (2022) A deep learning radiomics model may help to improve the prediction performance of preoperative grading in meningioma. Neuroradiology 64(7):1373-1382. doi:10.1007/s00234-022-02894-0 Chang CC, Tang EK, Wei YF et al (2023) Clinical radiomics-based machine learning versus three-dimension convolutional neural network analysis for differentiation of thymic epithelial tumors from other prevascular mediastinal tumors on chest computed tomography scan. Front Oncol 13:1105100. doi:10.3389/fonc.2023.1105100 Luo W (2021) Predicting Cervical Cancer Outcomes: Statistics, Images, and Machine Learning. Front Artif Intell 4:627369. doi:10.3389/frai.2021.627369 Luo Y, Chen S, Valdes G (2020) Machine learning for radiation outcome modeling and prediction. Med Phys 47(5):e178-e184. doi:10.1002/mp.13570 Additional Declarations No competing interests reported. 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-3197925","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":221310486,"identity":"843aa0c7-a0f3-4f77-bf17-14432671cc9c","order_by":0,"name":"Huiling Liu","email":"","orcid":"","institution":"The Third Affillated Teaching Hospital of Xinjiang Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Huiling","middleName":"","lastName":"Liu","suffix":""},{"id":221310489,"identity":"3ebc1cd3-ab0d-4bb4-b759-c7575bb94e1b","order_by":1,"name":"Mi Lao","email":"","orcid":"","institution":"Binzhou People's Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Mi","middleName":"","lastName":"Lao","suffix":""},{"id":221310491,"identity":"7b468ee5-a689-404d-943d-a5589db0c057","order_by":2,"name":"Cheng Chang","email":"","orcid":"","institution":"The Third Affillated Teaching Hospital of Xinjiang Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Cheng","middleName":"","lastName":"Chang","suffix":""},{"id":221310492,"identity":"e647a3f4-c76a-42ad-9619-ca363d52902a","order_by":3,"name":"Yalin Zhang","email":"","orcid":"","institution":"The Third Affillated Teaching Hospital of Xinjiang Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Yalin","middleName":"","lastName":"Zhang","suffix":""},{"id":221310495,"identity":"a603536a-95e9-46ce-86fa-e87a1836cb44","order_by":4,"name":"Yong Yin","email":"","orcid":"","institution":"Shandong Cancer Hospital and Institute","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Yong","middleName":"","lastName":"Yin","suffix":""},{"id":221310497,"identity":"2eee0815-a5f7-4309-b76e-62842f29b4bb","order_by":5,"name":"Ruozheng Wang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA30lEQVRIiWNgGAWjYHACxsc/KmzkGJiBzISKGqK0MBsznEkzBmt5cOYYUVrYpBnbDiU2gOx72MJMWL3B7eOPjQvYDqRvOM57gCGxgY2Bv707Ab+WcwmJj2fw3MndcJgvgSFxhwyDxJmzG/BqMTvDcNiAR+JZ7sxmHgOGxDNsDAYSuYS0MLZJ8BgcTpcEa2ljJkYLM5s0T8LhBH5mYrXYn2FjNpxxIM2wn5kv4UDCmWM8BP0i2cP+8MHHfzbybPxnDz78UVEjx9/ei18LEuBhOAAmSQAkKR4Fo2AUjIKRBADbiEcsF2sxcAAAAABJRU5ErkJggg==","orcid":"","institution":"The Third Affillated Teaching Hospital of Xinjiang Medical University","correspondingAuthor":true,"submittingAuthor":false,"prefix":"","firstName":"Ruozheng","middleName":"","lastName":"Wang","suffix":""}],"badges":[],"createdAt":"2023-07-24 04:44:24","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3197925/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3197925/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":40724291,"identity":"607708d7-f794-4e07-ad73-6783ff41f5e7","added_by":"auto","created_at":"2023-07-28 14:18:33","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2005109,"visible":true,"origin":"","legend":"\u003cp\u003eFlow chart of patients selection.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-3197925/v1/f7b3069efd009cd8c8f252de.png"},{"id":40726064,"identity":"4a67c401-fd5b-4130-a5ad-29ce5e210424","added_by":"auto","created_at":"2023-07-28 14:34:33","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":6667198,"visible":true,"origin":"","legend":"\u003cp\u003eWorkflow of this study. A Tumor segmentation. B PET and CT radiomics features extraction . C Feature selection. D Performance evaluation.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-3197925/v1/8749551ee02e798c24d10d04.png"},{"id":40724820,"identity":"28b657aa-af21-4650-89bc-c1a1c822bd79","added_by":"auto","created_at":"2023-07-28 14:26:33","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":4807517,"visible":true,"origin":"","legend":"\u003cp\u003eCT and PET radiomics feature selection using the least absolute shrinkage and selection operator (LASSO) algorithm. A LASSO coefficient profiles of CT radiomics features. B LASSO coefficient profiles of PET radiomics features. C Mean square error path obtained through tenfold cross-validation for CT radiomics feature selection process. D Mean square error path obtained through tenfold cross-validation for PET radiomics feature selection process.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-3197925/v1/11aed0c16ce0fe25553e1da0.png"},{"id":40724296,"identity":"0fa00bdb-7f37-40b9-8909-349d26d61723","added_by":"auto","created_at":"2023-07-28 14:18:33","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":623598,"visible":true,"origin":"","legend":"\u003cp\u003eThe five PET radiomics features are selected and shown.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-3197925/v1/15be02f4d65a69bafef473e3.png"},{"id":40724295,"identity":"d78f2960-b70f-4e28-9357-bdbfd9a4169c","added_by":"auto","created_at":"2023-07-28 14:18:33","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1479068,"visible":true,"origin":"","legend":"\u003cp\u003eThe receiver operating characteristic (ROC) curve of radiomics models in training and validation cohort. A the training cohort. B the validation cohort.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-3197925/v1/0c23dca9c56cc0593ef576dc.png"},{"id":40724292,"identity":"6b1d9705-cd9a-4627-81cb-373326a02ac4","added_by":"auto","created_at":"2023-07-28 14:18:33","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":1289954,"visible":true,"origin":"","legend":"\u003cp\u003eThe decision curve analysis (DCA) of radiomics models in training and validation cohort. (a) the training cohort. (b) the validation cohort.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-3197925/v1/c1684a435e7cb28c05805333.png"},{"id":43641775,"identity":"682a73a3-4ceb-4c6a-9846-95c529bbb611","added_by":"auto","created_at":"2023-09-25 15:22:37","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1581484,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3197925/v1/3de6b54a-17eb-4fe3-969c-591963b8a173.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"18F-FDG PET radiomics-based machine learning model for differentiating pathological subtypes in locally advanced cervical cancer","fulltext":[{"header":"Introduction","content":"\u003cp\u003eBased on the data from GLOBOCAN 2020, cervical cancer ranks as the fourth most prevalent cancer among women globally (Saleh et al. \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). In 2020, there were approximately 604,000 new cases and 342,000 deaths from cervical cancer globally, accounting for approximately 3.1% and 3.4% of the total new cases and deaths from female cancers respectively (Sung et al. \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). According to the latest cancer data in China, in 2015, there were approximately 111,000 reported new cases of cervical cancer and approximately 34,000 deaths, accounting for 6.24% and 3.94% of new cancer cases and deaths in Chinese women, respectively. Notably, cervical cancer has exhibited a trend of affecting younger individuals, with the incidence rate significantly increasing after the age of 25 and peaking between the ages of 50 and 54 (Zheng et al. \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Squamous cell carcinoma (SCC) is the most common pathological subtype of cervical cancer, accounting for about 70\u0026ndash;75% of cases, followed by adenocarcinoma (AC) which represents approximately 10\u0026ndash;25% of cases. The incidence of AC has been observed to increase in recent decades (Gadducci et al. \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Smith et al. \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Although SCC and AC share many risk factors such as HPV infection, multiple sexual partners, and long-term use of oral contraceptives, they have different histological origins (Gallardo-Alvarado et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). However, despite these differences, SCC and AC are still considered as a whole (Williams et al. \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). According to international and country-specific guidelines for locally advanced cervical cancer (LACC), there are no distinct differences in treatment strategies for AC and SCC (Pujade-Lauraine et al. \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Patients with LACC who receive radiation therapy or concurrent chemoradiotherapy have a worse prognosis for AC compared to SCC, highlighting the need for alternative treatment options specifically for AC cases (Hu et al. \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Therefore, there is a critical need to advance research on early-stage differentiation of pathological subtypes of cervical cancer. It is essential to uncover the differences between AC and SCC from multiple perspectives, explore the underlying reasons for these differences, and develop personalized treatment strategies and plans, as this holds significant importance (Fujiwara et al. \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eBiopsy serves as the standard for qualitative diagnosis of cervical cancer. However, biopsy is an invasive procedure associated with risks of bleeding and infection. Point-to-point biopsy performed on larger tumors only evaluate a small portion of the sample, resulting in sampling bias and an inability to comprehensively assess tumor heterogeneity (Fan et al. \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Kim et al. \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Nassiri et al. \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Previous study investigated the value of apparent diffusion coefficient (ADC) in evaluating the pathological subtypes and grading of cervical cancer and found that the ADC value of AC was higher than that of SCC (Karunya et al. \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). However, Winfield et al. discovered that ADC cannot differentiate between SCC and AC (Winfield et al. \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Shao et al. found that only the parameter β derived from fractional order calculus models based on Diffusion-weighted imaging (DWI) can distinguish the pathological subtypes of cervical cancer, with SCC having significantly lower β values than ADC (Shao et al. \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eRadiomics is a rapidly growing field of research that utilizes medical images to extract quantitative features, converting them into high-dimensional data for analysis and exploration. This technique enhances our understanding of diseases and provides valuable support for clinical decision-making (McCague et al. \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Litvin et al. \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Malignant tumors exhibit significant spatial variation within the tumor at morphological and histopathological levels, including cellularity, vascularization, extracellular matrix, and necrotic components (Davnall et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Patkulkar et al. \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). As a non-invasive tool, radiomics quantifies intra-tumoral heterogeneity and is widely used in diagnosis, treatment response evaluation, and survival prediction (Liu et al. \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Recently, magnetic resonance imaging (MRI) radiomics has been employed to differentiate between SCC and AC (Wang et al. \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). To the best of our knowledge, there is currently no research investigating the use of radiomics based on PET/CT images, utilizing diverse machine learning (ML) algorithms, to differentiate between these pathological subtypes. Therefore, the aim of this study was to develop and validate an optimal ML model based on pretherapeutic PET/CT for differentiating between SCC and AC in cervical cancer.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cp\u003e \u003cb\u003ePatients\u003c/b\u003e \u003c/p\u003e \u003cp\u003e This single-center retrospective study was conducted in accordance with the Declaration of Helsinki. Ethical approval (No. SDTHEC2023006030) was obtained from the Institutional Review Board of the Affiliated Cancer Hospital of Shandong First Medical University, and the requirement for written informed consent was waived. The study included patients with a diagnosis of cervical cancer between September 2015 and February 2022. Inclusion criteria were as follows: (1) pathologically confirmed cervical cancer, (2) underwent \u003csup\u003e18\u003c/sup\u003eF-FDG PET/CT, and (3) complete clinical data retrieved from the electronic medical records. Exclusion criteria included: (1) a history of any previous anticancer treatment, (2) pathological types other than squamous cell carcinoma (SCC) and adenocarcinoma (ADC), (3) patients with a diagnosis of other unrelated malignant tumors, (4) presence of extensive abdominal metastasis, (5) poor PET/CT image quality, or (6) primary maximal tumor diameter less than 1.0 cm.\u003c/p\u003e \u003cp\u003eUltimately, a total of 195 patients were enrolled and randomly divided into a training cohort (n\u0026thinsp;=\u0026thinsp;136) and a validation cohort (n\u0026thinsp;=\u0026thinsp;59) in a 7:3 ratio. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e illustrates a flow chart outlining the process of patient selection. The clinical information of the patients, including age, pathology, maximal tumor diameter (MTD) on PET/CT images, menopausal status, lymph node metastasis (LNM), red blood cell count, and others, was collected from electronic medical records.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003ePET/CT examination\u003c/b\u003e \u003c/p\u003e \u003cp\u003eAll patients underwent \u003csup\u003e18\u003c/sup\u003eF-FDG PET/CT using the Philips Gemini TF PET/CT scanner (Phillips Medical Systems, Netherlands) with a standardized scan setup and parameters. The \u003csup\u003e18\u003c/sup\u003eF-FDG was generated by the MINItrace cyclotron from GE Healthcare, ensuring a radiochemical purity of over 95%. Prior to scanning, patients were required to fast for at least 6 hours, and their fasting blood glucose levels were monitored to ensure levels below 140 mg/dL (7.8mmol/L). An intravenous injection of 4.4 MBq/kg of \u003csup\u003e18\u003c/sup\u003eF-FDG was administered, followed by a resting period of approximately 1 hour before a whole-body PET/CT scan was performed. Patients were positioned in a supine position with their arms raised above their head, and the scan covered from the top of the head to the mid-thigh. During the scan, patients were instructed to breathe freely, and respiratory gating techniques were utilized to acquire the images. CT data was used for attenuation correction, and the images were reconstructed using the ordered subset expectation maximization method, resulting in fused images in the transverse, coronal, and sagittal planes.\u003c/p\u003e \u003cp\u003e \u003cb\u003eTumor Segmentation\u003c/b\u003e \u003c/p\u003e \u003cp\u003ePET images were attenuated, corrected, reconstructed in multiple layers, and then fused with CT images. The resulting images were imported into MIM Maestro version 7.1.7 (MIM Software Inc., Cleveland, OH, USA). The regions of interest (ROIs) were delineated using a fixed threshold value at 42% of the maximum standardized uptake value (SUVmax) of the primary tumor. Regions corresponding to the bladder were manually excluded from the analysis. For the obtained ROIs, various parameters such as metabolic active tumor volume (MTV), mean standardized uptake value (SUVmean), total lesion glycolysis (TLG), and SUVmax were calculated using MIM Software. The contoured ROIs were then transferred to PET and CT images using rigid registration. Another experienced oncologist carefully reviewed and modified the transferred results on a slice-by-slice basis.\u003c/p\u003e \u003cp\u003e \u003cb\u003eFeature extraction and normalization\u003c/b\u003e \u003c/p\u003e \u003cp\u003eA total of 1409 CT radiomics features and 1409 PET radiomics features were extracted from each segmented ROI using the AccuContour software version 3.2 (Manteia Medical Technologies Co. Ltd., Xiamen, China), which is a commercial software that allows for standardized preprocessing of medical imaging data. The radiomics features based on the original images include shape features, first-order intensity histogram features, gray-level co-occurrence matrix (GLCM) features, gray-level run-length matrix (GLRLM) features, gray-level size zone matrix (GLSZM) features, neighboring gray-tone difference matrices (NGTDM), and gray-level dependence matrix (GLDM) features.\u003c/p\u003e \u003cp\u003e \u003cb\u003eFeature selection and model development\u003c/b\u003e \u003c/p\u003e \u003cp\u003eAll features were standardized to Z scores with the mean and standard deviation. The Pearson correlation coefficient (PCC) for each feature pair was calculated to evaluate their similarity (29), and if the PCC value exceeded 0.9, one of the features was randomly eliminated. After this process, the dimension of the feature space was reduced, and each feature became independent of one another. Then, the least absolute shrinkage and selection operator (LASSO) regression analysis with 10-fold cross-validation was employed to select the effective radiomics features. Clinical features were selected using logistic regression analysis. Separate models with good prediction performance were built to differentiate pathological subtypes in locally advanced cervical cancer. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e illustrates the radiomics workflow of this study.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eQuantitative data that follow a normal distribution were presented as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (s), while qualitative data were expressed as frequency (percentage). The patient characteristics between the training and validation cohorts were compared using various statistical tests, such as the Pearson chi-square test, Fisher's exact test, Student's t-test, and Mann-Whitney U test. Clinical features were selected using univariate and multivariate logistic regression analysis. Six ML classifiers, including logistic regression (LR), Naive Bayes (NB), support vector machine (SVM), k-nearest neighbors (KNN), light gradient boosting machine (lightGBM), and multilayer perceptron neural network (MLP), were used to build a model to differentiate pathological subtypes. The optimal ML model was selected based on its area under the curve (AUC), accuracy (ACC), sensitivity (SEN), and specificity (SPE). The AUC values were compared between different models using the DeLong test. The data analyses were performed using SPSS software (Version 25.0, IBM Corp., Armonk, NY, USA) and R software (Version 3.4.0, R Foundation for Statistical Computing, Vienna, Austria). A two-sided \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e \u003cb\u003eClinical characteristics and PET metabolic parameters\u003c/b\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e presents the clinical characteristics and PET metabolic parameters of 195 patients with LACC. The results of the univariate logistic regression analysis are provided in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. None of the clinical features or PET metabolic parameters showed significant differentiation ability for the pathological subtypes.\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\u003eClinical characteristics and PET metabolic parameters in the training cohort and validation cohort.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTraining cohort\u003c/p\u003e \u003cp\u003e (N\u0026thinsp;=\u0026thinsp;136)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eValidation cohort\u003c/p\u003e \u003cp\u003e(N\u0026thinsp;=\u0026thinsp;59)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cem\u003et\u003c/em\u003e/χ\u003csup\u003e2\u003c/sup\u003e/\u003cem\u003eZ\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003e\u003cem\u003eP-\u003c/em\u003evalue\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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e54.29\u0026thinsp;\u0026plusmn;\u0026thinsp;9.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e53.29\u0026thinsp;\u0026plusmn;\u0026thinsp;11.55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.612\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.541\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePathology\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.919\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.166\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSCC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e115(84.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e45(76.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e21(15.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e14(23.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAbortion\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.389\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.239\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNO\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e77(56.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e28(47.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYES\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e59(43.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e31(52.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMTD (cm)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.96\u0026thinsp;\u0026plusmn;\u0026thinsp;1.62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5.22\u0026thinsp;\u0026plusmn;\u0026thinsp;1.67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-1.062\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.290\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLNM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.174\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.676\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNO\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e48(35.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e19(32.20%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYES\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e88(64.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e40(67.80%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePara-aortic LNM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.289\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.591\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNO\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e106(77.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e48(81.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYES\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e30(22.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e11(18.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMenopause\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.05\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.823\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNO\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e53(39.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e24(40.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYES\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e83(61.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e35(59.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSUVmax (SUVbw)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e15.72\u0026thinsp;\u0026plusmn;\u0026thinsp;6.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e16.20\u0026thinsp;\u0026plusmn;\u0026thinsp;5.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-0.503\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.615\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMTV (ml)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e29.43(15.53,54.09)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e33.42(21.42,59.93)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-1.202\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.230\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSUVmean (SUVbw)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9.25\u0026thinsp;\u0026plusmn;\u0026thinsp;3.66\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e9.59\u0026thinsp;\u0026plusmn;\u0026thinsp;3.44\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-0.613\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.540\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTLG (SUVbw*ml)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e253.07(119.97,552.82)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e331.91(154.36,662.90)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-1.061\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.289\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWBC count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6.81\u0026thinsp;\u0026plusmn;\u0026thinsp;2.68\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e7.24\u0026thinsp;\u0026plusmn;\u0026thinsp;2.99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.236\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.316\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRBC count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.11\u0026thinsp;\u0026plusmn;\u0026thinsp;0.49\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e117.61\u0026thinsp;\u0026plusmn;\u0026thinsp;23.73\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-1.378\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.170\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePlt count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e290.82\u0026thinsp;\u0026plusmn;\u0026thinsp;99.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e323.24\u0026thinsp;\u0026plusmn;\u0026thinsp;109.59\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-1.934\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.056\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003elymphocyte count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.66\u0026thinsp;\u0026plusmn;\u0026thinsp;0.54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.63\u0026thinsp;\u0026plusmn;\u0026thinsp;0.55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.275\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.783\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eneutrophile count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.51\u0026thinsp;\u0026plusmn;\u0026thinsp;2.42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4.92\u0026thinsp;\u0026plusmn;\u0026thinsp;2.41\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-1.069\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.286\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHb count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e120.18\u0026thinsp;\u0026plusmn;\u0026thinsp;16.51\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e117.61\u0026thinsp;\u0026plusmn;\u0026thinsp;23.73\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.755\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.452\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\u003eSCC, squamous cell carcinoma; AC, adenocarcinoma; MTD, maximal tumor diameter; LNM, lymph node metastasis; SUV, standardized uptake value; MTV, metabolic active tumor volume; TLG, total lesion glycolysis; WBC, white blood cell; RBC, red blood cell; WBC, white blood cell; Plt, blood platelet; Hb, hemoglobin; \u003csup\u003ea\u003c/sup\u003e \u003cem\u003et\u003c/em\u003e value; \u003csup\u003eb\u003c/sup\u003e χ\u003csup\u003e2\u003c/sup\u003e value ; \u003csup\u003ec\u003c/sup\u003e \u003cem\u003eZ\u003c/em\u003e value;\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\u003eUnivariate logistic regression analysis to differentiate pathological subtypes in the training cohort.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"5\" nameend=\"c7\" namest=\"c3\"\u003e \u003cp\u003eUnivariate logistic analysis\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eOR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e95% \u003cem\u003eCI\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e-value\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge (years)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.022\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.986\u0026ndash;1.060\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.228\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAbortion\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.915\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.647\u0026ndash;1.295\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.617\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMTD (cm)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.859\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.680\u0026ndash;1.087\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.206\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLNM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.642\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.304\u0026ndash;1.355\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.245\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePara-aortic LNM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.739\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.284\u0026ndash;1.921\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.535\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMenopause\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.974\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.462\u0026ndash;2.056\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.945\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSUVmax (SUVbw)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.968\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.909\u0026ndash;1.031\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.316\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMTV (ml)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.996\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.986\u0026ndash;1.006\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.414\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSUVmean (SUVbw)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.933\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.837\u0026ndash;1.040\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.211\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTLG (SUVbw*ml)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.999\u0026ndash;1.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.385\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWBC count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.911\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.778\u0026ndash;1.067\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.249\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRBC count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.995\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.496\u0026ndash;1.997\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.989\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePlt count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.996\u0026ndash;1.003\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.806\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003elymphocyte count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.780\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.386\u0026ndash;1.575\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.488\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eneutrophile count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.929\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.780\u0026ndash;1.107\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.410\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHb count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.991\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.972\u0026ndash;1.010\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.755\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\u003eMTD, maximal tumor diameter; LNM, lymph node metastasis; MTV, metabolic active tumor volume; SUV, standardized uptake value; TLG, total lesion glycolysis; WBC, white blood cell; RBC, red blood cell; WBC, white blood cell; Plt, blood platelet; Hb, hemoglobin;\u003c/p\u003e \u003cp\u003e \u003cb\u003eRadiomics features and models development\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThere were 1409 radiomics features extracted from the ROI of CT and PET images, respectively. Among them, a total of 391 and 242 radiomics features were selected from the CT and PET images, respectively, based on the PCC. Subsequently, the LASSO regression analysis was performed to select one CT radiomics feature (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA, \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC) and five PET radiomics features (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB, \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD, \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eRadiomics models performance\u003c/b\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e presents a summary of the prediction performance in distinguishing between AC and SCC using various ML classifiers in the training and validation cohorts. The LightGBM model exhibited superior performance in terms of AUC, ACC, SEN, SPE compared to the other ML models, and was consequently employed as the ML algorithm for differentiating the described pathological subtypes.\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\u003ePerformance of machine learning classifiers for differentiating pathological subtypes in training and validation cohort.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"13\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c11\" colnum=\"11\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c12\" colnum=\"12\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c13\" colnum=\"13\"\u003e\u003c/div\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eML\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eDS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c7\" namest=\"c3\"\u003e \u003cp\u003ePET radiomics model\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c13\" namest=\"c9\"\u003e \u003cp\u003eCT radiomics model\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e95% \u003cem\u003eCI\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eACC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eSEN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eSPE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e95% \u003cem\u003eCI\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003eACC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003eSEN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003eSPE\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.916\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.852\u0026ndash;0.979\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.919\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.714\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.957\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.597\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.441\u0026ndash;0.753\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.779\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.429\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0.843\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.779\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.631\u0026ndash;0.928\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.814\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.571\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.889\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.521\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.330\u0026ndash;0.711\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.746\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.286\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0.909\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eNB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.848\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.739\u0026ndash;0.957\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.919\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.667\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.965\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.684\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.549\u0026ndash;0.820\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.603\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.762\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0.574\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.719\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.517\u0026ndash;0.921\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.847\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.643\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.911\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.524\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.334\u0026ndash;0.712\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.746\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.286\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0.909\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eSVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.941\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.885\u0026ndash;0.998\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.941\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.857\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.957\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.612\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.465\u0026ndash;0.760\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.632\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.619\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0.635\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.811\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.647\u0026ndash;0.975\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.864\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.786\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.889\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.484\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.287\u0026ndash;0.681\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.780\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.214\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0.977\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eKNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.931\u0026ndash;0.989\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.824\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.791\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.802\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.735\u0026ndash;0.870\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.559\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0.478\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.700\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.535\u0026ndash;0.865\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.847\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.357\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.417\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.253\u0026ndash;0.582\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.763\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.071\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eLightGBM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.955\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.922\u0026ndash;0.988\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.868\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.952\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.852\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.752\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.642\u0026ndash;0.862\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.713\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.667\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0.761\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.851\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.715\u0026ndash;0.986\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.915\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.643\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.513\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.339\u0026ndash;0.688\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.661\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.286\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0.814\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eMLP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.930\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.877\u0026ndash;0.984\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.809\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.905\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.791\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.597\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.440\u0026ndash;0.753\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.779\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.429\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0.843\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.816\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.667\u0026ndash;0.965\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.847\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.643\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.911\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.521\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.330\u0026ndash;0.711\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.746\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.286\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0.909\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\u003eML, machine learning; DS, data set; CI, confidence interval; ACC, Accuracy; SEN, Sensitivity; SPE, Specificity; LR, logistic regression; T, training cohort; V, validation cohort; NB, Naive Bayes; SVM, support vector machine; KNN, k-nearest neighbors; lightGBM, light gradient boosting machine; MLP, multilayer perceptron neural network.\u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e illustrates the ROC curves of the CT radiomics model, PET radiomics model, and combined model. In the training cohort, the combined radiomics model demonstrated the best differentiation performance (AUC\u0026thinsp;=\u0026thinsp;0.968), followed by the PET radiomics model (AUC\u0026thinsp;=\u0026thinsp;0.955), while the differentiation performance of the CT radiomics model was average (AUC\u0026thinsp;=\u0026thinsp;0.752). The DeLong test indicated that there was no statistically significant difference between the combined radiomics model and the PET radiomics model (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.347). Nevertheless, both the combined radiomics model and the PET radiomics model outperformed the CT radiomics model significantly (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001). In the validation cohort, the PET radiomics model had the best differentiation effectiveness (AUC\u0026thinsp;=\u0026thinsp;0.851), followed by the combined radiomics model (AUC\u0026thinsp;=\u0026thinsp;0.842), while the differentiation performance of the CT radiomics model was poor (AUC\u0026thinsp;=\u0026thinsp;0.513). The DeLong test showed no statistically significant difference between the combined radiomics model and the PET radiomics model (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.776). However, both the combined radiomics model and the PET radiomics model were significantly better than the CT radiomics model (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.005 and \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.007, respectively).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eWe evaluated the clinical utility of the three radiomics models by plotting decision curve analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e), which revealed that the PET radiomics model outperformed the other models in terms of accuracy and effectiveness.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003ePathological diagnosis is considered the gold standard for the detection of cervical cancer, with cervical cytology and cervical biopsy being the primary recommended methods (Rajaram and Gupta \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Cervical cytology examines abnormal cells obtained from the cervical transformation zone. However, AC may sometimes result in cytological false negatives (Sasieni et al. \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). Moreover, point-to-point biopsies are unable to comprehensively evaluate tumor heterogeneity. Conventional imaging examinations also struggle to differentiate between different pathological subtypes of cervical cancer and analyze tumor heterogeneity. In this study, we did not find clinical features and PET metabolic parameters that could be used to distinguish pathological subtypes. However, we successfully developed six ML models based on PET and CT images respectively, among which the lightGBM model based on PET radiomics features performed excellent in distinguishing AC and SCC.\u003c/p\u003e \u003cp\u003ePrevious literature has indicated that CT radiomics features exhibit better predictive performance than PET radiomics features in predicting survival, and CT radiomics features are also more abundant than PET features (Liu et al. \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Kirienko et al. \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). However, this study is contrary to that, as the selected PET radiomics features are significantly more than CT radiomics features, and the performance of the PET radiomics model in distinguishing SCC and AC is notably superior to that of the CT radiomics model. Furthermore, the Delong test show that although there is a slight improvement in performance when combining PET radiomics features with CT radiomics features, the increase in AUC value does not reach statistical significance (the p-values of the training and validation cohort are 0.347 and 0.776, respectively). This finding highlights the importance of functional imaging-based radiomics research in differentiating tumor pathological types, as it suggests that the PET radiomics model can effectively discriminate between SCC and AC without assistance. MRI is widely utilized for primary tumor evaluation in cervical cancer. Technological advancements have introduced various functional MR sequences, including DWI, dynamic contrast-enhanced (DCE) imaging, and perfusion-weighted imaging (Matani et al. \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). The DWI sequence enables differentiation of cellular structures based on water molecule diffusion differences, quantified by the ADC map. DCE imaging is extensively studied and indirectly characterizes tumor perfusion by examining tumor microvasculature. Perfusion-weighted imaging, using a contrast agent such as gadolinium, assesses enhancement and other pharmacokinetic characteristics on T1 or T2-weighted sequences (Matani et al. \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Wang et al. also achieved good differentiation between SCC and AC using a multi-parameter MRI radiomics model based on ADC, enhanced T1-weighted imaging, and other anatomical and functional sequences (Wang et al. \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). However, the differentiation performance of the multiparametric MRI-based radiomics model (AUC\u0026thinsp;=\u0026thinsp;0.89) was lower than that of the purely PET radiomics model in the present study (AUC\u0026thinsp;=\u0026thinsp;0.955). These findings demonstrate the advantages of PET radiomics features over multiparametric MRI radiomics features to a certain extent. However, it is important to consider that the observed differences can be attributed not only to the tumor's heterogeneity, specifically related to molecular mechanisms, cell arrangement, and tissue morphology in the two pathological types, but also to variations in tumor cell metabolism and their close association with different imaging techniques.\u003c/p\u003e \u003cp\u003eCampos-Parra et al. found that compared to AC, SCC exhibits higher activation levels of key cancer pathways, such as IL-17, JAK/STAT, and Ras signaling (Campos-Parra et al. \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). High-risk human papilloma virus (HPV)-16 infection is more common in SCC, while HPV-18 and HPV-45 are more frequently observed in AC (Campos-Parra et al. \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Wang et al. \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Priego-Hern\u0026aacute;ndez et al. discovered that cervical cancer and HPV-16 positive cell lines have increased expression of HIF-1αand glucose metabolism-related genes (GLUT1, LDHA, CAIX, MCT4, and BSG genes) (Priego-Hern\u0026aacute;ndez et al. \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Furthermore, there are significant variations in the expression of glucose metabolism-related genes between SCC and AC (Martinez-Morales et al. \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Choi et al. demonstrated that tumor FDG uptake is associated with glucose transporters (Glut-1 and Glut-3), with SCC exhibiting higher expression intensity and proportion of Glut-1 compared to AC. Consequently, SCC demonstrates higher SUVmax and stronger FDG uptake capacity (Choi et al. \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Therefore, the differential expression of pathogenic molecular mechanisms, especially glucose metabolism genes, determines the metabolic differences of tumor cells, while cell arrangement and tissue morphology determine the spatial heterogeneity of tumor cells. The tumor heterogeneity revealed by PET images manifests these metabolic differences and spatial heterogeneity of tumor cells. The tumor heterogeneity revealed by PET images manifests these metabolic differences and spatial heterogeneity of tumor cells.\u003c/p\u003e \u003cp\u003eIn this study, we employed six ML algorithms to develop models for distinguishing SCC and AC. Among the algorithms, the radiomics model constructed by the LightGBM algorithm exhibited excellent differentiation performance, accuracy, sensitivity, and specificity, with a relatively balanced performance. This finding is consistent with a similar study conducted by Lam et al., who investigated the correlation between radiomics features and tumor mutation burden in glioma based on MRI images using LR, SVM, and six other ML algorithms (Lam et al. \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). They found that the radiomics model constructed by the LightGBM algorithm also demonstrated the best discriminative performance with relatively balanced sensitivity and specificity. Furthermore, researchers have successfully achieved good discriminative performance in distinguishing low-grade and high-grade meningiomas using the LightGBM algorithm for both radiomics and deep learning models (Yang et al. \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Similarly, Chang et al. constructed LightGBM and convolutional neural network (CNN) models based on non-contrast CT and enhanced images to differentiate thymic epithelial tumors from other anterior mediastinal tumors (Chang et al. \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The results demonstrated that the LightGBM model outperformed the CNN model in both the non-contrast CT dataset and the enhanced CT dataset. The LightGBM algorithm, which is based on the gradient boosting decision tree (GBDT) model, optimizes the search for optimal split points and the tree growth process. It supports efficient parallel training and possesses advantages such as faster training speed, lower memory consumption, better accuracy, and quick processing of massive data, making it widely applicable. Therefore, ML can better handle complex nonlinear relationships in large-scale datasets and hold great potential for clinical applications (Luo \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). However, it is important to acknowledge that ML models and algorithms also have limitations, including overfitting and lack of interpretability. Overfitting can undermine predictive performance, while the lack of interpretability can hinder the use of ML (Luo et al. \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Hence, it is essential to prioritize the future optimization of ML algorithms and conduct independent validations to verify their performance.\u003c/p\u003e \u003cp\u003eThere were several limitations in this study. Firstly, it was a retrospective and preliminary study, which introduced a potential selection bias despite the use of strict inclusion and exclusion criteria. Secondly, HPV status was not available for some patients when retrieving the electronic medical record system. Lastly, this research was conducted in a single-center with a relatively small sample size. To improve the generalizability of the model, it is necessary to investigate a larger sample size from multiple centers in future research.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn conclusion, the PET-based radiomics model constructed in the present study can be used as an effective tool for differentiating pathological subtypes in patients with locally advanced cervical cancer.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor contributions\u0026nbsp;\u003c/strong\u003eRW and YY: contributions to conception and design; HL: contributions to acquisition, analysis, and interpretation of data, and drafting the manuscript; ML: revised critically for important intellectual content; CC and YZ: participated in acquisition of data. All authors have reviewed and approved this version of the article, and due care has been taken to ensure the integrity of the work.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eFunding\u0026nbsp;\u003c/strong\u003eThis work was supported by the Special Funds Project of Central Guidance on Local Science and Technology Development (ZYYD2022B18), the State Key Laboratory of Pathogenesis, Prevention and Treatment of High Incidence Diseases in Central Asian Fund (SKL-HIDCA-2020-GJ4) and the Key Research and Development Program of Xinjiang Uygur Autonomous Region of China (2022B03019-5).\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eData availability\u003c/strong\u003e The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of interest\u003c/strong\u003e The authors declare that they have no conflict of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInformed consent\u003c/strong\u003e Informed patient consent was not applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSaleh M, Virarkar M, Javadi S et al (2020) Cervical Cancer: 2018 Revised International Federation of Gynecology and Obstetrics Staging System and the Role of Imaging. AJR Am J Roentgenol 214(5):1182-1195. doi:10.2214/ajr.19.21819\u003c/li\u003e\n\u003cli\u003eSung H, Ferlay J, Siegel RL et al (2021) Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 71(3):209-249. doi:10.3322/caac.21660\u003c/li\u003e\n\u003cli\u003eZheng R, Zhang S, Zeng H et al (2022) Cancer incidence and mortality in China, 2016. Journal of the National Cancer Center 2(1):1-9. doi:10.1016/j.jncc.2022.02.002\u003c/li\u003e\n\u003cli\u003eGadducci A, Guerrieri ME, Cosio S (2019) Adenocarcinoma of the uterine cervix: Pathologic features, treatment options, clinical outcome and prognostic variables. Crit Rev Oncol Hematol 135(103-114. doi:10.1016/j.critrevonc.2019.01.006\u003c/li\u003e\n\u003cli\u003eSmith AJB, Beavis AL, Rositch AF et al (2023) Disparities in Diagnosis and Treatment of Cervical Adenocarcinoma Compared With Squamous Cell Carcinoma: An Analysis of the National Cancer Database, 2004-2017. J Low Genit Tract Dis 27(1):29-34. doi:10.1097/lgt.0000000000000702\u003c/li\u003e\n\u003cli\u003eGallardo-Alvarado L, Cant\u0026uacute;-de Le\u0026oacute;n D, Ramirez-Morales R et al (2022) Tumor histology is an independent prognostic factor in locally advanced cervical carcinoma: A retrospective study. BMC Cancer 22(1):401. doi:10.1186/s12885-022-09506-3\u003c/li\u003e\n\u003cli\u003eWilliams NL, Werner TL, Jarboe EA et al (2015) Adenocarcinoma of the cervix: should we treat it differently? Curr Oncol Rep 17(4):17. doi:10.1007/s11912-015-0440-6\u003c/li\u003e\n\u003cli\u003ePujade-Lauraine E, Tan DSP, Leary A et al (2022) Comparison of global treatment guidelines for locally advanced cervical cancer to optimize best care practices: A systematic and scoping review. Gynecol Oncol. doi:10.1016/j.ygyno.2022.08.013\u003c/li\u003e\n\u003cli\u003eHu K, Wang W, Liu X et al (2018) Comparison of treatment outcomes between squamous cell carcinoma and adenocarcinoma of cervix after definitive radiotherapy or concurrent chemoradiotherapy. Radiat Oncol 13(1):249. doi:10.1186/s13014-018-1197-5\u003c/li\u003e\n\u003cli\u003eFujiwara K, Monk B, Devouassoux-Shisheboran M (2014) Adenocarcinoma of the uterine cervix: why is it different? Curr Oncol Rep 16(12):416. doi:10.1007/s11912-014-0416-y\u003c/li\u003e\n\u003cli\u003eFan A, Zhang L, Wang C et al (2017) Analysis of clinical factors correlated with the accuracy of colposcopically directed biopsy. Arch Gynecol Obstet 296(5):965-972. doi:10.1007/s00404-017-4500-z\u003c/li\u003e\n\u003cli\u003eKim SI, Kim SJ, Suh DH et al (2020) Pathologic discrepancies between colposcopy-directed biopsy and loop electrosurgical excision procedure of the uterine cervix in women with cytologic high-grade squamous intraepithelial lesions. J Gynecol Oncol 31(2):e13. doi:10.3802/jgo.2020.31.e13\u003c/li\u003e\n\u003cli\u003eNassiri S, Aminimoghaddam S, Sadaghian MR et al (2022) Evaluation of the diagnostic accuracy of the cervical biopsy under colposcopic vision. Eur J Transl Myol 32(4). doi:10.4081/ejtm.2022.10670\u003c/li\u003e\n\u003cli\u003eKarunya RJ, Tharani P, John S et al (2017) Role of Functional Magnetic Resonance Imaging Derived Parameters as Imaging Biomarkers and Correlation with Clinicopathological Features in Carcinoma of Uterine Cervix. J Clin Diagn Res 11(8):Xc06-xc11. doi:10.7860/jcdr/2017/29165.10426\u003c/li\u003e\n\u003cli\u003eWinfield JM, Orton MR, Collins DJ et al (2017) Separation of type and grade in cervical tumours using non-mono-exponential models of diffusion-weighted MRI. Eur Radiol 27(2):627-636. doi:10.1007/s00330-016-4417-0\u003c/li\u003e\n\u003cli\u003eShao X, An L, Liu H et al (2022) Cervical Carcinoma: Evaluation Using Diffusion MRI With a Fractional Order Calculus Model and its Correlation With Histopathologic Findings. Front Oncol 12:851677. doi:10.3389/fonc.2022.851677\u003c/li\u003e\n\u003cli\u003eMcCague C, Ramlee S, Reinius M et al (2023) Introduction to radiomics for a clinical audience. Clin Radiol 78(2):83-98. doi:10.1016/j.crad.2022.08.149\u003c/li\u003e\n\u003cli\u003eLitvin AA, Burkin DA, Kropinov AA et al (2021) Radiomics and Digital Image Texture Analysis in Oncology (Review). Sovrem Tekhnologii Med 13(2):97-104. doi:10.17691/stm2021.13.2.11\u003c/li\u003e\n\u003cli\u003eDavnall F, Yip CS, Ljungqvist G et al (2012) Assessment of tumor heterogeneity: an emerging imaging tool for clinical practice? Insights Imaging 3(6):573-89. doi:10.1007/s13244-012-0196-6\u003c/li\u003e\n\u003cli\u003ePatkulkar PA, Subbalakshmi AR, Jolly MK et al (2023) Mapping Spatiotemporal Heterogeneity in Tumor Profiles by Integrating High-Throughput Imaging and Omics Analysis. ACS Omega 8(7):6126-6138. doi:10.1021/acsomega.2c06659\u003c/li\u003e\n\u003cli\u003eLiu Z, Wang S, Dong D et al (2019) The Applications of Radiomics in Precision Diagnosis and Treatment of Oncology: Opportunities and Challenges. Theranostics 9(5):1303-1322. doi:10.7150/thno.30309\u003c/li\u003e\n\u003cli\u003eWang W, Jiao Y, Zhang L et al (2022) Multiparametric MRI-based radiomics analysis: differentiation of subtypes of cervical cancer in the early stage. Acta Radiol 63(6):847-856. doi:10.1177/02841851211014188\u003c/li\u003e\n\u003cli\u003eRajaram S, Gupta B (2021) Screening for cervical cancer: Choices \u0026amp; dilemmas. Indian J Med Res 154(2):210-220. doi:10.4103/ijmr.IJMR_857_20\u003c/li\u003e\n\u003cli\u003eSasieni P, Castanon A, Cuzick J (2009) Screening and adenocarcinoma of the cervix. Int J Cancer 125(3):525-9. doi:10.1002/ijc.24410\u003c/li\u003e\n\u003cli\u003eLiu S, Li R, Liu Q et al (2022) Radiomics model of 18F-FDG PET/CT imaging for predicting disease-free survival of early-stage uterine cervical squamous cancer. Cancer Biomark 33(2):249-259. doi:10.3233/cbm-210201\u003c/li\u003e\n\u003cli\u003eKirienko M, Cozzi L, Antunovic L et al (2018) Prediction of disease-free survival by the PET/CT radiomic signature in non-small cell lung cancer patients undergoing surgery. Eur J Nucl Med Mol Imaging 45(2):207-217. doi:10.1007/s00259-017-3837-7\u003c/li\u003e\n\u003cli\u003eMatani H, Patel AK, Horne ZD et al (2022) Utilization of functional MRI in the diagnosis and management of cervical cancer. Front Oncol 12:1030967. doi:10.3389/fonc.2022.1030967\u003c/li\u003e\n\u003cli\u003eCampos-Parra AD, P\u0026eacute;rez-Quintanilla M, Mart\u0026iacute;nez-Gutierrez AD et al (2022) Molecular Differences between Squamous Cell Carcinoma and Adenocarcinoma Cervical Cancer Subtypes: Potential Prognostic Biomarkers. Curr Oncol 29(7):4689-4702. doi:10.3390/curroncol29070372\u003c/li\u003e\n\u003cli\u003eWang WP, An JS, Yao HW et al (2019) Prevalence and attribution of high-risk HPV in different histological types of cervical cancer. Zhonghua Fu Chan Ke Za Zhi 54(5):293-300. doi:10.3760/cma.j.issn.0529-567x.2019.05.002 \u003c/li\u003e\n\u003cli\u003ePriego-Hern\u0026aacute;ndez VD, Arizmendi-Izazaga A, Soto-Flores DG et al (2022) Expression of HIF-1\u0026alpha; and Genes Involved in Glucose Metabolism Is Increased in Cervical Cancer and HPV-16-Positive Cell Lines. Pathogens 12(1). doi:10.3390/pathogens12010033\u003c/li\u003e\n\u003cli\u003eMartinez-Morales P, Mor\u0026aacute;n Cruz I, Roa-de la Cruz L et al (2021) Hallmarks of glycogene expression and glycosylation pathways in squamous and adenocarcinoma cervical cancer. PeerJ 9:e12081. doi:10.7717/peerj.12081\u003c/li\u003e\n\u003cli\u003eChoi WH, Yoo Ie R, O JH et al (2015) Is the Glut expression related to FDG uptake in PET/CT of non-small cell lung cancer patients? Technol Health Care 23 Suppl 2:S311-8. doi:10.3233/thc-150967\u003c/li\u003e\n\u003cli\u003eLam LHT, Chu NT, Tran TO et al (2022) A Radiomics-Based Machine Learning Model for Prediction of Tumor Mutational Burden in Lower-Grade Gliomas. Cancers (Basel) 14(14). doi:10.3390/cancers14143492\u003c/li\u003e\n\u003cli\u003eYang L, Xu P, Zhang Y et al (2022) A deep learning radiomics model may help to improve the prediction performance of preoperative grading in meningioma. Neuroradiology 64(7):1373-1382. doi:10.1007/s00234-022-02894-0\u003c/li\u003e\n\u003cli\u003eChang CC, Tang EK, Wei YF et al (2023) Clinical radiomics-based machine learning versus three-dimension convolutional neural network analysis for differentiation of thymic epithelial tumors from other prevascular mediastinal tumors on chest computed tomography scan. Front Oncol 13:1105100. doi:10.3389/fonc.2023.1105100\u003c/li\u003e\n\u003cli\u003eLuo W (2021) Predicting Cervical Cancer Outcomes: Statistics, Images, and Machine Learning. Front Artif Intell 4:627369. doi:10.3389/frai.2021.627369\u003c/li\u003e\n\u003cli\u003eLuo Y, Chen S, Valdes G (2020) Machine learning for radiation outcome modeling and prediction. Med Phys 47(5):e178-e184. doi:10.1002/mp.13570\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":"Locally advanced cervical cancer, positron emission tomography, radiomics, adenocarcinoma, squamous cell carcinoma","lastPublishedDoi":"10.21203/rs.3.rs-3197925/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3197925/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003ePurpose\u003c/h2\u003e \u003cp\u003eTo determine diagnostic performance of \u003csup\u003e18\u003c/sup\u003eF-fluorodeoxyglucose (FDG) positron emission tomography (PET) and computed tomography (CT) radiomics-based machine learning (ML) for classification of cervical adenocarcinoma (AC) and squamous cell carcinoma (SCC).\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003e A total of 195 patients with locally advanced cervical cancer were enrolled in this study, and randomly allocated to training cohort (n\u0026thinsp;=\u0026thinsp;136) and validation cohort (n\u0026thinsp;=\u0026thinsp;59) in a ratio of 7:3. Radiomics features were extracted from pretreatment \u003csup\u003e18\u003c/sup\u003eF-FDG PET/CT and selected by the Pearson correlation coefficient and the least absolute shrinkage and selection operator regression analysis. Six ML classifiers were trained and validated, and the best-performing classifier was selected based on accuracy, sensitivity, specificity, and area under the curve (AUC). The performance of different models was assessed and compared using the DeLong test.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eFive PET and one CT radiomics features were selected and incorporated into the ML classifiers. The PET radiomics model constructed based on the lightGBM algorithm had an accuracy of 0.915 and an AUC of 0.851 (95% CI, 0.715\u0026ndash;0.986) in the validation cohort, which were higher than that of the CT radiomics model (accuracy: 0.661; AUC: 0.513 [95% CI, 0.339\u0026ndash;0.688]). The DeLong test revealed no significant difference in AUC between the combined radiomics model and the PET radiomics model in both the training cohort (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.347) and the validation cohort (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.776).\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eThe \u003csup\u003e18\u003c/sup\u003eF-FDG PET radiomics model can be used as a clinically applicable tool for differentiating pathological subtypes in patients with locally advanced cervical cancer.\u003c/p\u003e","manuscriptTitle":"18F-FDG PET radiomics-based machine learning model for differentiating pathological subtypes in locally advanced cervical cancer","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2023-07-28 14:18:28","doi":"10.21203/rs.3.rs-3197925/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":"f4ca1e45-74fd-4f08-b07f-f04675d4b993","owner":[],"postedDate":"July 28th, 2023","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2023-09-25T15:14:30+00:00","versionOfRecord":[],"versionCreatedAt":"2023-07-28 14:18:28","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-3197925","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3197925","identity":"rs-3197925","version":["v1"]},"buildId":"rHA-KDH7Qsr4HCuvH75dn","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}