Predicting the Molecular Subtypes of 2021 WHO Grade 4 Glioma by a Multiparametric MRI-Based Machine Learning Model

Predicting the Molecular Subtypes of 2021 WHO Grade 4 Glioma by a Multiparametric MRI-Based Machine Learning Model | 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 Predicting the Molecular Subtypes of 2021 WHO Grade 4 Glioma by a Multiparametric MRI-Based Machine Learning Model Wenji Xu, Yangyang Li, Jie Zhang, Zhiyi Zhang, Pengxin Shen, Xiaochun Wang, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5288001/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 develop and validate a machine learning (ML) model using multiparametric MRI for the preoperative differentiation of 2021 World Health Organization (WHO) grade 4 astrocytoma and glioblastoma (GBM) (Task 1), and to stratify grade 4 astrocytoma to distinguish isocitrate dehydrogenase-mutant (IDH-mut) from IDH-wild-type (IDH-wt) (Task 2). Additionally, to evaluate the model’s prognostic value. Materials and methods: We retrospectively analyzed 320 glioma patients from three hospitals. Cases were randomly divided into training and validation sets with a 7:3 ratio. Features were extracted from tumor and edema on contrast-enhanced T1-weighted imaging (CE-T1WI) and T2 fluid-attenuated inversion recovery (T2-FLAIR). Extreme gradient boosting (XGBoost) was utilized for constructing ML, clinical, and combined models. Model performance was evaluated with receiver operating characteristic (ROC) curves, decision curves, and calibration curves. Stability was evaluated using six additional classifiers. Kaplan-Meier (KM) survival analysis and the log-rank test assessed the model’s prognostic value. Results : In Task 1 (grade 4 vs GBM) and Task 2 (IDH-mut grade 4 vs IDH-wt grade 4), the combined model (AUC = 0.911 and 0.854, 0.902 and 0.909) and the optimal ML model (AUC = 0.902 and 0.855, 0.904 and 0.895) significantly outperformed the clinical model (AUC = 0.671 and 0.656, 0.619 and 0.605) in both the training and validation sets. Survival analysis showed the combined model performed similarly to molecular subtype in both tasks ( P = 0.966 and P = 0.793). Conclusion : The multiparametric MRI ML model effectively distinguished grade 4 astrocytoma from GBM and differentiated IDH-mut from IDH-wt grade 4 astrocytoma. Additionally, the model provides reliable survival stratification for glioma patients with various molecular subtypes. Astrocytoma Glioblastoma Magnetic resonance imaging Machine learning Molecular subtype Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction The 2021 World Health Organization (WHO) Classification of central nervous system (CNS) tumors emphasizes the integration of molecular parameters with histological findings for accurate grading[ 1 ]. Notably, isocitrate dehydrogenase-mutant (IDH-mut) grade 2–3 astrocytoma is now reclassified as astrocytoma, IDH-mutant, WHO grade 4, in the presence of homozygous deletion of cyclin-dependent kinase inhibitor A/B (CDKN2A/B), the regardless of necrosis and/or microvascular proliferation. Similarly, IDH wild-type (IDH-wt) grade 2–3 astrocytoma is reclassified as astrocytoma, IDH wild-type, with molecular features of glioblastoma (mGBM), WHO grade 4, if any or a combination of telomerase reverse transcriptase (TERT) promoter mutation, epidermal growth factor receptor (EGFR) amplification, and chromosome + 7 /−10 copy number changes. These molecularly characterized grade 2–3 astrocytoma are high-risk molecular subtypes with poor prognosis[ 1 , 2 ]. Despite poor prognosis, IDH-wt grade 4 astrocytoma patients demonstrated longer overall survival (OS) and progression-free survival (PFS) compared to glioblastoma (GBM) patients[ 3 ]. The two IDH variants of grade 4 astrocytoma exhibit different biological behaviors and clinical outcomes, with IDH-mut astrocytoma showing less aggressive behavior and better prognosis under similar treatment protocols[ 2 , 4 ]. Therefore, accurately distinguishing the different molecular subtypes of WHO grade 4 glioma is highly relevant for prognostic stratification and personalized treatment. However, current molecular diagnostics depend on invasive biopsies, necessitating a reliable non-invasive method for predicting the molecular subtypes of glioma. Magnetic Resonance Imaging (MRI) is extensively utilized for glioma diagnosis and monitoring but falls short in revealing histological and molecular details. Radiomics, by extracting high-throughput image features, offers insights into tumor heterogeneity, thereby enhancing diagnostic and therapeutic accuracy[ 5 , 6 ]. Machine learning (ML) algorithms can process radiomics data to predict tissue characteristics and identify molecular features of glioma[ 7 , 8 ]. Previous studies have developed MRI-based radiomics and ML models for glioma grading, mostly adhering to the 2016 or earlier WHO classifications, or focusing on glioma grading[ 9 – 17 ]. The value of differentiating grade 4 glioma from GBM, and further distinguishing grade 4 IDH-mut from IDH-wt astrocytoma, remains under investigated. Therefore, our study aims to construct a ML model using multiparametric MRI to differentiate grade 4 astrocytoma from GBM (Task 1), and further stratify grade 4 astrocytoma into IDH-mut and IDH-wt subtypes (Task 2). Additionally, we seek to analyze its corresponding prognostic value in OS. Materials and Methods Patient population This study conducted in accordance with the Helsinki Declaration ethical standards. The Ethics Board of the First Hospital of Shanxi Medical University (approval number: KYYJ-2023-058) has approved for this retrospective study. Given the retrospective nature of the study and the anonymous nature of the data, the requirement for informed consent from each patient has been waived. This study collected data from 320 patients with pathologically confirmed glioma from three institutions (First Hospital of Shanxi Medical University, Shanxi Provincial People's Hospital and Shanxi Bethune Hospital) based on inclusion and exclusion criteria between February 2011 and December 2023. Patients were divided into training and validation sets in a 7:3 ratio. OS was defined as the time from initial diagnosis to death or last follow-up, with survival data available for 312 patients. Inclusion criteria: 1) Histological diagnosis of diffuse glioma; 2) No prior radiotherapy, chemotherapy, or surgery before MRI examination; 3) No history of craniocerebral surgery or other systemic malignancies; 4) Availability of molecular information, including IDH, 1p/19q, methyl guanine methyl transferase promoter methylation (MGMTmet), CDKN2A/B, EGFR, TERT, and chromosome 7/10 status. Exclusion criteria: 1) 2021 WHO CNS grade 2-3 astrocytoma; 2) Oligodendroglioma; 3) Incomplete or poor-quality MRI images; 4) Incomplete molecular information. A flowchart of patient selection and machine learning classification is shown in Figure 1 . Machine learning-based classification We constructed two independent binary classification tasks: 1) distinguishing grade 4 astrocytoma from GBM (Task 1: grade 4 vs GBM), and 2) stratifying grade 4 astrocytoma into IDH-mut and IDH-wt subtypes (Task 2: IDH-mut grade 4 vs IDH-wt grade 4). Clinical-radiological characteristics collection Clinical-radiological characteristics included gender, age, MGMTmet, treatment, Karnofsky Performance Status (KPS) score, tumor number, tumor margin, intratumoral hemorrhage, intratumoral necrosis, peritumoral edema, maximum diameter, midline shift, enhancement pattern, enhancement quality, tumor crosses midline (TCM), edema crosses midline (ECM), cortical involvement, deep white matter invasion, pial invasion, and ependymal invasion. Molecular biomarker detection methods are described in supplementary materials. MRI image acquisition and preprocessing MR Images were obtained using 3T scanners (Signa HDxt from GE Healthcare, USA and Skyra from Siemens Healthineers, Germany) with axial T2-weighted imaging fluid attenuated inversion recovery (T2WI-FLAIR) and contrast-enhanced T1WI (CE-T1WI). Scanning parameters are described in supplementary materials. Images were preprocessed to standardize signal intensity on a scale of 100, corrected for N4ITK bias fields, and resampled to a voxel size of 1 mm × 1 mm × 1 mm, with voxel intensity discretized using a fixed bin width of 25. Image preprocessing used FeAture Explorer V.0.5.7 (FAE, https:// github.com/salan668/FAE) and 3Dslicer version 5.7.20240325 (https://www.slicer.org). Image segmentation, feature extraction and selection T2WI-FLAIR images were registered to CE-T1WI using a rigid registration algorithm. A radiologist with 8 years of experience manually delineated tumor and edema areas on CE-T1WI and T2-FLAIR, respectively. The volume of interest (VOIs) included CE-T1WI tumor (T1C tumor), CE-T1WI edema (T1C edema), and T2WI-FLAIR tumor (T2F tumor). Thirty patients underwent repeat segmentation to calculate the intraclass correlation coefficient (ICC) for inter-observer agreement. Feature selection was performed on the training set. Features with an ICC ≥ 0.75 were selected, followed by z-score normalization. Normality was assessed using the Shapiro-Wilk test. Features were retained if P < 0.05 using independent sample t-tests or Mann-Whitney U tests, depending on distribution. The least absolute shrinkage and selection operator (LASSO) selected the best regularization parameters through ten-fold cross-validation. Image segmentation utilized ITK-SNAP software (http://www.itksnap.org/, version 4.0.0). Features were extracted using FAE. The radiomics workflow is shown in Figure 2 . Region of interest (ROI) segmentation is shown in Supplementary Figure 1 . Construction of the ML model We employed Extreme Gradient Boosting (XGBoost) to build ML models for two tasks. Based on the selected features of each sequence and their combination, ML models for single and combined sequences were constructed. The ML model with the best predictive performance on the validation set was selected as the optimal ML model, and the Rad-score was calculated. Construction of the clinical model Univariate logistic regression (LR) analysis screened clinical-radiological characteristics, and variables with P < 0.05 were included in multivariate LR analysis. Significant variables were used to establish the clinical model using XGBoost. Construction of the combined model and nomogram A combined model was constructed using XGBoost based on the Rad-score and independent clinical-radiological risk factors. A nomogram was generated using LR to visually discriminate between grade 4 astrocytoma and GBM. Model evaluation and model comparison The model’s performance in predicting WHO grade 4 glioma molecular subtypes was evaluated using the Receiver Operating Characteristic (ROC) curve, assessing Area Under the Curve (AUC), sensitivity (SEN), specificity (SPE) and accuracy (ACC). The DeLong's test compared predictive performance between models, with statistical significance set at P < 0.05. Decision and calibration curves evaluated model calibration and clinical utility. Various ML algorithms, including LR, support vector machines (SVM), multilayer perceptrons (MLP), linear discriminant analysis (LDA), random forest (RF), and Naive Bayes (NB), were used to evaluate the generalization and stability of the combined model. Survival analysis Kaplan-Meier (KM) survival analysis and log-rank test evaluate the prognostic value of the molecular subtype prediction model based on the combined model. The Z-test compared the prognostic value between the combined model and the molecular subtype. Statistical analysis Statistical analyses were performed using R version 4.2.3 (http://www.Rproject.org). Numerical variables were expressed as mean ± standard deviations. The Shapiro-Wilk test assessed normality. Independent sample t-test or Mann-Whitney U-test was used based on the distribution. Categorical variables were expressed as frequencies (percentages) and evaluated using Pearson’s chi-square test or Fisher’s exact test. P < 0.05 was considered statistically significant. Results Clinical-radiological baseline characteristics Among 320 patients enrolled based on the 2021 WHO CNS tumor classification, there were 196 GBMs, 41 IDH-mut grade 4, and 83 IDH-wt grade 4 astrocytomas. For Task 1 (grade 4 vs GBM), 224 and 96 cases were in the training and validation set, respectively. For Task 2 (IDH-mut grade 4 vs IDH-wt grade 4), 118 and 51 cases were in the training and validation set. Of the 312 patients with survival information, there were 189 GBMs, 41 IDH-mut grade 4, and 82 IDH-wt grade 4 astrocytomas. Except for MGMTmet ( P = 0.027) in Task 1 and peritumoral edema ( P = 0.033) in Task 2, there were no significant differences in clinical-radiological characteristics between the training and validation sets for both tasks. Baseline characteristics are detailed in Table 1 . Table 1 Clinical-radiological baseline characteristics between training and validation sets of the two tasks Task 1 (grade 4 vs GBM) Task 2 (IDH-mut grade 4 vs IDH-wt grade 4 ) Variables Training set (n = 224) Validation set (n = 96) P -value Training set (n = 87) Validation set (n = 37) P -value Gender Male Female 130 (58.04%) 94 (41.96%) 53 (55.21%) 43 (44.79%) 0.712 48 (55.17%) 39 (44.83%) 23 (62.16%) 14 (37.84%) 0.554 Age(years) 53.76 ± 13.48 52.23 ± 13.62 0.201 49.26 ± 14.71 52.11 ± 14.21 0.216 MGMTmet Yes No 136 (60.71%) 88 (39.29%) 45 (46.88%) 51 (53.12%) 0.027 64 (73.56%) 23 (26.44%) 21 (56.76%) 16 (43.24%) 0.090 Treatment Surgery Combination therapy 88 (39.29%) 136 (60.71%) 38 (39.58%) 58 (60.42%) 1.000 38 (43.68%) 49 (56.32%) 19 (51.35%) 18 (48.65%) 0.440 KPS 78.64 ± 12.08 79.23 ± 11.46 0.913 78.46 ± 12.11 79.78 ± 6.68 0.884 Tumor number Single Multiple 118 (52.68%) 106 (47.32%) 54 (56.25%) 42 (43.75%) 0.625 59 (67.82%) 28 (32.18%) 21 (56.76%) 16 (43.24%) 0.305 Tumor margin Clear Non-clear 97 (43.30%) 127 (56. 70%) 44 (45.83%) 52 (54.17%) 0.713 31 (35.63%) 56 (64.37%) 20 (54.05%) 17 (45.95%) 0.073 Intratumoral hemorrhage Yes No 54 (24.11%) 170 (75.89%) 23 (23.96%) 73 (76.04%) 1.000 15 (17.24%) 72 (82.76%) 8 (21.62%) 29 (78.38%) 0.617 Intratumoral necrosis Yes No 180 (80.36%) 44 (19.64%) 80 (83.33%) 16 (16.67%) 0.640 61 (70.11%) 26 (29.89%) 30 (81.08%) 7 (18.92%) 0.269 Peritumoral edema Yes No 212 (94.64%) 12 (5.36%) 94 (97.92%) 2 (2.08%) 0.243 76 (87.36%) 11 (12.64%) 37 (100.00%) 0 (0.00%) 0.033 Maximum diameter 4.73 ± 1.82 4.74 ± 1.96 0.840 4.66 ± 1.94 4.64 ± 1.52 0.960 Midline shift Yes No 111 (47.45%) 113 (52.55%) 48 (50.00%) 48 (50.00%) 0.456 44 (50.57%) 43 (49.43%) 22 (59.46%) 15 (40.54%) 0.433 Enhancement pattern No reinforcement Annular reinforcement Nodular enhancement Mixed reinforcement 32 (14.29%) 103 (45.98%) 30 (13.39%) 59 (26.34%) 10 (10.42%) 41 (42.71%) 10 (10.42%) 35 (36.46%) 0.316 24 (27.59%) 36 (41.38%) 13(14.94%) 14 (16.09%) 6 (16.22%) 15 (40.54%) 6 (16.22%) 10 (27.03%) 0.392 Enhancement quality No reinforcement Reinforcement 31 (13.84%) 193 (86.16%) 9 (9.38%) 87 (90.62%) 0.357 24 (27.59%) 63 (72.41%) 5 (13.51%) 32 (86.49%) 0.108 TCM Yes No 183 (81.70%) 41 (18.30%) 75 (78.12%) 21 (21.88%) 0.537 67 (77.01%) 20 (22.99%) 29 (78.38%) 8 (21.62%) 1.000 ECM Yes No 177 (79.02%) 47 (20.98%) 66 (68.75%) 30 (31.25%) 0.063 60 (68.97%) 27 (31.03%) 29 (78.38%) 8 (21.62%) 0.384 Cortical involvement Yes No 66 (29.46%) 158 (70.54%) 28 (29.17%) 68 (70.83%) 1.000 25 (28.74%) 62 (71.26%) 11 (29.73%) 26 (70.27%) 1.000 Deep white matter invasion Yes No 62 (27.68%) 162 (72.32%) 28 (29.17%) 68 (70.83%) 0.788 23 (26.44%) 64 (73.56%) 10 (27.03%) 27 (72.97%) 1.000 Pial invasion Yes No 132 (58.93%) 92(41.07%) 59 (61.46%) 37 (38.54%) 0.710 41 (47.13%) 46 (52.87%) 23 (62.16%) 14 (37.84%) 0.169 Ependymal invasion Yes No 168 (75.00%) 56 (25.00%) 72 (75.00%) 24 (25.00%) 1.000 60 (68.97%) 27 (31.03%) 30 (81.08%) 7 (18.92%) 0.193 GBM: glioblastoma; IDH-mut: Isocitrate dehydrogenase-mutant; IDH-wt: Isocitrate dehydrogenase wild-type; MGMTmet: methyl guanine methyl transferase promoter methylation; KPS: karnofsky kerformance status score; TCM: tumor across midline, ECM: edema across midline. Construction of the clinical model Univariate and multivariate LR results are presented in Table 2 . For Task 1, age ( P = 0.014) and MGMTmet ( P = 0.028) were significant. For Task 2, ECM ( P = 0.005) and deep white matter invasion ( P = 0.003) were significant. Table 2 Univariate and multivariate logistic regression analysis in the training sets of the two tasks Task 1 (grade 4 vs GBM) Task 2 (IDH-mut grade 4 vs IDH-wt grade 4 ) Variables Univariate analysis Multivariate analysis Univariate analysis Multivariate analysis OR (95%CI) P OR (95%CI) P OR (95%CI) P OR (95%CI) P Gender 1.103(0.638–1.918) 0.727 1.120 (0.458–2.729) 0.802 Age(years) 1.034 (1.013–1.056) 0.002 1.028 (1.006–1.052) 0.014 0.979 (0.948–1.009) 0.170 MGMTmet 0.564 (0.316–0.991) 0.049 0.503 (0.269–0.918) 0.028 1.700 (0.612–5.247) 0.326 Treatment 1.487 (0.856–2.585) 0.159 0.831 (0.340–2.035) 0.684 KPS 0.987 (0.963–1.010) 0.271 1.003 (0.967–1.044) 0.863 Tumor number 1.557 (0.903–2.706) 0.113 0.673 (0.244–1.745) 0.426 Tumor margin 0.971 (0.561–1.676) 0.917 1.167 (0.465–3.032) 0.745 Intratumoral hemorrhage 1.776 (0.925–3.555) 0.093 0.940 (0.268–2.959) 0.918 Intratumoral necrosis 1.696 (0.867–3.308) 0.120 1.643 (0.617–4.752) 0.335 Peritumoral edema 5.480 (1.581–25.261) 0.013 1.960 (0.424–10.969) 0.405 2.625 (0.621–18.028) 0.238 Maximum diameter 1.062 (0.915–1.238) 0.432 0.888 (0.699–1.118) 0.316 Midline shift 0.972 (0.565–1.671) 0.918 1.184 (0.488–2.895) 0.709 Enhancement pattern 1.465 (1.117–1.944) 0.006 1.085 (0.765–1.552) 0.649 1.492 (0.966–2.348) 0.075 Enhancement quality 4.333 (1.97–10.127) <0.001 2.568 (0.832–8.27) 0.105 1.396 (0.518–4.067) 0.521 TCM 1.387 (0.692–2.752) 0.350 0.733 (0.264–2.113) 0.555 ECM 1.462 (0.756–2.804) 0.254 0.338 (0.129–0.866) 0.025 0.211 (0.067–0.602) 0.005 Cortical involvement 0.892 (0.496–1.621) 0.705 1.400(0.526–3.651) 0.493 Deep white matter invasion 1.024 (0.562–1.894) 0.939 3.594 (1.344–9.956) 0.012 5.688 (1.912–18.609) 0.003 Pial invasion 1.423 (0.822–2.465) 0.208 2.217 (0.909–5.575) 0.084 Ependymal invasion 1.350 (0.725–2.496) 0.340 2.333 (0.857–7.121) 0.112 GBM: glioblastoma; IDH-mut: Isocitrate dehydrogenase-mutant; IDH-wt: Isocitrate dehydrogenase wild-type; OR: odds ratio; CI: confidence interval; MGMTmet: methyl guanine methyl transferase promoter methylation; KPS: karnofsky performance status score; TCM: tumor crosses midline, ECM: edema crosses midline. The AUCs of the clinical models were 0.671 and 0.619 (training set) and 0.656 and 0.605 (validation set) for Tasks 1 and Tasks 2, respectively (Fig. 3 and Table 3 ). Table 3 The diagnostic performance of clinical model, ML model, and combined model in the training and validation sets of the two tasks Task 1 (grade 4 vs GBM) Task 2 (IDH-mut grade 4 vs IDH-wt grade 4 ) Model Set AUC (95% CI) Cut-off SEN SPE ACC AUC (95% CI) Cut-off SEN SPE ACC T1C Edema Training 0.843 (0.787–0.892) 0.667 0.707 0.833 0.754 0.838 (0.745–0.912) 0.399 0.700 0.842 0.793 Validation 0.829 (0.734–0.911) 0.505 0.875 0.750 0.823 0.792 (0.618–0.941) 0.385 0.818 0.692 0.730 T1C Tumor Training 0.825 (0.768–0.877) 0.681 0.700 0.786 0.732 0.735 (0.623–0.844) 0.382 0.800 0.614 0.678 Validation 0.722 (0.608–0.830) 0.584 0.821 0.625 0.740 0.750 (0.515–0.926) 0.373 0.818 0.654 0.703 T2F Tumor Training 0.777 (0.711–0.838) 0.558 0.886 0.571 0.768 0.722 (0.612–0.827) 0.395 0.733 0.649 0.678 Validation 0.629 (0.491–0.741) 0.549 0.821 0.425 0.656 0.640 (0.421–0.817) 0.370 0.909 0.423 0.568 Optimal ML Training 0.902 (0.861–0.939) 0.619 0.821 0.821 0.821 0.904 (0.836–0.953) 0.375 0.933 0.789 0.839 Validation 0.855 (0.769–0.928) 0.675 0.732 0.875 0.792 0.895 (0.769–0.991) 0.380 0.909 0.846 0.864 Clinical Training 0.671 (0.596–0.743) 0.664 0.693 0.583 0.652 0.619 (0.515–0.721) 0.400 0.467 0.772 0.667 Validation 0.656 (0.492–0.766) 0.680 0.357 0.875 0.573 0.605 (0.453–0.766) 0.400 0.364 0.846 0.703 Combined Training 0.911 (0.869–0.945) 0.531 0.871 0.786 0.839 0.902 (0.838–0.952) 0.398 0.933 0.789 0.839 Validation 0.854 (0.769–0.928) 0.674 0.750 0.825 0.781 0.909 (0.794–0.987) 0.477 0.909 0.846 0.865 GBM: glioblastoma; IDH-mut: Isocitrate dehydrogenase-mutant; IDH-wt: Isocitrate dehydrogenase wild-type; AUC: area under curve; CI: confidence interval; SEN: sensitivity; SPE: specificity; ACC: accuracy; T1C: contrast-enhanced T1-weighted imaging; T2F: T2-weighted imaging fluid attenuated inversion recovery; ML: machine learning. Construction of the ML model A total of 1688 radiomics features were extracted. After feature selection and reduction, 11 T1C edema, 14 T1C tumor, and 10 T2F tumor features for Task 1 and 9 T1C edema, 6 T1C tumor, and 3 T2F tumor features for Task 2 were retained. These features were then amalgamated separately, resulting in final sets of 35 and 18 features for constructing combined-sequence ML models for the two tasks. The combined-sequence ML model achieved the highest AUCs in the validation set for Task 1 (training set = 0.902, validation set = 0.855) and Task 2 (training set = 0.904, validation set = 0.895), making them the optimal ML models (Fig. 3 and Table 3 ). Rad-score was calculated. Construction of the combined model and nomogram In Task 1, the combined model had the highest AUC in the training set (0.911) but was slightly lower than the optimal ML model in the validation set (0.854). In Task 2, the combined model had the highest AUC in the validation set (0.909) but lower than the optimal ML model in the training set (0.902) (Fig. 3). Nomogram based on the combined model further facilitates individualized discrimination between grade 4 astrocytoma and GBM (Fig. 4 a). Their representative MR Images and case nomograms are shown in Fig. 4 b-e. Model evaluation and comparison DeLong's test demonstrated significant differences in AUCs between the optimal ML model and the combined model, as compared to the clinical model for both tasks in the training and validation sets (Task 1: all P < 0.001; Task 2: training set P < 0.001, validation set P = 0.011 and P = 0.004). No significant differences were found between the combined and optimal ML models in the training ( P = 0.148, P = 0.907) and validation ( P = 0.934, P = 0.514) sets for both tasks. Delong’s test results are shown in Fig. 3c, d, g, h . The combined model demonstrated good calibration in both tasks, though moderate in the validation set for Task 2 (Fig. 5 a, b, e, f). Decision curve analysis (DCA) indicated that the combined model and the optimal ML model provided significantly better net benefits in predicting glioma molecular subtypes compared to the clinical model. For Task 1, threshold probabilities were 0.12 to 0.92 and 0.02 to 0.96 (training set, Fig. 5 c), and 0.18 to 0.86 and 0.08 to 0.92 (validation set, Fig. 5 d). For Task 2, threshold probabilities ranged from 0.04 to 0.70 and 0.02 to 0.98 (training set, Fig. 5 g), and from 0.06 to 0.66 and 0.02 to 0.96 (validation set, Fig. 5 h). The combined model offered a limited net benefit compared to the optimal ML model, which offered greater net benefits across multiple threshold ranges. We used six other ML methods for building the combined model to test its stability and reliability. In both tasks, the RF model achieved prediction performances of 1.0 and 0.998 in the training sets, outperforming other models, while XGBoost showed slightly lower performance than some others. The remaining models performed similarly, with validation set AUCs ranging from 0.854 to 0.881 (Task 1) and 0.958 (Task 2) to 0.979, respectively (Fig. 6 a-h). DeLong's test revealed no statistically significant differences in AUCs among models in the training sets, except for RF and XGBoost (Fig. 6 i, k), and no significant differences across all models in the validation sets (Fig. 6 j, l). Survival analysis Using the training set cutoff values from the combined model for the two tasks (0.531 and 0.398), patients were divided into high-risk and low-risk groups. In Task 1 (489 vs 993 days) and Task 2 (721 vs 1297 days), the average OS of low-risk patients was significantly longer than that of high-risk patients, reflecting the different prognoses associated with various molecular subtypes. KM analysis revealed significant differences between these groups for the combined model and molecular subtype in Task 1 and Task 2 (all P < 0.001). Z-test indicated no statistically significant difference in prognostic value between the molecular subtype and combined model in both tasks ( P = 0.966, P = 0.793) (Fig. 7 ). Discussion This study constructed two machine learning tasks to predict molecular subtypes of 2021 WHO grade 4 glioma using multiparametric MRI, clinical-radiological characteristics, and their combination. The ML model performed well in distinguishing grade 4 astrocytoma from GBM and discriminating IDH-mut grade 4 astrocytoma from IDH-wt grade 4 astrocytoma. Additionally, the combined model effectively stratified cases into high-risk and low-risk groups according to OS, with prognostic performance comparable to molecular subtype. Predictive Value of the Clinical Model Our study identified age and MGMTmet status as significant predictors for distinguishing grade 4 astrocytoma from GBM (Task 1) with AUCs of 0.671 (training set) and 0.656 (validation set). ECM and deep white matter invasion were significant predictors for differentiating IDH-mut from IDH-wt grade 4 astrocytoma (Task 2) with AUCs of 0.619 (training set) and 0.605 (validation set). GBM patients were older on average than grade 4 astrocytoma patients (55.3 vs 50.1 years), and IDH-wt patients were older than IDH-mut patients (52.3 vs 45.6 years), consistent with a previous study[ 18 ]. Older age may be associated with higher malignancy potential. Studies indicated that younger age is associated with better prognosis, whereas older age is linked to poor survival in adult glioma patients[ 19 , 20 ], and our study indirectly supports this. Additionally, a higher proportion of MGMTmet was observed in grade 4 astrocytoma compared to GBM (67.5% vs 49.0%), suggesting it may be associated with less aggressive tumor behavior. Previous studies have shown that MGMTmet is linked to an improved response to temozolomide and longer OS[ 2 , 21 , 22 ], consistent with our findings. IDH-wt patients exhibited higher rates of ECM (79.5% vs 56.1%) and lower rates of deep white matter invasion (16.9% vs 46.3%) compared to IDH-mut patients. Despite these findings, the clinical model demonstrated limited predictive power, underscoring the challenge of relying solely on clinical-radiological features for precise molecular subtypes of 2021 WHO glioma. This limitation is likely due to the inherent heterogeneity of glioma, where clinical and radiological characteristics may not fully capture the underlying molecular alterations. Predictive Value of the ML Model The ML model significantly outperformed the clinical models in distinguishing grade 4 astrocytoma from GBM and discriminating IDH-mut from IDH-wt grade 4 astrocytoma. The optimal ML model for both tasks has strong predictive performance in the training (AUC = 0.902 and 0.904) and validation (0.855 and 0.895) sets. Currently, only Wei et al.[ 23 ] conducted a relevant study in which they developed a subregion-based MRI RadioFusionOmics model to discriminate between grade 4 astrocytoma and GBM, achieving AUCs of 0.976 and 0.974 for the training and validation cohorts, respectively, consistent with our finding. However, we further stratified grade 4 astrocytoma into IDH-mut and IDH-wt subtypes and evaluated the prognostic value of the combined model. Previous studies employing radiomics or ML models to distinguish IDH-mut from IDH-wt GBM were based on the 2016 WHO CNS criteria[ 24 – 27 ]. No studies currently have addressed differentiating IDH-mut and IDHwt astrocytoma reclassified as grade 4 astrocytoma under the 2021 WHO CNS criteria. Our multiparametric MRI includes T1C edema, T1C tumor, and T2F tumor. In both tasks, T1C edema, besides the optimal ML model and the combined model, showed the highest AUC for the training (0.843 and 0.838) and validation (0.829 and 0.792) sets. The peritumoral region adjacent to GBM, a mix of infiltrative tumor and vasogenic edema, is often the site of recurrence[ 16 , 27 ]. Thus, tumor and edema regions in grade 4 glioma reflect tumor heterogeneity. Although peritumoral edema is not a significant predictor of the 2021 glioma molecular subtypes, its proportion in GBM is higher than in grade 4 astrocytoma (98.5% vs 90.5%). This suggests peritumoral edema may encompass potential heterogeneity between grade 4 astrocytoma and GBM, consistent with a previous study[ 23 ]. Predictive Value of the Combined Model The combined model, integrating Rad-score and clinical-radiological characteristics, aimed to leverage the strengths of both approaches. However, its performance was similar to the optimal ML model. In Task 1, the AUCs for the training and validation sets were 0.911 vs 0.902 ( P = 0.148) and 0.854 vs 0.855 ( P = 0.934), respectively. In Task 2, the AUCs for the training and validation sets were 0.902 vs 0.904 ( P = 0.907) and 0.909 vs 0.895 ( P = 0.514), respectively. These findings indicate limited added value from incorporating clinical-radiological features. The nomogram based on the combined model of Task 1 offers a practical tool for individualized risk prediction and clinical decision-making, providing an intuitive and clinically interpretable visual representation to aid clinicians in stratifying patients and tailoring treatment strategies. Model evaluation and comparison We evaluated the combined model using various ML algorithms, including LR, SVM, MLP, LDA, RF, and NB. Most algorithms showed consistent performance. This consistency underscores the reliability of the combined model and its potential for broad clinical application. Survival analysis We further evaluated the prognostic value of the combined model. It effectively stratified patients into high-risk and low-risk groups, with prognostic value comparable to molecular subtype. This confirmed that our model can accurately predict glioma molecular subtype and holds substantial prognostic value, offering a new perspective for clinical decision-making. Limitations This study has several limitations. Firstly, the retrospective nature and reliance on data from three institutions may introduce selection bias. Prospective validation on a larger multicenter cohort is necessary to confirm the findings. Secondly, different equipment and scanning parameters used across the three institutions, despite image preprocessing, may have influenced the radiomic features. Future studies should aim for uniform scanning parameters. Thirdly, although this study included multi-sequence MRI of tumor and edema regions, future research should incorporate more functional imaging modalities, such as diffusion-weighted imaging (DWI) and perfusionweighted imaging (PWI), to further enhance the model's predictive power. Conclusion In conclusion, the multiparametric MRI machine learning model effectively differentiated grade 4 astrocytoma from GBM and distinguished between IDH-mut and IDH-wt grade 4 astrocytoma. Additionally, the model stratified various molecular subtypes of glioma patients into high-risk and low-risk groups according to OS, offering a new perspective for clinical decision-making. Abbreviations ACC Accuracy AUC Area under the curve AMP Amplification CDKN Cyclin-dependent kinase inhibitor CE Contrast-enhanced CI Confidence interval CNS Central nervous system DCA Decision curve analysis DWI Diffusion-weighted imaging ECM Edema crosses midline EGFR Epidermal growth factor receptor FLAIR Fluid attenuated inversion recovery GBM Glioblastoma GLCM Gray Level CO-Occurrence Matrix GLDM Gray Level Dependence Matrix GLRLM Gray-Level Run-Length Matrix GLSZM Gray-level size zone matrix ICC Intraclass correlation coefficient IDH-mut Isocitrate dehydrogenase mutant IDH-wt Isocitrate dehydrogenase wild-type KM Kaplan-Meier KPS Karnofsky performance status LASSO Least absolute shrinkage and selection operator LDA Linear discriminant analysis LR Logistic regression mGBM Molecular features of glioblastoma MGMTmet Methyl guanine methyl transferase promoter methylation ML Machine learning MRI Magnetic resonance imaging MLP Multi-layer perceptrons NB Naive Bayes NGTDM Neighborhood Gray Tone Difference Matrix OR Odds ratio OS Overall survival PFS Progression-free survival PWI Perfusion-weighted imaging RF Random forest ROC Receiver operating characteristic ROI Region of interest SEN Sensitivity SPE Specificity SVM Support vector machines TCM Tumor crosses midline TERT Telomerase reverse transcriptase T1WI T1-weighted imaging T2WI T2-weighted imaging VOI Volume of interest WHO World Health Organization XGBoost Extreme Gradient Boosting Declarations Ethics approval and consent to participate: The Ethics Board of the First Hospital of Shanxi Medical University (approval number: KYYJ-2023-058) has approved for this study. Patient consent was waived due to the retrospective nature of the study and the anonymous nature of the data. Consent for publication: Written informed consent for publication was obtained from all participants. Availability of data and materials The datasets generated or analyzed during the study are not publicly available due to institutional regulations but are available from the corresponding author on reasonable request. Competing interests: The authors declare that they have no competing interests. Funding This work was supported by the National Natural Science Foundation of China [grant numbers 82071893, 82371941 to Yan Tan]; the Research Project Supported by Shanxi Scholarship Council of China [grant number 2023-186 to Yan Tan]; and Shanxi Province Higher Education "Billion Project" Science and Technology Guidance Project [grant number BYJL017 to Yan Tan]. Authors' contributions: Wenji Xu : Data processing, Writing - original draft. Yangyang Li : Data processing, Writing - original draft. Jie Zhang : Data curation. Zhiyi Zhang and Pengxin Shen : Data processing, Software. Xiaochun Wang : Writing - review & editing, Project administration. Guoqiang Yang : Data processing, Writing - review & editing, Project administration. Jiangfeng Du : Data processing, Writing - review & editing. Hui Zhang : Supervision, Writing - review & editing. Yan Tan : Methodology, Writing - review & editing, Project administration. Acknowledgements: Not applicable. References Louis DN, Perry A, Wesseling P, Brat DJ, Cree IA, Figarella-Branger D, et al. The 2021 WHO Classification of Tumors of the Central Nervous System: a summary. Neuro Oncol. 2021;23:1231–51. Horbinski C, Berger T, Packer RJ, Wen PY. Clinical implications of the 2021 edition of the WHO classification of central nervous system tumours. Nat Rev Neurol. 2022;18:515–29. Ramos-Fresnedo A, Pullen MW, Perez-Vega C, Domingo RA, Akinduro OO, Almeida JP, et al. The survival outcomes of molecular glioblastoma IDH-wildtype: a multicenter study. J Neurooncol. 2022;157:177–85. Gritsch S, Batchelor TT, Gonzalez Castro LN. Diagnostic, therapeutic, and prognostic implications of the 2021 World Health Organization classification of tumors of the central nervous system. Cancer. 2022;128:47–58. Lambin P, Leijenaar RTH, Deist TM, Peerlings J, De Jong EEC, Van Timmeren J, et al. Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol. 2017;14:749–62. Lambin P, Rios-Velazquez E, Leijenaar R, Carvalho S, Van Stiphout RGPM, Granton P, et al. Radiomics: Extracting more information from medical images using advanced feature analysis. European Journal of Cancer. 2012;48:441–6. Avanzo M, Wei L, Stancanello J, Vallières M, Rao A, Morin O, et al. Machine and deep learning methods for radiomics. Med Phys. 2020;47:e185–202. Moodi F, Khodadadi Shoushtari F, Ghadimi DJ, Valizadeh G, Khormali E, Salari HM, et al. Glioma Tumor Grading Using Radiomics on Conventional MRI : A Comparative Study of WHO 2021 and WHO 2016 Classification of Central Nervous Tumors. Magnetic Resonance Imaging. 2023;jmri.29146. Tian Q, Yan L, Zhang X, Zhang X, Hu Y, Han Y, et al. Radiomics strategy for glioma grading using texture features from multiparametric MRI. Magnetic Resonance Imaging. 2018;48:1518–28. Chiu F-Y, Le NQK, Chen C-Y. A Multiparametric MRI-Based Radiomics Analysis to Efficiently Classify Tumor Subregions of Glioblastoma: A Pilot Study in Machine Learning. JCM. 2021;10:2030. Lin K, Cidan W, Qi Y, Wang X. Glioma grading prediction using multiparametric magnetic resonance imaging‐based radiomics combined with proton magnetic resonance spectroscopy and diffusion tensor imaging. Medical Physics. 2022;49:4419–29. Ding J, Zhao R, Qiu Q, Chen J, Duan J, Cao X, et al. Developing and validating a deep learning and radiomic model for glioma grading using multiplanar reconstructed magnetic resonance contrast-enhanced T1-weighted imaging: a robust, multi-institutional study. Quant Imaging Med Surg. 2022;12:1517–28. Vijithananda SM, Jayatilake ML, Gonçalves TC, Rato LM, Weerakoon BS, Kalupahana TD, et al. Texture feature analysis of MRI-ADC images to differentiate glioma grades using machine learning techniques. Sci Rep. 2023;13:15772. Xing X, Zhu M, Chen Z, Yuan Y. Comprehensive learning and adaptive teaching: Distilling multi-modal knowledge for pathological glioma grading. Medical Image Analysis. 2024;91:102990. Malik N, Geraghty B, Dasgupta A, Maralani PJ, Sandhu M, Detsky J, et al. MRI radiomics to differentiate between low grade glioma and glioblastoma peritumoral region. J Neurooncol. 2021;155:181–91. Szekeres D, Jetty SN, Soni N. The Role of Multiparametric MRI in Diagnosing and Grading Glioma. Neurology India. 2023;71:1274–5. Naser MA, Deen MJ. Brain tumor segmentation and grading of lower-grade glioma using deep learning in MRI images. Computers in Biology and Medicine. 2020;121:103758. Lee D, Riestenberg RA, Haskell-Mendoza A, Bloch O. Diffuse astrocytic glioma, IDH-Wildtype, with molecular features of glioblastoma, WHO grade IV: A single-institution case series and review. J Neurooncol. 2021;152:89–98. Weller M, Van Den Bent M, Preusser M, Le Rhun E, Tonn JC, Minniti G, et al. EANO guidelines on the diagnosis and treatment of diffuse gliomas of adulthood. Nat Rev Clin Oncol. 2021;18:170–86. Park YW, Kim S, Park CJ, Ahn SS, Han K, Kang S-G, et al. Adding radiomics to the 2021 WHO updates may improve prognostic prediction for current IDH-wildtype histological lower-grade gliomas with known EGFR amplification and TERT promoter mutation status. Eur Radiol. 2022;32:8089–98. Agarwal A, Edgar MA, Desai A, Gupta V, Soni N, Bathla G. Molecular GBM versus Histopathological GBM: Radiology-Pathology-Genetic Correlation and the New WHO 2021 Definition of Glioblastoma. AJNR Am J Neuroradiol. 2024;ajnr.A8225. Zeng C, Song X, Zhang Z, Cai Q, Cai J, Horbinski C, et al. Dissection of transcriptomic and epigenetic heterogeneity of grade 4 gliomas: implications for prognosis. acta neuropathol commun. 2023;11:133. Wei R, Lu S, Lai S, Liang F, Zhang W, Jiang X, et al. A subregion-based RadioFusionOmics model discriminates between grade 4 astrocytoma and glioblastoma on multisequence MRI. J Cancer Res Clin Oncol. 2024;150:73. Pasquini L, Napolitano A, Tagliente E, Dellepiane F, Lucignani M, Vidiri A, et al. Deep Learning Can Differentiate IDH-Mutant from IDH-Wild GBM. J Pers Med. 2021;11. Calabrese E, Villanueva-Meyer JE, Cha S. A fully automated artificial intelligence method for non-invasive, imaging-based identification of genetic alterations in glioblastomas. Sci Rep. 2020;10:11852. Kandalgaonkar P, Sahu A, Saju AC, Joshi A, Mahajan A, Thakur M, et al. Predicting IDH subtype of grade 4 astrocytoma and glioblastoma from tumor radiomic patterns extracted from multiparametric magnetic resonance images using a machine learning approach. Front Oncol. 2022;12:879376. Cheng J, Liu J, Yue H, Bai H, Pan Y, Wang J. Prediction of Glioma Grade Using Intratumoral and Peritumoral Radiomic Features From Multiparametric MRI Images. IEEE/ACM Trans Comput Biol Bioinform. 2022;19:1084–95. Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterial.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5288001","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":372799473,"identity":"ff474e32-7a41-4b75-bc4d-5be03ac2f328","order_by":0,"name":"Wenji Xu","email":"","orcid":"","institution":"Shanxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Wenji","middleName":"","lastName":"Xu","suffix":""},{"id":372799474,"identity":"e68733f1-1fd0-4c3d-b12e-54f7c3513558","order_by":1,"name":"Yangyang Li","email":"","orcid":"","institution":"Shanxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yangyang","middleName":"","lastName":"Li","suffix":""},{"id":372799475,"identity":"e6bac687-53f5-414f-979e-d5dca5b4308a","order_by":2,"name":"Jie Zhang","email":"","orcid":"","institution":"Shanxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jie","middleName":"","lastName":"Zhang","suffix":""},{"id":372799476,"identity":"b3b3bcda-ef4b-4ba1-bf6f-015ecf2c3d20","order_by":3,"name":"Zhiyi Zhang","email":"","orcid":"","institution":"Shanxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Zhiyi","middleName":"","lastName":"Zhang","suffix":""},{"id":372799477,"identity":"50a5e96e-f9c3-4c6d-bf63-d6e4fff3fb88","order_by":4,"name":"Pengxin Shen","email":"","orcid":"","institution":"Shanxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Pengxin","middleName":"","lastName":"Shen","suffix":""},{"id":372799478,"identity":"70b897e1-a501-40f8-b761-081f9f425ee0","order_by":5,"name":"Xiaochun Wang","email":"","orcid":"","institution":"First Hospital of Shanxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xiaochun","middleName":"","lastName":"Wang","suffix":""},{"id":372799479,"identity":"c0f2d3bf-e74e-44aa-a780-1b35425ac948","order_by":6,"name":"Guoqiang Yang","email":"","orcid":"","institution":"First Hospital of Shanxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Guoqiang","middleName":"","lastName":"Yang","suffix":""},{"id":372799480,"identity":"49dd9a76-71b8-4d11-a8ef-206bb8f2d5d1","order_by":7,"name":"Jiangfeng Du","email":"","orcid":"","institution":"First Hospital of Shanxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jiangfeng","middleName":"","lastName":"Du","suffix":""},{"id":372799481,"identity":"0696771b-3067-4eb0-ba43-5c58659faaec","order_by":8,"name":"hui zhang","email":"","orcid":"","institution":"First Hospital of Shanxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"hui","middleName":"","lastName":"zhang","suffix":""},{"id":372799482,"identity":"957f5a08-3a1f-4f14-837e-83c27e758adb","order_by":9,"name":"Yan Tan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA8UlEQVRIiWNgGAWjYPACCTl5hvMPH3xgkCCslgdCWRgbNp5hNpxBgpaKxIbDZ9iEeYhxkT372WMSP3dIMDa2nT3GbPPHIo+/gfnhoxv4bOHJS5PsPSPBzM5zLu1xbptEscQBNmPjHLwOyzGT4G2TYGOcccDcOLdBIrHhAA+bNF4t/G/MJP+2SfAw3H9gJm3xRyJxPkEtEjlm0kBbJBgOnDGTZmCTSNxAUMuNN8bWsm0SBoYNx5INe9skEjceJuAX9v4cw5tv2+rq5zMcPvjgx5+6xHnHmx8+xqcFCFjQoo8Zv3Kwkg+E1YyCUTAKRsGIBgDXMkrQ8dv+pQAAAABJRU5ErkJggg==","orcid":"","institution":"First Hospital of Shanxi Medical University","correspondingAuthor":true,"prefix":"","firstName":"Yan","middleName":"","lastName":"Tan","suffix":""}],"badges":[],"createdAt":"2024-10-18 08:53:44","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5288001/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5288001/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":68698848,"identity":"3057fcfe-68b5-4b0c-bd93-6a1478681630","added_by":"auto","created_at":"2024-11-11 07:09:18","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":546992,"visible":true,"origin":"","legend":"\u003cp\u003ePatient flow and machine learning-based classification chart. WHO: World Health Organization; IDH-mut: Isocitrate dehydrogenase-mutant; IDHwt: Isocitrate dehydrogenase wild-type; TERTp \u003csup\u003eMUT\u003c/sup\u003e: telomerase reverse transcriptase promoter mutation; EGFR \u003csup\u003eAMP\u003c/sup\u003e: epidermal growth factor receptor amplification; GBM: glioblastoma; mGBM: molecular glioblastoma.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-5288001/v1/c2d7a8240c65ce1899429c03.png"},{"id":68698834,"identity":"a3743ac8-9606-426f-a3b8-22e0f9236782","added_by":"auto","created_at":"2024-11-11 07:09:18","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":658642,"visible":true,"origin":"","legend":"\u003cp\u003eThe workflow of radiomics analysis. The radiomics workflow includes ROI segmentation, feature extraction, feature selection and ML model construction and model evaluation. GLCM: Gray Level Co-occurrence Matrix; GLRLM: Gray Level Run Length Matrix; GLSZM: Gray Level Size Zone Matrix; GLDM: Gray Level Dependence Matrix; NGTDM: Neighborhood Gray Tone Difference Matrix; ICC: intraclass correlation coefficient; M-W: Mann-Whitney test; LASSO: Least Absolute Shrinkage and Selection Operator; XGBoost: Extreme Gradient Boosting; ML: machine learning.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-5288001/v1/b6e4bd7af2eb34aec8961fa6.png"},{"id":68697767,"identity":"8f8f2571-adc1-4ae1-8618-6808fc434ba5","added_by":"auto","created_at":"2024-11-11 07:01:18","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":667494,"visible":true,"origin":"","legend":"\u003cp\u003eROC curves of the ML models, clinical models, and combined models in the training (a, e) and validation sets (b, f) for two tasks. DeLong’s test p-valued heat maps of the training (c, g) and validation sets (d, h) for two tasks. ROC: Receiver Operating Characteristic; ML: machine learning; T1C: contrast-enhanced T1-weighted imaging; T2F: T2-weighted imaging fluid attenuated inversion recovery.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-5288001/v1/21f5976a490ca17e64e4f1ab.png"},{"id":68699454,"identity":"99f0537d-13dd-4eb1-8023-72c18a3de541","added_by":"auto","created_at":"2024-11-11 07:17:18","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1111424,"visible":true,"origin":"","legend":"\u003cp\u003ea. Nomogram of the combined model for Task 1 (grade 4 vs GBM). The nomogram incorporates Rad-score and independent clinical-radiological risk factors. b-e. Representative MR images and case nomograms for grade 4 astrocytoma and GBM patients. b. CE-T1WI and T2WI-FLAIR images of a 50-year-old male patient with grade 4 astrocytoma. c. Rad-score = 1.36, calculates a score of 172 total points based on the nomogram, with a prediction probability of 0.141; d. CE-T1WI and T2WIFLAIR images of a 53-year-old male patient with GBM. e. Rad-score=8.53, calculates a score of 268 total points based on the nomogram, with a prediction probability of 0.937. MGMTmet: methyl guanine methyl transferase promoter methylation\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-5288001/v1/d3927d14ba71c566d7428297.png"},{"id":68698835,"identity":"67a609c4-c3d7-4197-bc8e-c219120d9dcb","added_by":"auto","created_at":"2024-11-11 07:09:18","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":461291,"visible":true,"origin":"","legend":"\u003cp\u003eCalibration curves of the combined model in the training (a, c) and validation sets (b, f) for two tasks. Decision curve analysis (DCA) of the training (c, g) and validation sets (d, h) of the clinical model, optimal ML model, and combined model for two tasks. ML: machine learning.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-5288001/v1/726e0a8e0c86c083e649a3bf.png"},{"id":68697769,"identity":"b127e5e9-33ba-42d7-aa60-777dd8c1c662","added_by":"auto","created_at":"2024-11-11 07:01:18","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":994693,"visible":true,"origin":"","legend":"\u003cp\u003eAUCs for the training (a, c, e, g) and validation sets (b, d, f, h) of the combined model constructed by 7 classifiers. DeLong’s test p-valued heat maps of training (i, k) and validation sets (j, l) for two tasks. ROC: Receiver Operating Characteristic; RF: Random Forest; SVM: support vector machines, NB: Naive Bayes; XGboost: extreme gradient boosting; LDA: linear discriminant analysis; MLP: multi-layer perceptrons; LR: logistic regression.\u003c/p\u003e","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-5288001/v1/ca8386f9596e86aced778116.png"},{"id":68697771,"identity":"ffffe9c1-e99b-48fa-b4ca-3edc9a75e000","added_by":"auto","created_at":"2024-11-11 07:01:18","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":223267,"visible":true,"origin":"","legend":"\u003cp\u003eKaplan-Meier survival analysis of the combined model for two tasks. The combined model (dotted line) effectively stratified Task 1 and Task 2 cases into highrisk (red line) and low-risk (green line) groups, with significant prognostic differences. The combined model performed comparable to the molecular subtype (solid line) in two tasks.\u003c/p\u003e","description":"","filename":"Figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-5288001/v1/680fbf2eed94da92bee558b0.png"},{"id":71434523,"identity":"95fcea8e-c7d5-4f0f-89a6-6d3a37ea2b94","added_by":"auto","created_at":"2024-12-15 08:31:44","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":6272905,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5288001/v1/7c373cc8-b20f-4545-9d90-f99c0815669c.pdf"},{"id":68697773,"identity":"7549870f-cdfa-46c3-821e-f280c01b4f97","added_by":"auto","created_at":"2024-11-11 07:01:18","extension":"docx","order_by":11,"title":"","display":"","copyAsset":false,"role":"supplement","size":1115221,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-5288001/v1/e7d85973bd3c41a27341cdde.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Predicting the Molecular Subtypes of 2021 WHO Grade 4 Glioma by a Multiparametric MRI-Based Machine Learning Model","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe 2021 World Health Organization (WHO) Classification of central nervous system (CNS) tumors emphasizes the integration of molecular parameters with histological findings for accurate grading[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Notably, isocitrate dehydrogenase-mutant (IDH-mut) grade 2\u0026ndash;3 astrocytoma is now reclassified as astrocytoma, IDH-mutant, WHO grade 4, in the presence of homozygous deletion of cyclin-dependent kinase inhibitor A/B (CDKN2A/B), the regardless of necrosis and/or microvascular proliferation. Similarly, IDH wild-type (IDH-wt) grade 2\u0026ndash;3 astrocytoma is reclassified as astrocytoma, IDH wild-type, with molecular features of glioblastoma (mGBM), WHO grade 4, if any or a combination of telomerase reverse transcriptase (TERT) promoter mutation, epidermal growth factor receptor (EGFR) amplification, and chromosome\u0026thinsp;+\u0026thinsp;7 /\u0026minus;10 copy number changes. These molecularly characterized grade 2\u0026ndash;3 astrocytoma are high-risk molecular subtypes with poor prognosis[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eDespite poor prognosis, IDH-wt grade 4 astrocytoma patients demonstrated longer overall survival (OS) and progression-free survival (PFS) compared to glioblastoma (GBM) patients[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. The two IDH variants of grade 4 astrocytoma exhibit different biological behaviors and clinical outcomes, with IDH-mut astrocytoma showing less aggressive behavior and better prognosis under similar treatment protocols[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Therefore, accurately distinguishing the different molecular subtypes of WHO grade 4 glioma is highly relevant for prognostic stratification and personalized treatment. However, current molecular diagnostics depend on invasive biopsies, necessitating a reliable non-invasive method for predicting the molecular subtypes of glioma.\u003c/p\u003e \u003cp\u003eMagnetic Resonance Imaging (MRI) is extensively utilized for glioma diagnosis and monitoring but falls short in revealing histological and molecular details. Radiomics, by extracting high-throughput image features, offers insights into tumor heterogeneity, thereby enhancing diagnostic and therapeutic accuracy[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Machine learning (ML) algorithms can process radiomics data to predict tissue characteristics and identify molecular features of glioma[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Previous studies have developed MRI-based radiomics and ML models for glioma grading, mostly adhering to the 2016 or earlier WHO classifications, or focusing on glioma grading[\u003cspan additionalcitationids=\"CR10 CR11 CR12 CR13 CR14 CR15 CR16\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. The value of differentiating grade 4 glioma from GBM, and further distinguishing grade 4 IDH-mut from IDH-wt astrocytoma, remains under investigated. Therefore, our study aims to construct a ML model using multiparametric MRI to differentiate grade 4 astrocytoma from GBM (Task 1), and further stratify grade 4 astrocytoma into IDH-mut and IDH-wt subtypes (Task 2). Additionally, we seek to analyze its corresponding prognostic value in OS.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003e\u003cstrong\u003ePatient population\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study conducted in accordance with the Helsinki Declaration ethical standards. The Ethics Board of the First Hospital of Shanxi Medical University (approval number: KYYJ-2023-058) has approved for this retrospective study. Given the retrospective nature of the study and the anonymous nature of the data, the requirement for informed consent from each patient has been waived.\u0026nbsp;This study collected data from 320 patients with pathologically confirmed glioma from three institutions (First Hospital of Shanxi Medical University, Shanxi Provincial People\u0026apos;s Hospital and Shanxi Bethune Hospital) based on inclusion and exclusion criteria between February 2011 and December 2023. Patients were divided into training and validation sets in a 7:3 ratio. OS was defined as the time from initial diagnosis to death or last follow-up, with survival data available for 312 patients.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eInclusion criteria: 1) Histological diagnosis of diffuse glioma; 2) No prior radiotherapy, chemotherapy, or surgery before MRI examination; 3) No history of craniocerebral surgery or other systemic malignancies; 4) Availability of molecular information, including IDH, 1p/19q, methyl guanine methyl transferase promoter methylation (MGMTmet), CDKN2A/B, EGFR, TERT, and chromosome 7/10 status. Exclusion criteria: 1) 2021 WHO CNS grade 2-3 astrocytoma; 2) Oligodendroglioma; 3) Incomplete or poor-quality MRI images; 4) Incomplete molecular information. A flowchart of patient selection and machine learning classification is shown in \u003cstrong\u003eFigure 1\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMachine learning-based classification\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe constructed two independent binary classification tasks: 1) distinguishing grade 4 astrocytoma from GBM (Task 1: grade 4 vs GBM), and 2) stratifying grade 4 astrocytoma into IDH-mut and IDH-wt subtypes (Task 2: IDH-mut grade 4 vs IDH-wt grade 4).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical-radiological characteristics collection\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eClinical-radiological characteristics included gender, age, MGMTmet, treatment, Karnofsky Performance Status (KPS) score, tumor number, tumor margin, intratumoral hemorrhage, intratumoral necrosis, peritumoral edema, maximum diameter, midline shift, enhancement pattern, enhancement quality, tumor crosses midline (TCM), edema crosses midline (ECM), cortical involvement, deep white matter invasion, pial invasion, and ependymal invasion. Molecular biomarker detection methods are described in supplementary materials.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMRI image acquisition and preprocessing\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMR Images were obtained using 3T scanners (Signa HDxt from GE Healthcare, USA and Skyra from Siemens Healthineers, Germany) with axial T2-weighted imaging fluid attenuated inversion recovery (T2WI-FLAIR) and contrast-enhanced T1WI (CE-T1WI). Scanning parameters are described in supplementary materials.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eImages were preprocessed to standardize signal intensity on a scale of 100, corrected for N4ITK bias fields, and resampled to a voxel size of 1 mm \u0026times; 1 mm \u0026times; 1 mm, with voxel intensity discretized using a fixed bin width of 25. Image preprocessing used FeAture Explorer V.0.5.7 (FAE, https:// github.com/salan668/FAE) and 3Dslicer version 5.7.20240325 (https://www.slicer.org).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImage segmentation, feature extraction and selection\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eT2WI-FLAIR images were registered to CE-T1WI using a rigid registration algorithm. A radiologist with 8 years of experience manually delineated tumor and edema areas on CE-T1WI and T2-FLAIR, respectively. The volume of interest (VOIs) included CE-T1WI tumor (T1C tumor), CE-T1WI edema (T1C edema), and T2WI-FLAIR tumor (T2F tumor). Thirty patients underwent repeat segmentation to calculate the intraclass correlation coefficient (ICC) for inter-observer agreement.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFeature selection was performed on the training set. Features with an ICC \u0026ge; 0.75 were selected, followed by z-score normalization. Normality was assessed using the Shapiro-Wilk test. Features were retained if \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05 using independent sample t-tests or Mann-Whitney U tests, depending on distribution. The least absolute shrinkage and selection operator (LASSO) selected the best regularization parameters through ten-fold cross-validation.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eImage segmentation utilized ITK-SNAP software (http://www.itksnap.org/, version 4.0.0). Features were extracted using FAE. The radiomics workflow is shown in \u003cstrong\u003eFigure 2\u003c/strong\u003e. Region of interest (ROI) segmentation is shown in \u003cstrong\u003eSupplementary Figure 1\u003c/strong\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConstruction of the ML model\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe employed Extreme Gradient Boosting (XGBoost) to build ML models for two tasks. Based on the selected features of each sequence and their combination, ML models for single and combined sequences were constructed. The ML model with the best predictive performance on the validation set was selected as the optimal ML model, and the Rad-score was calculated.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConstruction of the clinical model\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eUnivariate logistic regression (LR) analysis screened clinical-radiological characteristics, and variables with \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05 were included in multivariate LR analysis. Significant variables were used to establish the clinical model using XGBoost. \u003cstrong\u003eConstruction of the combined model and nomogram\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA combined model was constructed using XGBoost based on the Rad-score and independent clinical-radiological risk factors. A nomogram was generated using LR to visually discriminate between grade 4 astrocytoma and GBM.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eModel evaluation and model comparison\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe model\u0026rsquo;s performance in predicting WHO grade 4 glioma molecular subtypes was evaluated using the Receiver Operating Characteristic (ROC) curve, assessing Area Under the Curve (AUC), sensitivity (SEN), specificity (SPE) and accuracy (ACC). The DeLong\u0026apos;s test compared predictive performance between models, with statistical significance set at \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05. Decision and calibration curves evaluated model calibration and clinical utility.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eVarious ML algorithms, including LR, support vector machines (SVM), multilayer perceptrons (MLP), linear discriminant analysis (LDA), random forest (RF), and Naive Bayes (NB), were used to evaluate the generalization and stability of the combined model.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSurvival analysis\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eKaplan-Meier (KM) survival analysis and log-rank test evaluate the prognostic value of the molecular subtype prediction model based on the combined model. The Z-test compared the prognostic value between the combined model and the molecular subtype.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical analysis\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eStatistical analyses were performed using R version 4.2.3 (http://www.Rproject.org). Numerical variables were expressed as mean \u0026plusmn; standard deviations. The Shapiro-Wilk test assessed normality. Independent sample t-test or Mann-Whitney U-test was used based on the distribution. Categorical variables were expressed as frequencies (percentages) and evaluated using Pearson\u0026rsquo;s chi-square test or Fisher\u0026rsquo;s exact test. \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05 was considered statistically significant.\u0026nbsp;\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eClinical-radiological baseline characteristics\u003c/h2\u003e \u003cp\u003eAmong 320 patients enrolled based on the 2021 WHO CNS tumor classification, there were 196 GBMs, 41 IDH-mut grade 4, and 83 IDH-wt grade 4 astrocytomas. For Task 1 (grade 4 vs GBM), 224 and 96 cases were in the training and validation set, respectively. For Task 2 (IDH-mut grade 4 vs IDH-wt grade 4), 118 and 51 cases were in the training and validation set. Of the 312 patients with survival information, there were 189 GBMs, 41 IDH-mut grade 4, and 82 IDH-wt grade 4 astrocytomas. Except for MGMTmet (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.027) in Task 1 and peritumoral edema (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.033) in Task 2, there were no significant differences in clinical-radiological characteristics between the training and validation sets for both tasks. Baseline characteristics are detailed in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\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-radiological baseline characteristics between training and validation sets of the two tasks\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=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" 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=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" 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\" colspan=\"3\" nameend=\"c5\" namest=\"c3\"\u003e \u003cp\u003eTask 1 (grade 4 vs GBM)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c9\" namest=\"c7\"\u003e \u003cp\u003eTask 2 (IDH-mut \u003csup\u003egrade 4\u003c/sup\u003e vs IDH-wt \u003csup\u003egrade 4\u003c/sup\u003e)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTraining set\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;224)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eValidation set\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;96)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e-value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eTraining set\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;87)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eValidation set\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;37)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e-value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eGender\u003c/b\u003e\u003c/p\u003e \u003cp\u003eMale\u003c/p\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e130 (58.04%)\u003c/p\u003e \u003cp\u003e94 (41.96%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e53 (55.21%)\u003c/p\u003e \u003cp\u003e43 (44.79%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.712\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e48 (55.17%)\u003c/p\u003e \u003cp\u003e39 (44.83%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e23 (62.16%)\u003c/p\u003e \u003cp\u003e14 (37.84%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.554\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAge(years)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e53.76\u0026thinsp;\u0026plusmn;\u0026thinsp;13.48\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c4\"\u003e \u003cp\u003e52.23\u0026thinsp;\u0026plusmn;\u0026thinsp;13.62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.201\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e49.26\u0026thinsp;\u0026plusmn;\u0026thinsp;14.71\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e52.11\u0026thinsp;\u0026plusmn;\u0026thinsp;14.21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.216\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMGMTmet\u003c/b\u003e\u003c/p\u003e \u003cp\u003eYes\u003c/p\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\u003e136 (60.71%)\u003c/p\u003e \u003cp\u003e88 (39.29%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e45 (46.88%)\u003c/p\u003e \u003cp\u003e51 (53.12%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.027\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e64 (73.56%)\u003c/p\u003e \u003cp\u003e23 (26.44%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e21 (56.76%)\u003c/p\u003e \u003cp\u003e16 (43.24%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.090\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTreatment\u003c/b\u003e\u003c/p\u003e \u003cp\u003eSurgery\u003c/p\u003e \u003cp\u003eCombination therapy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e88 (39.29%)\u003c/p\u003e \u003cp\u003e136 (60.71%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e38 (39.58%)\u003c/p\u003e \u003cp\u003e58 (60.42%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e38 (43.68%)\u003c/p\u003e \u003cp\u003e49 (56.32%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e19 (51.35%)\u003c/p\u003e \u003cp\u003e18 (48.65%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.440\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eKPS\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e78.64\u0026thinsp;\u0026plusmn;\u0026thinsp;12.08\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c4\"\u003e \u003cp\u003e79.23\u0026thinsp;\u0026plusmn;\u0026thinsp;11.46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.913\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e78.46\u0026thinsp;\u0026plusmn;\u0026thinsp;12.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e79.78\u0026thinsp;\u0026plusmn;\u0026thinsp;6.68\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.884\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTumor number\u003c/b\u003e\u003c/p\u003e \u003cp\u003eSingle\u003c/p\u003e \u003cp\u003eMultiple\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e118 (52.68%)\u003c/p\u003e \u003cp\u003e106 (47.32%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e54 (56.25%)\u003c/p\u003e \u003cp\u003e42 (43.75%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.625\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e59 (67.82%)\u003c/p\u003e \u003cp\u003e28 (32.18%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e21 (56.76%)\u003c/p\u003e \u003cp\u003e16 (43.24%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.305\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTumor margin\u003c/b\u003e\u003c/p\u003e \u003cp\u003eClear\u003c/p\u003e \u003cp\u003eNon-clear\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e97 (43.30%)\u003c/p\u003e \u003cp\u003e127 (56. 70%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e44 (45.83%)\u003c/p\u003e \u003cp\u003e52 (54.17%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.713\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e31 (35.63%)\u003c/p\u003e \u003cp\u003e56 (64.37%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e20 (54.05%)\u003c/p\u003e \u003cp\u003e17 (45.95%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.073\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eIntratumoral hemorrhage\u003c/b\u003e\u003c/p\u003e \u003cp\u003eYes\u003c/p\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\u003e54 (24.11%)\u003c/p\u003e \u003cp\u003e170 (75.89%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e23 (23.96%)\u003c/p\u003e \u003cp\u003e73 (76.04%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e15 (17.24%)\u003c/p\u003e \u003cp\u003e72 (82.76%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e8 (21.62%)\u003c/p\u003e \u003cp\u003e29 (78.38%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.617\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eIntratumoral necrosis\u003c/b\u003e\u003c/p\u003e \u003cp\u003eYes\u003c/p\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\u003e180 (80.36%)\u003c/p\u003e \u003cp\u003e44 (19.64%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e80 (83.33%)\u003c/p\u003e \u003cp\u003e16 (16.67%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.640\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e61 (70.11%)\u003c/p\u003e \u003cp\u003e26 (29.89%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e30 (81.08%)\u003c/p\u003e \u003cp\u003e7 (18.92%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.269\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePeritumoral edema\u003c/b\u003e\u003c/p\u003e \u003cp\u003eYes\u003c/p\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\u003e212 (94.64%)\u003c/p\u003e \u003cp\u003e12 (5.36%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e94 (97.92%)\u003c/p\u003e \u003cp\u003e2 (2.08%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.243\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e76 (87.36%)\u003c/p\u003e \u003cp\u003e11 (12.64%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e37 (100.00%)\u003c/p\u003e \u003cp\u003e0 (0.00%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.033\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMaximum diameter\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e4.73\u0026thinsp;\u0026plusmn;\u0026thinsp;1.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c4\"\u003e \u003cp\u003e4.74\u0026thinsp;\u0026plusmn;\u0026thinsp;1.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.840\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e4.66\u0026thinsp;\u0026plusmn;\u0026thinsp;1.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c8\"\u003e \u003cp\u003e4.64\u0026thinsp;\u0026plusmn;\u0026thinsp;1.52\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.960\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMidline shift\u003c/b\u003e\u003c/p\u003e \u003cp\u003eYes\u003c/p\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\u003e111 (47.45%)\u003c/p\u003e \u003cp\u003e113 (52.55%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e48 (50.00%)\u003c/p\u003e \u003cp\u003e48 (50.00%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.456\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e44 (50.57%)\u003c/p\u003e \u003cp\u003e43 (49.43%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e22 (59.46%)\u003c/p\u003e \u003cp\u003e15 (40.54%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.433\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eEnhancement pattern\u003c/b\u003e\u003c/p\u003e \u003cp\u003eNo reinforcement\u003c/p\u003e \u003cp\u003eAnnular reinforcement\u003c/p\u003e \u003cp\u003eNodular enhancement\u003c/p\u003e \u003cp\u003eMixed reinforcement\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e32 (14.29%)\u003c/p\u003e \u003cp\u003e103 (45.98%)\u003c/p\u003e \u003cp\u003e30 (13.39%)\u003c/p\u003e \u003cp\u003e59 (26.34%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10 (10.42%)\u003c/p\u003e \u003cp\u003e41 (42.71%)\u003c/p\u003e \u003cp\u003e10 (10.42%)\u003c/p\u003e \u003cp\u003e35 (36.46%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.316\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e24 (27.59%)\u003c/p\u003e \u003cp\u003e36 (41.38%)\u003c/p\u003e \u003cp\u003e13(14.94%)\u003c/p\u003e \u003cp\u003e14 (16.09%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e6 (16.22%)\u003c/p\u003e \u003cp\u003e15 (40.54%)\u003c/p\u003e \u003cp\u003e6 (16.22%)\u003c/p\u003e \u003cp\u003e10 (27.03%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.392\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eEnhancement quality\u003c/b\u003e\u003c/p\u003e \u003cp\u003eNo reinforcement\u003c/p\u003e \u003cp\u003eReinforcement\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e31 (13.84%)\u003c/p\u003e \u003cp\u003e193 (86.16%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e9 (9.38%)\u003c/p\u003e \u003cp\u003e87 (90.62%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.357\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e24 (27.59%)\u003c/p\u003e \u003cp\u003e63 (72.41%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e5 (13.51%)\u003c/p\u003e \u003cp\u003e32 (86.49%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.108\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTCM\u003c/b\u003e\u003c/p\u003e \u003cp\u003eYes\u003c/p\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\u003e183 (81.70%)\u003c/p\u003e \u003cp\u003e41 (18.30%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e75 (78.12%)\u003c/p\u003e \u003cp\u003e21 (21.88%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.537\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e67 (77.01%)\u003c/p\u003e \u003cp\u003e20 (22.99%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e29 (78.38%)\u003c/p\u003e \u003cp\u003e8 (21.62%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eECM\u003c/b\u003e\u003c/p\u003e \u003cp\u003eYes\u003c/p\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\u003e177 (79.02%)\u003c/p\u003e \u003cp\u003e47 (20.98%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e66 (68.75%)\u003c/p\u003e \u003cp\u003e30 (31.25%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.063\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e60 (68.97%)\u003c/p\u003e \u003cp\u003e27 (31.03%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e29 (78.38%)\u003c/p\u003e \u003cp\u003e8 (21.62%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.384\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCortical involvement\u003c/b\u003e\u003c/p\u003e \u003cp\u003eYes\u003c/p\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\u003e66 (29.46%)\u003c/p\u003e \u003cp\u003e158 (70.54%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e28 (29.17%)\u003c/p\u003e \u003cp\u003e68 (70.83%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e25 (28.74%)\u003c/p\u003e \u003cp\u003e62 (71.26%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e11 (29.73%)\u003c/p\u003e \u003cp\u003e26 (70.27%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDeep white matter invasion\u003c/b\u003e\u003c/p\u003e \u003cp\u003eYes\u003c/p\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\u003e62 (27.68%)\u003c/p\u003e \u003cp\u003e162 (72.32%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e28 (29.17%)\u003c/p\u003e \u003cp\u003e68 (70.83%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.788\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e23 (26.44%)\u003c/p\u003e \u003cp\u003e64 (73.56%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e10 (27.03%)\u003c/p\u003e \u003cp\u003e27 (72.97%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePial invasion\u003c/b\u003e\u003c/p\u003e \u003cp\u003eYes\u003c/p\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\u003e132 (58.93%)\u003c/p\u003e \u003cp\u003e92(41.07%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e59 (61.46%)\u003c/p\u003e \u003cp\u003e37 (38.54%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.710\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e41 (47.13%)\u003c/p\u003e \u003cp\u003e46 (52.87%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e23 (62.16%)\u003c/p\u003e \u003cp\u003e14 (37.84%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.169\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eEpendymal invasion\u003c/b\u003e\u003c/p\u003e \u003cp\u003eYes\u003c/p\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\u003e168 (75.00%)\u003c/p\u003e \u003cp\u003e56 (25.00%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e72 (75.00%)\u003c/p\u003e \u003cp\u003e24 (25.00%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e60 (68.97%)\u003c/p\u003e \u003cp\u003e27 (31.03%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e30 (81.08%)\u003c/p\u003e \u003cp\u003e7 (18.92%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.193\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\u003eGBM: glioblastoma; IDH-mut: Isocitrate dehydrogenase-mutant; IDH-wt: Isocitrate dehydrogenase wild-type; MGMTmet: methyl guanine methyl transferase promoter methylation; KPS: karnofsky kerformance status score; TCM: tumor across midline, ECM: edema across midline.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eConstruction of the clinical model\u003c/h2\u003e \u003cp\u003eUnivariate and multivariate LR results are presented in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. For Task 1, age (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.014) and MGMTmet (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.028) were significant. For Task 2, ECM (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.005) and deep white matter invasion (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.003) were significant.\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 and multivariate logistic regression analysis in the training sets of the two tasks\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"11\"\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 \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=\"4\" nameend=\"c6\" namest=\"c3\"\u003e \u003cp\u003eTask 1 (grade 4 vs GBM)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c11\" namest=\"c8\"\u003e \u003cp\u003eTask 2 (IDH-mut \u003csup\u003egrade 4\u003c/sup\u003e vs IDH-wt \u003csup\u003egrade 4\u003c/sup\u003e)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003eUnivariate analysis\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003eMultivariate analysis\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003eUnivariate analysis\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c11\" namest=\"c10\"\u003e \u003cp\u003eMultivariate 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 (95%CI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eOR (95%CI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003eOR (95%CI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003eOR (95%CI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eGender\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.103(0.638\u0026ndash;1.918)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.727\u003c/p\u003e \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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1.120 (0.458\u0026ndash;2.729)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.802\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAge(years)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.034 (1.013\u0026ndash;1.056)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.002\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.028 (1.006\u0026ndash;1.052)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.014\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.979 (0.948\u0026ndash;1.009)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.170\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMGMTmet\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.564 (0.316\u0026ndash;0.991)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.049\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.503 (0.269\u0026ndash;0.918)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.028\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1.700 (0.612\u0026ndash;5.247)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.326\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTreatment\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.487 (0.856\u0026ndash;2.585)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.159\u003c/p\u003e \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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.831 (0.340\u0026ndash;2.035)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.684\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eKPS\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.987 (0.963\u0026ndash;1.010)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.271\u003c/p\u003e \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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1.003 (0.967\u0026ndash;1.044)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.863\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTumor number\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.557 (0.903\u0026ndash;2.706)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.113\u003c/p\u003e \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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.673 (0.244\u0026ndash;1.745)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.426\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTumor margin\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.971 (0.561\u0026ndash;1.676)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.917\u003c/p\u003e \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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1.167 (0.465\u0026ndash;3.032)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.745\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eIntratumoral hemorrhage\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.776 (0.925\u0026ndash;3.555)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.093\u003c/p\u003e \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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.940 (0.268\u0026ndash;2.959)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.918\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eIntratumoral necrosis\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.696 (0.867\u0026ndash;3.308)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.120\u003c/p\u003e \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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1.643 (0.617\u0026ndash;4.752)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.335\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePeritumoral edema\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5.480 (1.581\u0026ndash;25.261)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.013\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.960 (0.424\u0026ndash;10.969)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.405\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2.625 (0.621\u0026ndash;18.028)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.238\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMaximum diameter\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.062 (0.915\u0026ndash;1.238)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.432\u003c/p\u003e \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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.888 (0.699\u0026ndash;1.118)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.316\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMidline shift\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.972 (0.565\u0026ndash;1.671)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.918\u003c/p\u003e \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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1.184 (0.488\u0026ndash;2.895)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.709\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eEnhancement pattern\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.465 (1.117\u0026ndash;1.944)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.006\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.085 (0.765\u0026ndash;1.552)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.649\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1.492 (0.966\u0026ndash;2.348)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.075\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eEnhancement quality\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.333 (1.97\u0026ndash;10.127)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2.568 (0.832\u0026ndash;8.27)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.105\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1.396 (0.518\u0026ndash;4.067)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.521\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTCM\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.387 (0.692\u0026ndash;2.752)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.350\u003c/p\u003e \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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.733 (0.264\u0026ndash;2.113)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.555\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eECM\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.462 (0.756\u0026ndash;2.804)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.254\u003c/p\u003e \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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.338 (0.129\u0026ndash;0.866)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.025\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.211 (0.067\u0026ndash;0.602)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.005\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCortical involvement\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.892 (0.496\u0026ndash;1.621)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.705\u003c/p\u003e \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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1.400(0.526\u0026ndash;3.651)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.493\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDeep white matter invasion\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.024 (0.562\u0026ndash;1.894)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.939\u003c/p\u003e \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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e3.594 (1.344\u0026ndash;9.956)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.012\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e5.688 (1.912\u0026ndash;18.609)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.003\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePial invasion\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.423 (0.822\u0026ndash;2.465)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.208\u003c/p\u003e \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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2.217 (0.909\u0026ndash;5.575)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.084\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eEpendymal invasion\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.350 (0.725\u0026ndash;2.496)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.340\u003c/p\u003e \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\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2.333 (0.857\u0026ndash;7.121)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.112\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eGBM: glioblastoma; IDH-mut: Isocitrate dehydrogenase-mutant; IDH-wt: Isocitrate dehydrogenase wild-type; OR: odds ratio; CI: confidence interval; MGMTmet: methyl guanine methyl transferase promoter methylation; KPS: karnofsky performance status score; TCM: tumor crosses midline, ECM: edema crosses midline.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe AUCs of the clinical models were 0.671 and 0.619 (training set) and 0.656 and 0.605 (validation set) for Tasks 1 and Tasks 2, respectively (Fig.\u0026nbsp;3 and Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe diagnostic performance of clinical model, ML model, and combined model in the training and validation sets of the two tasks\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"12\"\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 \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\u003eTask 1 (grade 4 vs GBM)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"5\" nameend=\"c12\" namest=\"c8\"\u003e \u003cp\u003eTask 2 (IDH-mut \u003csup\u003egrade 4\u003c/sup\u003e vs IDH-wt \u003csup\u003egrade 4\u003c/sup\u003e)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSet\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003cp\u003e(95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCut-off\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSEN\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eSPE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eACC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003cp\u003e(95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eCut-off\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c10\"\u003e \u003cp\u003eSEN\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c11\"\u003e \u003cp\u003eSPE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c12\"\u003e \u003cp\u003eACC\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eT1C Edema\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTraining\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.843\u003c/p\u003e \u003cp\u003e(0.787\u0026ndash;0.892)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.667\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.707\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.833\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.754\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.838\u003c/p\u003e \u003cp\u003e(0.745\u0026ndash;0.912)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.399\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.700\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.842\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.793\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eValidation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.829\u003c/p\u003e \u003cp\u003e(0.734\u0026ndash;0.911)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.505\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.875\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.750\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.823\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.792\u003c/p\u003e \u003cp\u003e(0.618\u0026ndash;0.941)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.385\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.818\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.692\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.730\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eT1C Tumor\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTraining\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.825\u003c/p\u003e \u003cp\u003e(0.768\u0026ndash;0.877)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.681\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.700\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\u003e0.732\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.735\u003c/p\u003e \u003cp\u003e(0.623\u0026ndash;0.844)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.382\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.800\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.614\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.678\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eValidation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.722\u003c/p\u003e \u003cp\u003e(0.608\u0026ndash;0.830)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.584\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.821\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.625\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.740\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.750\u003c/p\u003e \u003cp\u003e(0.515\u0026ndash;0.926)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.373\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.818\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.654\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.703\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eT2F Tumor\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTraining\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.777\u003c/p\u003e \u003cp\u003e(0.711\u0026ndash;0.838)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.558\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.886\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\u003e0.768\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.722\u003c/p\u003e \u003cp\u003e(0.612\u0026ndash;0.827)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.395\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.733\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.649\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.678\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eValidation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.629\u003c/p\u003e \u003cp\u003e(0.491\u0026ndash;0.741)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.549\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.821\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.425\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.656\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.640\u003c/p\u003e \u003cp\u003e(0.421\u0026ndash;0.817)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.370\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.909\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.423\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.568\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eOptimal ML\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTraining\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.902\u003c/p\u003e \u003cp\u003e(0.861\u0026ndash;0.939)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.619\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.821\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.821\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.821\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.904\u003c/p\u003e \u003cp\u003e(0.836\u0026ndash;0.953)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.375\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.933\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.789\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.839\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eValidation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.855\u003c/p\u003e \u003cp\u003e(0.769\u0026ndash;0.928)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.675\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.732\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.875\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.792\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.895\u003c/p\u003e \u003cp\u003e(0.769\u0026ndash;0.991)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.380\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.909\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.846\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.864\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eClinical\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTraining\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.671\u003c/p\u003e \u003cp\u003e(0.596\u0026ndash;0.743)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.664\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.693\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.583\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.652\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.619\u003c/p\u003e \u003cp\u003e(0.515\u0026ndash;0.721)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.400\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.467\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.772\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.667\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eValidation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.656\u003c/p\u003e \u003cp\u003e(0.492\u0026ndash;0.766)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.680\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.357\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.875\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.573\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.605\u003c/p\u003e \u003cp\u003e(0.453\u0026ndash;0.766)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.400\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.364\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.846\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.703\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCombined\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTraining\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.911\u003c/p\u003e \u003cp\u003e(0.869\u0026ndash;0.945)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.531\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.871\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\u003e0.839\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.902\u003c/p\u003e \u003cp\u003e(0.838\u0026ndash;0.952)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.398\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.933\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.789\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.839\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eValidation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.854\u003c/p\u003e \u003cp\u003e(0.769\u0026ndash;0.928)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.674\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.750\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.825\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.781\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.909\u003c/p\u003e \u003cp\u003e(0.794\u0026ndash;0.987)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.477\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.909\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0.846\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0.865\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\u003eGBM: glioblastoma; IDH-mut: Isocitrate dehydrogenase-mutant; IDH-wt: Isocitrate dehydrogenase wild-type; AUC: area under curve; CI: confidence interval; SEN: sensitivity; SPE: specificity; ACC: accuracy; T1C: contrast-enhanced T1-weighted imaging; T2F: T2-weighted imaging fluid attenuated inversion recovery; ML: machine learning.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eConstruction of the ML model\u003c/h2\u003e \u003cp\u003eA total of 1688 radiomics features were extracted. After feature selection and reduction, 11 T1C edema, 14 T1C tumor, and 10 T2F tumor features for Task 1 and 9 T1C edema, 6 T1C tumor, and 3 T2F tumor features for Task 2 were retained. These features were then amalgamated separately, resulting in final sets of 35 and 18 features for constructing combined-sequence ML models for the two tasks.\u003c/p\u003e \u003cp\u003eThe combined-sequence ML model achieved the highest AUCs in the validation set for Task 1 (training set\u0026thinsp;=\u0026thinsp;0.902, validation set\u0026thinsp;=\u0026thinsp;0.855) and Task 2 (training set\u0026thinsp;=\u0026thinsp;0.904, validation set\u0026thinsp;=\u0026thinsp;0.895), making them the optimal ML models (Fig.\u0026nbsp;3 and Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Rad-score was calculated.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eConstruction of the combined model and nomogram\u003c/h2\u003e \u003cp\u003eIn Task 1, the combined model had the highest AUC in the training set (0.911) but was slightly lower than the optimal ML model in the validation set (0.854). In Task 2, the combined model had the highest AUC in the validation set (0.909) but lower than the optimal ML model in the training set (0.902) (Fig.\u0026nbsp;3). Nomogram based on the combined model further facilitates individualized discrimination between grade 4 astrocytoma and GBM (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e4\u003c/span\u003ea). Their representative MR Images and case nomograms are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e4\u003c/span\u003eb-e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eModel evaluation and comparison\u003c/h2\u003e \u003cp\u003eDeLong's test demonstrated significant differences in AUCs between the optimal ML model and the combined model, as compared to the clinical model for both tasks in the training and validation sets (Task 1: all \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001; Task 2: training set \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001, validation set \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.011 and \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.004). No significant differences were found between the combined and optimal ML models in the training (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.148, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.907) and validation (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.934, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.514) sets for both tasks. Delong\u0026rsquo;s test results are shown in \u003cb\u003eFig.\u0026nbsp;3c, d, g, h\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eThe combined model demonstrated good calibration in both tasks, though moderate in the validation set for Task 2 (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003ea, b, e, f). Decision curve analysis (DCA) indicated that the combined model and the optimal ML model provided significantly better net benefits in predicting glioma molecular subtypes compared to the clinical model. For Task 1, threshold probabilities were 0.12 to 0.92 and 0.02 to 0.96 (training set, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003ec), and 0.18 to 0.86 and 0.08 to 0.92 (validation set, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003ed). For Task 2, threshold probabilities ranged from 0.04 to 0.70 and 0.02 to 0.98 (training set, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003eg), and from 0.06 to 0.66 and 0.02 to 0.96 (validation set, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003eh).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe combined model offered a limited net benefit compared to the optimal ML model, which offered greater net benefits across multiple threshold ranges.\u003c/p\u003e \u003cp\u003eWe used six other ML methods for building the combined model to test its stability and reliability. In both tasks, the RF model achieved prediction performances of 1.0 and 0.998 in the training sets, outperforming other models, while XGBoost showed slightly lower performance than some others. The remaining models performed similarly, with validation set AUCs ranging from 0.854 to 0.881 (Task 1) and 0.958 (Task 2) to 0.979, respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e6\u003c/span\u003ea-h). DeLong's test revealed no statistically significant differences in AUCs among models in the training sets, except for RF and XGBoost (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e6\u003c/span\u003ei, k), and no significant differences across all models in the validation sets (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e6\u003c/span\u003ej, l).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eSurvival analysis\u003c/h2\u003e \u003cp\u003eUsing the training set cutoff values from the combined model for the two tasks (0.531 and 0.398), patients were divided into high-risk and low-risk groups. In Task 1 (489 vs 993 days) and Task 2 (721 vs 1297 days), the average OS of low-risk patients was significantly longer than that of high-risk patients, reflecting the different prognoses associated with various molecular subtypes. KM analysis revealed significant differences between these groups for the combined model and molecular subtype in Task 1 and Task 2 (all \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Z-test indicated no statistically significant difference in prognostic value between the molecular subtype and combined model in both tasks (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.966, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.793) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e7\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study constructed two machine learning tasks to predict molecular subtypes of 2021 WHO grade 4 glioma using multiparametric MRI, clinical-radiological characteristics, and their combination. The ML model performed well in distinguishing grade 4 astrocytoma from GBM and discriminating IDH-mut grade 4 astrocytoma from IDH-wt grade 4 astrocytoma. Additionally, the combined model effectively stratified cases into high-risk and low-risk groups according to OS, with prognostic performance comparable to molecular subtype.\u003c/p\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003ePredictive Value of the Clinical Model\u003c/h2\u003e \u003cp\u003eOur study identified age and MGMTmet status as significant predictors for distinguishing grade 4 astrocytoma from GBM (Task 1) with AUCs of 0.671 (training set) and 0.656 (validation set). ECM and deep white matter invasion were significant predictors for differentiating IDH-mut from IDH-wt grade 4 astrocytoma (Task 2) with AUCs of 0.619 (training set) and 0.605 (validation set).\u003c/p\u003e \u003cp\u003eGBM patients were older on average than grade 4 astrocytoma patients (55.3 vs 50.1 years), and IDH-wt patients were older than IDH-mut patients (52.3 vs 45.6 years), consistent with a previous study[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Older age may be associated with higher malignancy potential. Studies indicated that younger age is associated with better prognosis, whereas older age is linked to poor survival in adult glioma patients[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], and our study indirectly supports this. Additionally, a higher proportion of MGMTmet was observed in grade 4 astrocytoma compared to GBM (67.5% vs 49.0%), suggesting it may be associated with less aggressive tumor behavior. Previous studies have shown that MGMTmet is linked to an improved response to temozolomide and longer OS[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e], consistent with our findings. IDH-wt patients exhibited higher rates of ECM (79.5% vs 56.1%) and lower rates of deep white matter invasion (16.9% vs 46.3%) compared to IDH-mut patients.\u003c/p\u003e \u003cp\u003eDespite these findings, the clinical model demonstrated limited predictive power, underscoring the challenge of relying solely on clinical-radiological features for precise molecular subtypes of 2021 WHO glioma. This limitation is likely due to the inherent heterogeneity of glioma, where clinical and radiological characteristics may not fully capture the underlying molecular alterations.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003ePredictive Value of the ML Model\u003c/h2\u003e \u003cp\u003eThe ML model significantly outperformed the clinical models in distinguishing grade 4 astrocytoma from GBM and discriminating IDH-mut from IDH-wt grade 4 astrocytoma. The optimal ML model for both tasks has strong predictive performance in the training (AUC\u0026thinsp;=\u0026thinsp;0.902 and 0.904) and validation (0.855 and 0.895) sets. Currently, only Wei et al.[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e] conducted a relevant study in which they developed a subregion-based MRI RadioFusionOmics model to discriminate between grade 4 astrocytoma and GBM, achieving AUCs of 0.976 and 0.974 for the training and validation cohorts, respectively, consistent with our finding. However, we further stratified grade 4 astrocytoma into IDH-mut and IDH-wt subtypes and evaluated the prognostic value of the combined model. Previous studies employing radiomics or ML models to distinguish IDH-mut from IDH-wt GBM were based on the 2016 WHO CNS criteria[\u003cspan additionalcitationids=\"CR25 CR26\" citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. No studies currently have addressed differentiating IDH-mut and IDHwt astrocytoma reclassified as grade 4 astrocytoma under the 2021 WHO CNS criteria.\u003c/p\u003e \u003cp\u003eOur multiparametric MRI includes T1C edema, T1C tumor, and T2F tumor. In both tasks, T1C edema, besides the optimal ML model and the combined model, showed the highest AUC for the training (0.843 and 0.838) and validation (0.829 and 0.792) sets. The peritumoral region adjacent to GBM, a mix of infiltrative tumor and vasogenic edema, is often the site of recurrence[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. Thus, tumor and edema regions in grade 4 glioma reflect tumor heterogeneity. Although peritumoral edema is not a significant predictor of the 2021 glioma molecular subtypes, its proportion in GBM is higher than in grade 4 astrocytoma (98.5% vs 90.5%). This suggests peritumoral edema may encompass potential heterogeneity between grade 4 astrocytoma and GBM, consistent with a previous study[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e].\u003c/p\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003ePredictive Value of the Combined Model\u003c/h2\u003e \u003cp\u003eThe combined model, integrating Rad-score and clinical-radiological characteristics, aimed to leverage the strengths of both approaches. However, its performance was similar to the optimal ML model. In Task 1, the AUCs for the training and validation sets were 0.911 vs 0.902 (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.148) and 0.854 vs 0.855 (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.934), respectively. In Task 2, the AUCs for the training and validation sets were 0.902 vs 0.904 (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.907) and 0.909 vs 0.895 (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.514), respectively. These findings indicate limited added value from incorporating clinical-radiological features.\u003c/p\u003e \u003cp\u003eThe nomogram based on the combined model of Task 1 offers a practical tool for individualized risk prediction and clinical decision-making, providing an intuitive and clinically interpretable visual representation to aid clinicians in stratifying patients and tailoring treatment strategies.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003eModel evaluation and comparison\u003c/h2\u003e \u003cp\u003eWe evaluated the combined model using various ML algorithms, including LR, SVM, MLP, LDA, RF, and NB. Most algorithms showed consistent performance. This consistency underscores the reliability of the combined model and its potential for broad clinical application.\u003c/p\u003e \u003cdiv id=\"Sec25\" class=\"Section3\"\u003e \u003ch2\u003eSurvival analysis\u003c/h2\u003e \u003cp\u003eWe further evaluated the prognostic value of the combined model. It effectively stratified patients into high-risk and low-risk groups, with prognostic value comparable to molecular subtype. This confirmed that our model can accurately predict glioma molecular subtype and holds substantial prognostic value, offering a new perspective for clinical decision-making.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section3\"\u003e \u003ch2\u003eLimitations\u003c/h2\u003e \u003cp\u003eThis study has several limitations. Firstly, the retrospective nature and reliance on data from three institutions may introduce selection bias. Prospective validation on a larger multicenter cohort is necessary to confirm the findings. Secondly, different equipment and scanning parameters used across the three institutions, despite image preprocessing, may have influenced the radiomic features. Future studies should aim for uniform scanning parameters. Thirdly, although this study included multi-sequence MRI of tumor and edema regions, future research should incorporate more functional imaging modalities, such as diffusion-weighted imaging (DWI) and perfusionweighted imaging (PWI), to further enhance the model's predictive power.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn conclusion, the multiparametric MRI machine learning model effectively differentiated grade 4 astrocytoma from GBM and distinguished between IDH-mut and IDH-wt grade 4 astrocytoma. Additionally, the model stratified various molecular subtypes of glioma patients into high-risk and low-risk groups according to OS, offering a new perspective for clinical decision-making.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eACC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eAccuracy\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eAUC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eArea under the curve\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eAMP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eAmplification\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eCDKN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eCyclin-dependent kinase inhibitor\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eCE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eContrast-enhanced\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eCI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eConfidence interval\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eCNS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eCentral nervous system\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eDCA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eDecision curve analysis\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eDWI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eDiffusion-weighted imaging\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eECM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eEdema crosses midline\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eEGFR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eEpidermal growth factor receptor\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eFLAIR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eFluid attenuated inversion recovery\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eGBM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eGlioblastoma\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eGLCM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eGray Level CO-Occurrence Matrix\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eGLDM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eGray Level Dependence Matrix\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eGLRLM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eGray-Level Run-Length Matrix\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eGLSZM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eGray-level size zone matrix\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eICC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eIntraclass correlation coefficient\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eIDH-mut\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eIsocitrate dehydrogenase mutant\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eIDH-wt\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eIsocitrate dehydrogenase wild-type\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eKM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eKaplan-Meier\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eKPS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eKarnofsky performance status\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eLASSO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eLeast absolute shrinkage and selection operator\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eLDA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eLinear discriminant analysis\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eLR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eLogistic regression\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003emGBM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eMolecular features of glioblastoma\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eMGMTmet\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eMethyl guanine methyl transferase promoter methylation\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eML\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eMachine learning\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eMRI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eMagnetic resonance imaging\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eMLP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eMulti-layer perceptrons\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eNB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eNaive Bayes\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eNGTDM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eNeighborhood Gray Tone Difference Matrix\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eOR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eOdds ratio\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eOS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eOverall survival\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003ePFS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eProgression-free survival\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003ePWI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003ePerfusion-weighted imaging\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eRF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eRandom forest\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eROC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eReceiver operating characteristic\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eROI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eRegion of interest\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eSEN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eSPE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eSpecificity\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eSVM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eSupport vector machines\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eTCM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eTumor crosses midline\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eTERT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eTelomerase reverse transcriptase\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eT1WI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eT1-weighted imaging\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eT2WI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eT2-weighted imaging\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eVOI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eVolume of interest\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eWHO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eWorld Health Organization\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 25.4973%;\"\u003e\n \u003cp\u003eXGBoost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74.5027%;\"\u003e\n \u003cp\u003eExtreme Gradient Boosting\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Ethics Board of the First Hospital of Shanxi Medical University (approval number:\u0026nbsp;KYYJ-2023-058) has approved for this study.\u0026nbsp;Patient consent was waived due to the retrospective nature of the study\u0026nbsp;and the anonymous nature of the data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWritten informed consent for publication was obtained from all participants.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated or analyzed during the study are not publicly available due to institutional regulations but are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the National Natural Science Foundation of China [grant numbers 82071893, 82371941 to\u0026nbsp;Yan Tan]; the Research Project Supported by Shanxi Scholarship Council of China [grant number 2023-186 to\u0026nbsp;Yan Tan]; and Shanxi Province Higher Education \"Billion Project\" Science and Technology Guidance Project\u0026nbsp;[grant number\u0026nbsp;BYJL017 to\u0026nbsp;Yan Tan].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors' contributions:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWenji Xu\u003c/strong\u003e: Data processing, Writing - original draft. \u003cstrong\u003eYangyang Li\u003c/strong\u003e: Data processing, Writing - original draft. \u003cstrong\u003eJie Zhang\u003c/strong\u003e: Data curation. \u003cstrong\u003eZhiyi Zhang\u003c/strong\u003e and \u003cstrong\u003ePengxin Shen\u003c/strong\u003e: Data processing, Software. \u003cstrong\u003eXiaochun Wang\u003c/strong\u003e: Writing - review \u0026amp; editing, Project administration. \u003cstrong\u003eGuoqiang Yang\u003c/strong\u003e: Data processing, Writing - review \u0026amp; editing, Project administration. \u003cstrong\u003eJiangfeng Du\u003c/strong\u003e: Data processing, Writing - review \u0026amp; editing. \u003cstrong\u003eHui Zhang\u003c/strong\u003e: Supervision, Writing - review \u0026amp; editing. \u003cstrong\u003eYan Tan\u003c/strong\u003e: Methodology, Writing - review \u0026amp; editing, Project administration.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eLouis DN, Perry A, Wesseling P, Brat DJ, Cree IA, Figarella-Branger D, et al. The 2021 WHO Classification of Tumors of the Central Nervous System: a summary. Neuro Oncol. 2021;23:1231\u0026ndash;51. \u003c/li\u003e\n\u003cli\u003eHorbinski C, Berger T, Packer RJ, Wen PY. Clinical implications of the 2021 edition of the WHO classification of central nervous system tumours. Nat Rev Neurol. 2022;18:515\u0026ndash;29. \u003c/li\u003e\n\u003cli\u003eRamos-Fresnedo A, Pullen MW, Perez-Vega C, Domingo RA, Akinduro OO, Almeida JP, et al. The survival outcomes of molecular glioblastoma IDH-wildtype: a multicenter study. J Neurooncol. 2022;157:177\u0026ndash;85. \u003c/li\u003e\n\u003cli\u003eGritsch S, Batchelor TT, Gonzalez Castro LN. Diagnostic, therapeutic, and prognostic implications of the 2021 World Health Organization classification of tumors of the central nervous system. Cancer. 2022;128:47\u0026ndash;58. \u003c/li\u003e\n\u003cli\u003eLambin P, Leijenaar RTH, Deist TM, Peerlings J, De Jong EEC, Van Timmeren J, et al. Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol. 2017;14:749\u0026ndash;62. \u003c/li\u003e\n\u003cli\u003eLambin P, Rios-Velazquez E, Leijenaar R, Carvalho S, Van Stiphout RGPM, Granton P, et al. Radiomics: Extracting more information from medical images using advanced feature analysis. European Journal of Cancer. 2012;48:441\u0026ndash;6. \u003c/li\u003e\n\u003cli\u003eAvanzo M, Wei L, Stancanello J, Valli\u0026egrave;res M, Rao A, Morin O, et al. Machine and deep learning methods for radiomics. Med Phys. 2020;47:e185\u0026ndash;202. \u003c/li\u003e\n\u003cli\u003eMoodi F, Khodadadi Shoushtari F, Ghadimi DJ, Valizadeh G, Khormali E, Salari HM, et al. Glioma Tumor Grading Using Radiomics on Conventional MRI : A Comparative Study of WHO 2021 and WHO 2016 Classification of Central Nervous Tumors. Magnetic Resonance Imaging. 2023;jmri.29146. \u003c/li\u003e\n\u003cli\u003eTian Q, Yan L, Zhang X, Zhang X, Hu Y, Han Y, et al. Radiomics strategy for glioma grading using texture features from multiparametric MRI. Magnetic Resonance Imaging. 2018;48:1518\u0026ndash;28. \u003c/li\u003e\n\u003cli\u003eChiu F-Y, Le NQK, Chen C-Y. A Multiparametric MRI-Based Radiomics Analysis to Efficiently Classify Tumor Subregions of Glioblastoma: A Pilot Study in Machine Learning. JCM. 2021;10:2030. \u003c/li\u003e\n\u003cli\u003eLin K, Cidan W, Qi Y, Wang X. Glioma grading prediction using multiparametric magnetic resonance imaging‐based radiomics combined with proton magnetic resonance spectroscopy and diffusion tensor imaging. Medical Physics. 2022;49:4419\u0026ndash;29. \u003c/li\u003e\n\u003cli\u003eDing J, Zhao R, Qiu Q, Chen J, Duan J, Cao X, et al. Developing and validating a deep learning and radiomic model for glioma grading using multiplanar reconstructed magnetic resonance contrast-enhanced T1-weighted imaging: a robust, multi-institutional study. Quant Imaging Med Surg. 2022;12:1517\u0026ndash;28. \u003c/li\u003e\n\u003cli\u003eVijithananda SM, Jayatilake ML, Gon\u0026ccedil;alves TC, Rato LM, Weerakoon BS, Kalupahana TD, et al. Texture feature analysis of MRI-ADC images to differentiate glioma grades using machine learning techniques. Sci Rep. 2023;13:15772. \u003c/li\u003e\n\u003cli\u003eXing X, Zhu M, Chen Z, Yuan Y. Comprehensive learning and adaptive teaching: Distilling multi-modal knowledge for pathological glioma grading. Medical Image Analysis. 2024;91:102990. \u003c/li\u003e\n\u003cli\u003eMalik N, Geraghty B, Dasgupta A, Maralani PJ, Sandhu M, Detsky J, et al. MRI radiomics to differentiate between low grade glioma and glioblastoma peritumoral region. J Neurooncol. 2021;155:181\u0026ndash;91. \u003c/li\u003e\n\u003cli\u003eSzekeres D, Jetty SN, Soni N. The Role of Multiparametric MRI in Diagnosing and Grading Glioma. Neurology India. 2023;71:1274\u0026ndash;5. \u003c/li\u003e\n\u003cli\u003eNaser MA, Deen MJ. Brain tumor segmentation and grading of lower-grade glioma using deep learning in MRI images. Computers in Biology and Medicine. 2020;121:103758. \u003c/li\u003e\n\u003cli\u003eLee D, Riestenberg RA, Haskell-Mendoza A, Bloch O. Diffuse astrocytic glioma, IDH-Wildtype, with molecular features of glioblastoma, WHO grade IV: A single-institution case series and review. J Neurooncol. 2021;152:89\u0026ndash;98. \u003c/li\u003e\n\u003cli\u003eWeller M, Van Den Bent M, Preusser M, Le Rhun E, Tonn JC, Minniti G, et al. EANO guidelines on the diagnosis and treatment of diffuse gliomas of adulthood. Nat Rev Clin Oncol. 2021;18:170\u0026ndash;86. \u003c/li\u003e\n\u003cli\u003ePark YW, Kim S, Park CJ, Ahn SS, Han K, Kang S-G, et al. Adding radiomics to the 2021 WHO updates may improve prognostic prediction for current IDH-wildtype histological lower-grade gliomas with known EGFR amplification and TERT promoter mutation status. Eur Radiol. 2022;32:8089\u0026ndash;98. \u003c/li\u003e\n\u003cli\u003eAgarwal A, Edgar MA, Desai A, Gupta V, Soni N, Bathla G. Molecular GBM versus Histopathological GBM: Radiology-Pathology-Genetic Correlation and the New WHO 2021 Definition of Glioblastoma. AJNR Am J Neuroradiol. 2024;ajnr.A8225. \u003c/li\u003e\n\u003cli\u003eZeng C, Song X, Zhang Z, Cai Q, Cai J, Horbinski C, et al. Dissection of transcriptomic and epigenetic heterogeneity of grade 4 gliomas: implications for prognosis. acta neuropathol commun. 2023;11:133. \u003c/li\u003e\n\u003cli\u003eWei R, Lu S, Lai S, Liang F, Zhang W, Jiang X, et al. A subregion-based RadioFusionOmics model discriminates between grade 4 astrocytoma and glioblastoma on multisequence MRI. J Cancer Res Clin Oncol. 2024;150:73. \u003c/li\u003e\n\u003cli\u003ePasquini L, Napolitano A, Tagliente E, Dellepiane F, Lucignani M, Vidiri A, et al. Deep Learning Can Differentiate IDH-Mutant from IDH-Wild GBM. J Pers Med. 2021;11. \u003c/li\u003e\n\u003cli\u003eCalabrese E, Villanueva-Meyer JE, Cha S. A fully automated artificial intelligence method for non-invasive, imaging-based identification of genetic alterations in glioblastomas. Sci Rep. 2020;10:11852. \u003c/li\u003e\n\u003cli\u003eKandalgaonkar P, Sahu A, Saju AC, Joshi A, Mahajan A, Thakur M, et al. Predicting IDH subtype of grade 4 astrocytoma and glioblastoma from tumor radiomic patterns extracted from multiparametric magnetic resonance images using a machine learning approach. Front Oncol. 2022;12:879376. \u003c/li\u003e\n\u003cli\u003eCheng J, Liu J, Yue H, Bai H, Pan Y, Wang J. Prediction of Glioma Grade Using Intratumoral and Peritumoral Radiomic Features From Multiparametric MRI Images. IEEE/ACM Trans Comput Biol Bioinform. 2022;19:1084\u0026ndash;95. \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":"Astrocytoma, Glioblastoma, Magnetic resonance imaging, Machine learning; Molecular subtype","lastPublishedDoi":"10.21203/rs.3.rs-5288001/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5288001/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003ePurpose\u003c/strong\u003e: To develop and validate a machine learning (ML) model using multiparametric MRI for the preoperative differentiation of 2021 World Health Organization (WHO) grade 4 astrocytoma and glioblastoma (GBM) (Task 1), and to stratify grade 4 astrocytoma to distinguish isocitrate dehydrogenase-mutant (IDH-mut) from IDH-wild-type (IDH-wt) (Task 2). Additionally, to evaluate the model’s prognostic value.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMaterials and methods:\u003c/strong\u003e We retrospectively analyzed 320 glioma patients from three hospitals. Cases were randomly divided into training and validation sets with a 7:3 ratio. Features were extracted from tumor and edema on contrast-enhanced T1-weighted imaging (CE-T1WI) and T2 fluid-attenuated inversion recovery (T2-FLAIR). Extreme gradient boosting (XGBoost) was utilized for constructing ML, clinical, and combined models. Model performance was evaluated with receiver operating characteristic (ROC) curves, decision curves, and calibration curves. Stability was evaluated using six additional classifiers. Kaplan-Meier (KM) survival analysis and the log-rank test assessed the model’s prognostic value.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e: In Task 1 (grade 4 vs GBM) and Task 2 (IDH-mut grade 4 vs IDH-wt grade 4), the combined model (AUC = 0.911 and 0.854, 0.902 and 0.909) and the optimal ML model (AUC = 0.902 and 0.855, 0.904 and 0.895) significantly outperformed the clinical model (AUC = 0.671 and 0.656, 0.619 and 0.605) in both the training and validation sets. Survival analysis showed the combined model performed similarly to molecular subtype in both tasks (\u003cem\u003eP\u003c/em\u003e = 0.966 and \u003cem\u003eP\u003c/em\u003e = 0.793).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e: The multiparametric MRI ML model effectively distinguished grade 4 astrocytoma from GBM and differentiated IDH-mut from IDH-wt grade 4 astrocytoma. Additionally, the model provides reliable survival stratification for glioma patients with various molecular subtypes.\u003c/p\u003e","manuscriptTitle":"Predicting the Molecular Subtypes of 2021 WHO Grade 4 Glioma by a Multiparametric MRI-Based Machine Learning Model","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-11-11 07:01:13","doi":"10.21203/rs.3.rs-5288001/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":"e58d3bfb-7e63-478a-a6d7-d47b74a2847c","owner":[],"postedDate":"November 11th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-03-06T12:38:22+00:00","versionOfRecord":[],"versionCreatedAt":"2024-11-11 07:01:13","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5288001","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5288001","identity":"rs-5288001","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}