Optimizing machine learning-based multimodal radiomics for predicting <i>IDH</i> status in gliomas: A SHAP-based multicenter study.
Authors
Affiliations (7)
Affiliations (7)
- Department of Radiology, the First Affiliated Hospital of Shenzhen University, Health Science Center, Shenzhen Second People's Hospital, Shenzhen, China.
- Department of Radiology, Guangdong Provincial People's Hospital, Guangdong Academy of Medical Sciences, Guangzhou, Guangdong, China.
- School of Public Health, Southern Medical University, Guangzhou, China.
- The First People's Hospital of Foshan (The Affiliated Foshan Hospital of Southern University of Science and Technology), School of Medicine, Southern University of Science and Technology, Guangdong, China.
- Department of Radiology, Peking University Shenzhen Hospital, Shenzhen, Guangdong Province, China.
- Medical Imaging Center, Southern Medical University Shenzhen hospital, Shenzhen, Guangdong Province, China.
- Department of Radiology, The Fourth People's Hospital of Shenzhen (Shenzhen Samii Medical Center), Shenzhen, Guangdong, China.
Abstract
BackgroundPredicting isocitrate dehydrogenase (<i>IDH</i>) status is crucial in glioma management. Conventional MRI (cMRI) has limitations, but the clinical translation of radiomics and machine learning (ML) is often limited by single-center datasets and poor model interpretability.PurposeTo develop and validate an interpretable, multicenter ML model integrating cMRI with functional sequences (DWI and PWI) for predicting <i>IDH</i> status in gliomas.Material and MethodsThis retrospective study included 180 patients from four institutions (150 training, 30 external test). Radiomics features were extracted from cMRI (T1WI, T2WI, FLAIR, T1CE), DWI, and DSC-PWI (CBV maps). After feature selection, multiparametric MRI-based fusion radiomics models were built and compared using three ML algorithms across four segmentation strategies. The optimal model was explained using SHapley Additive exPlanation (SHAP).ResultsThe full-modality model (cMRI + DWI + PWI) with 3Dmodified segmentation achieved the best performance, with area under the curve of 0.840 (training) and 0.810 (external test). Incorporating functional sequences significantly improved prediction over cMRI alone. SHAP analysis identified key predictive features and provided individualized visual explanations for model decisions.ConclusionThe developed ML-SHAP model, integrating conventional and functional MRI, reliably predicts <i>IDH</i> status and demonstrates generalizability across multiple centers. This interpretable tool shows potential for supporting preoperative molecular diagnosis in glioma.