RADIOMICS-BASED MACHINE LEARNING FOR NON-INVASIVE PREDICTION OF IDH MUTATION STATUS IN DIFFUSE GLIOMAS: A SINGLE-CENTER STUDY FROM DOW UNIVERSITY OF HEALTH SCIENCES, OJHA CAMPUS, KARACHI, PAKISTAN

Abstract

Background: Isocitrate dehydrogenase (IDH) mutation status is a major prognostic biomarker in diffuse gliomas and a defining element of contemporary WHO classification, but it is usually established through invasive tissue sampling.

Objective: To develop and internally validate a machine-learning model using radiomics features extracted from routine MRI for prediction of IDH mutation status in diffuse gliomas.

Methods: This retrospective single-center study was conducted at the Department of Radiology, Dow University of Health Sciences (DUHS), Ojha Campus, Karachi, Pakistan. Adult patients with histopathologically confirmed diffuse gliomas and preoperative MRI performed between August 2025 and February 2026 were included. Imaging was acquired on a 3T Siemens MAGNETOM Skyra scanner. Tumor regions were manually segmented by two experienced neuroradiologists, and inter-observer agreement was assessed using the intraclass correlation coefficient. Radiomics features were extracted with PyRadiomics after N4 bias field correction, intensity normalization, and isotropic resampling. Feature selection was performed with least absolute shrinkage and selection operator (LASSO) using 5-fold cross-validation. Logistic regression, support vector machine, and random forest classifiers were trained and evaluated with stratified 5-fold cross-validation.

Results: A total of 158 patients were included, with a mean age of 44.1 ± 12.9 years; 39.9% had IDH-mutant tumors. Nine stable radiomics features were retained after feature selection. The best-performing model achieved an area under the ROC curve of 0.85, with sensitivity of 81.7%, specificity of 79.4%, and accuracy of 80.5% under cross-validation. These results are consistent with prior radiomics literature reporting good-to-excellent discrimination for IDH prediction.

Conclusion: Radiomics features derived from routine MRI demonstrated promising performance for non-invasive prediction of IDH mutation status in diffuse gliomas. External multicenter validation is needed before clinical application.