Radiomic features from the peritumoral region can be associated with epilepsy status of glioblastoma patients

Yeong Chul Yun^1,2,3*Johann M. E. Jende^1,2Katharina Holz^1,3Freya Garhöfer^1,3Sabine Wolf^1,3Anja Hohmann⁴Philipp Vollmuth¹Martin Bendszus¹Heinz-Peter Schlemmer²Felix Sahm^5,6Sabine Heiland¹Wolfgang Wick⁴Varun Venkataramani^4,7Felix T. Kurz^1,2,8

• ¹Department of Neuroradiology, Heidelberg University Hospital, Heidelberg, Germany
• ²Division of Radiology, German Cancer Research Center (DKFZ), Heidelberg, Baden-Württemberg, Germany
• ³Medical Faculty Heidelberg, University of Heidelberg, Heidelberg, Baden-Württemberg, Germany
• ⁴Neurologische Klinik, UniversitätsKlinikum Heidelberg, Heidelberg, Baden-Württemberg, Germany
• ⁵Clinical Cooperation Unit Neuropathology, German Cancer Consortium (DKTK) and German Cancer Research Center (DKFZ), Heidelberg, Germany
• ⁶Department of Neuropathology, Heidelberg University Hospital, Heidelberg, Germany
• ⁷Department of Functional Neuroanatomy, Heidelberg University, Heidelberg, Germany
• ⁸Department of Neuroradiology, Geneva University Hospitals, Geneva, Switzerland

Purpose: Identifying radiomics features that help to predict whether glioblastoma patients are prone to developing epilepsy may contribute to an improvement of preventive treatment and a better understanding of the underlying pathophysiology.In this retrospective study, 3 Tesla MRI data of 451 pre-treatment glioblastoma patients (mean age: 61.2 years ±11.8; 268 male, 183 female) were analyzed.336 patients reported no epilepsy, while 115 patients were diagnosed with symptomatic epilepsy. A total of 1546 radiomics features were extracted from contrast-enhancing tumor, peritumoral regions and normal-appearing white matter as regions of interest using PyRadiomics. The dataset was initially split into a training (70%) and a validation (30%) cohort. The training cohort was used for feature selection with ElasticNet and modeloptimization. Various machine-learning models, including logistic regression (LR), were used to predict epilepsy status. The models' performances were evaluated with the validation cohort and the area under the curve of receiver operating characteristics (AUC) was used as a measure. For identifying relevant features, permutation feature importance was applied.The performance of LR using radiomics features from only a single ROI in the validation cohort were AUC=0.83 (95%-CI: 0.76-0.91) and AUC=0.77 (95% CI: 0.69-0.85) for peritumoral and white matter region respectively. Most important features in peritumoral regions were shape features, while for the white matter region higher-order features from FLAIR were most relevant.Radiomics features from peritumoral and normal-appearing white matter can be associated with epilepsy status at diagnosis suggesting an important role of these regions for the development of epilepsy in glioblastoma patients.