Deep Neural Network-Based Risk Prediction of Glioblastoma Multiforme Recurrence

Abstract

This study aims to develop and evaluate deep neural network (DNN) models for accurately predicting the recurrence risk of glioblastoma multiforme (GBM) to enhance individualized treatment strategies and improve patient outcomes. This study implemented DNN architectures optimized using a hybrid differential evolution neural network (HDE-NN) framework to forecast GBM recurrence risk, particularly in patients at advanced disease stages. The models were trained and validated on a multimodal dataset comprising genomic profiles, imaging-derived metrics, and longitudinal clinical records from 780 GBM patients. Data were sourced from The Cancer Genome Atlas (TCGA) and institutional repositories. Performance was benchmarked against conventional machine learning models, including support vector machines (SVM), random forests (RF), and standard DNNs. The models were implemented in Python. The proposed HDE-optimized DNN achieved an accuracy of 94%, precision of 92%, recall of 90%, F1 score of 91%, and an AUC-ROC of 0.96. These metrics significantly outperformed baseline models, with improvements of 6–12% across evaluation criteria. Confidence intervals (95%) were computed via tenfold cross-validation, confirming statistical robustness. This research introduces a high-performance and generalizable deep learning framework for GBM recurrence prediction. By incorporating multi-source clinical and genomic data, the model demonstrates superior predictive capacity over traditional methods. These findings support the integration of AI-driven tools into GBM care workflows to improve prognosis assessment and personalize therapeutic interventions.