Enhancing AI-based decision support system with automatic brain tumor segmentation for EGFR mutation classification.
Authors
Affiliations (6)
Affiliations (6)
- College of Engineering, Computer Engineering Department, Koç University, Istanbul, Türkiye.
- Department of Electrical Engineering and Computer Sciences, University of California, Berkeley, CA, USA.
- Faculty of Science and Letters, Department of Statistics, Mimar Sinan Fine Arts University, Silahsör Cad. No. 71, Bomonti/Sisli, 34380, Istanbul, Türkiye. [email protected].
- Department of Neurosurgery, Istanbul Kanuni Sultan Süleyman Health Application and Research Center, Health Sciences University, Istanbul, Türkiye.
- Faculty of Medicine, Department of Medical Biology, Bandırma Onyedi Eylül University, Balıkesir, Türkiye.
- Institute for Data Science, University of California, Berkeley, CA, USA.
Abstract
Glioblastoma (GBM) carries poor prognosis; epidermal-growth-factor-receptor (EGFR) mutations further shorten survival. We propose a fully automated MRI-based decision-support system (DSS) that segments GBM and classifies EGFR status, reducing reliance on invasive biopsy. The segmentation module (UNet SI) fuses multiresolution, entropy-ranked shearlet features with CNN features, preserving fine detail through identity long-skip connections, to yield a Lightweight 1.9 M-parameter network. Tumour masks are fed to an Inception ResNet-v2 classifier via a 512-D bottleneck. The pipeline was five-fold cross-validated on 98 contrast-enhanced T1-weighted scans (Memorial Hospital; Ethics 24.12.2021/008) and externally validated on BraTS 2019. On the Memorial cohort UNet SI achieved Dice 0.873, Jaccard 0.853, SSIM 0.992, HD95 24.19 mm. EGFR classification reached Accuracy 0.960, Precision 1.000, Recall 0.871, AUC 0.94, surpassing published state-of-the-art results. Inference time is ≤ 0.18 s per slice on a 4 GB GPU. By combining shearlet-enhanced segmentation with streamlined classification, the DSS delivers superior EGFR prediction and is suitable for integration into routine clinical workflows.