AI and Cancer Research ── Brain ── 2026-11-08
Brain tumours: treatment, trials and AI
- Abstract
Preoperative differentiation of glioma from meningioma on MRI is important for treatment planning but depends on expert interpretation. A 3D ResNet-34 with CBAM attention was trained on four-modality MRI from 2251 BraTS 2023 patients (1801 for four-fold cross-validation; 450 held out for testing) and achieved 96% accuracy, an AUC of 0.993, and an expected calibration error of 0.0165 on the test set. For layer-wise explanation, Grad-CAM and SmoothGrad were applied at the four residual stages, and a reconstruction decoder was added to obtain full-resolution, decoder-mediated attribution at the 4 × 4 × 4 bottleneck, where standard Grad-CAM is too coarse. Four quantitative metrics gave complementary results: the decoder-assisted attribution had the highest pointing-game accuracy (80.2%) and the lowest deletion AUC, whereas Layer 1 had the highest insertion AUC and Layer 2 the highest mean faithfulness delta, with no significant faithfulness difference among Layers 1, 2, and 4. Glioma was localized by most layers, while meningioma was localized mainly through the decoder-assisted pathway. A preliminary clinical review and a normalization analysis are also reported. The decoder offers an annotation-free way to obtain full-resolution attribution from deep layers of volumetric models.
Journal IF-equivalent: 6.6 (OpenAlex 2-year mean citedness, value as of 2026-11-06, retrieved 2026-11-07; not the official Clarivate IF)Reference: Pagonis N, Plomaritis K, Kogias E, Foroglou N, Chouvarda I. From Classification to Clinical Trust: 3D Deep Learning and Layer-Wise Explainability for Brain Tumor MRI. Sensors. 2026 Nov 5;26(21):6993. doi:10.3390/s26216993.Checked: Abstract only - Abstract
BACKGROUND: Accurate classification of Isocitrate Dehydrogenase (IDH) mutation status is crucial for the diagnosis and treatment of adult-type diffuse gliomas as defined by the 2021 World Health Organization (WHO) Classification. However, conventional histological and molecular diagnostics are time-consuming, labor-intensive, costly, and often limited in accessibility. Recent advances in artificial intelligence have enabled automated and unbiased glioma classification directly from histopathological images, but challenges remain in capturing complex pathological and molecular features effectively. METHODS: In this study, we propose FOCUS (Feature Optimization and Cascaded Unified Screening), an integrated framework for IDH mutation status classification in gliomas using glioma pathology slides. The histopathological images are divided into patches, from which a Vision Transformer (ViT) extracts interpretable features related to nuclear morphology and tumor progression-associated protein expression. A Patch-based Graph Convolutional Network (PatchGCN) further aggregates spatial context to capture both local morphology and global tumor heterogeneity. A cascaded feature selection strategy is employed to refine biomarker identification, followed by a machine learning classifier for final classification. RESULTS: Using the proposed framework, the Support Vector Machine (SVM) demonstrated promising performance on a dataset of 73 patients (accuracy = 0.902, AUC = 0.909). CONCLUSIONS: These results highlight the potential of the FOCUS framework as a robust and interpretable tool for IDH mutation status classification, providing a computational approach to distinguish IDH-mutant (astrocytoma and oligodendroglioma) from IDH-wildtype (glioblastoma) tumors.
Journal IF-equivalent: 4.2 (OpenAlex 2-year mean citedness, value as of 2026-11-06, retrieved 2026-11-07; not the official Clarivate IF)Reference: Su Y, Li F, Yang H, Bai J, Lu J. FOCUS: a feature optimization and cascaded unified screening strategy for computer-aided pathology-based glioma molecular subtype classification. Diagn Pathol. 2026 Aug 8;21(1):84. doi:10.1186/s13000-026-01836-1.Checked: Abstract only