Focal Cortical Dysplasia Type II Detection Using Cross Modality Transfer Learning and Grad-CAM in 3D-CNNs for MRI Analysis

TL;DR

This study tackles the challenge of detecting FCD type II on MRI, a subtle and clinically critical condition in drug-resistant epilepsy. It employs 3D-CNNs based on ResNet architectures and investigates cross-modality transfer learning from segmentation tasks, coupled with Grad-CAM explainability and a novel Heat-Score metric to assess localization of epileptogenic regions. Results show that transfer learning, especially from 23 segmentation tasks, substantially boosts classification accuracy (up to $0.803$ on FLAIR with ResNet-50) and the Heat-Score (up to $2.940$), with significant improvements over training from scratch ($p<0.05$). The Heat-Score provides a quantified link between model attention and clinically relevant regions, suggesting enhanced interpretability and potential clinical utility for MRI-negative cases, while highlighting the need for multicenter validation. Overall, cross-modality TL and XAI approaches can meaningfully improve AI-assisted FCD-II diagnostics in MRI analysis, bridging predictive performance with clinical insight, aided by the Heat-Score as a practical interpretability metric.

Abstract

Focal cortical dysplasia (FCD) type II is a major cause of drug-resistant epilepsy, often curable only by surgery. Despite its clinical importance, the diagnosis of FCD is very difficult in MRI because of subtle abnormalities, leading to misdiagnosis. This study investigates the use of 3D convolutional neural networks (3D-CNNs) for FCD detection, using a dataset of 170 subjects (85 FCD patients and 85 controls) composed of T1-weighted and FLAIR MRI scans. In particular, it investigates the benefits obtained from cross-modality transfer learning and explainable artificial intelligence (XAI) techniques, in particular Gradient-weighted Class Activation Mapping (Grad-CAM). ResNet architectures (ResNet-18, -34, and -50) were implemented, employing transfer learning strategies that used pre-trained weights from segmentation tasks. Results indicate that transfer learning significantly enhances classification accuracy (up to 80.3%) and interpretability, as measured by a novel Heat-Score metric, which evaluates the model's focus on clinically relevant regions. Improvements in the Heat-Score metric underscore the model's seizure zone localization capabilities, bringing AI predictions and clinical insights closer together. These results highlight the importance of transfer learning, including cross-modality, and XAI in advancing AI-based medical diagnostics, especially for difficult-to-diagnose pathologies such as FCD.

Figures (2)

• Figure 1: Visual representation of 3D-ResNet used for our classification task.
• Figure 2: An example illustrating a correct classification with accurate localization (a), a correct classification despite incorrect localization (b), a near-correct localization (c), and an overly vague localization (d) of FCD-II. The segmentation contour is outlined in black.