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

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.

Low contrast detection and super-resolution in CT images: evaluation of a novel approach based on Centroidal Voronoi Tessellation

In this work, image analysis techniques used in astrophysics to detect low-contrast signals have been adapted in the processing of Computed Tomography (CT) images, combining Centroidal Voronoi Tessellation (CVT) and machine learning techniques. Several CT acquisitions were performed using a phantom containing cylindrical inserts of different diameters producing objects with different contrasts respect to background. The images of the phantom, tilted by a known angle with respect to the tomograph axis (to mimic the casual orientation of a clinical lesion), were acquired at various radiation doses (CTDIvol) and at different slice's thicknesses. The success in detecting the signal in the single image (slice) was always greater than 60%. The axis of each insert has always been correctly identified. A super-resolution 2D image was then generated by projecting the individual slices of the scan along this axis, thus increasing the CNR of the object scanned as a whole. CVT holds great promise for future use in medical imaging, for the identification of low-contrast lesions in homogeneous organs, such as the liver.