A Total Variation Regularized Framework for Epilepsy-Related MRI Image Segmentation

2025-10-06

AI Summary

Focal cortical dysplasia (FCD) poses significant challenges in 3D multimodal MRI segmentation due to its small volume, low contrast, scarcity of expert annotations, and difficulty in preserving anatomical consistency. To address these issues, we propose a Transformer-enhanced 3D encoder-decoder architecture and introduce a novel composite loss function integrating Dice loss with an anisotropic total variation (TV) regularization term—explicitly enforcing spatial smoothness and anatomical plausibility of segmentations while eliminating reliance on post-processing. Evaluated on a public dataset of 85 FCD cases, our method achieves a 11.9% improvement in Dice coefficient, a 13.3% gain in precision, and a substantial 61.6% reduction in false-positive clusters, outperforming all baseline methods. This work delivers an interpretable, robust, and clinically deployable end-to-end solution for precise pre-surgical FCD localization in epilepsy.

Technology Category

Computer Vision: 3D Computer VisionMachine Learning: Multimodal LearningConstraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAGWeb Mining and Content Analysis: Robustness and generalizability of Web mining methodsSecurity and Privacy: Data transparency and provenance

Abstract

Focal Cortical Dysplasia (FCD) is a primary cause of drug-resistant epilepsy and is difficult to detect in brain {magnetic resonance imaging} (MRI) due to the subtle and small-scale nature of its lesions. Accurate segmentation of FCD regions in 3D multimodal brain MRI images is essential for effective surgical planning and treatment. However, this task remains highly challenging due to the limited availability of annotated FCD datasets, the extremely small size and weak contrast of FCD lesions, the complexity of handling 3D multimodal inputs, and the need for output smoothness and anatomical consistency, which is often not addressed by standard voxel-wise loss functions. This paper presents a new framework for segmenting FCD regions in 3D brain MRI images. We adopt state-of-the-art transformer-enhanced encoder-decoder architecture and introduce a novel loss function combining Dice loss with an anisotropic {Total Variation} (TV) term. This integration encourages spatial smoothness and reduces false positive clusters without relying on post-processing. The framework is evaluated on a public FCD dataset with 85 epilepsy patients and demonstrates superior segmentation accuracy and consistency compared to standard loss formulations. The model with the proposed TV loss shows an 11.9% improvement on the Dice coefficient and 13.3% higher precision over the baseline model. Moreover, the number of false positive clusters is reduced by 61.6%

Problem

• Segmenting FCD lesions in 3D brain MRI images
• Addressing limited annotated data and subtle lesion characteristics
• Ensuring spatial smoothness and anatomical consistency in segmentation

Innovation

• Transformer-enhanced encoder-decoder architecture for segmentation
• Anisotropic Total Variation loss ensuring spatial smoothness
• Combined Dice and TV loss eliminating post-processing needs