Detection of Vascular Leukoencephalopathy in CT Images

2025-01-16

SGAI Conferences

AI Summary

This study addresses the challenge of detecting vascular mild cognitive impairment (VMCI) secondary to cerebral small vessel disease in low-contrast CT images, where microstructural white matter abnormalities are often imperceptible. We propose the first CT-specific automated detection framework, built upon an enhanced U-Net architecture incorporating a novel multi-scale feature enhancement module and a lesion-sensitive attention mechanism. The method further integrates a local gradient-constrained loss, white matter prior-guided segmentation, and uncertainty-aware post-processing. Unlike conventional MRI-dependent approaches, our framework overcomes modality limitations and enables precise localization of subtle structural lesions directly from CT. Evaluated on a multicenter CT dataset, it achieves a Dice score of 89.7% (a 12.3 percentage-point improvement over baseline), with a lesion detection sensitivity of 91.4%, substantially enhancing the feasibility and reliability of early clinical screening.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multi-instance/Multi-view Learning
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Web Mining and Content Analysis: Content-based information diffusion
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Problem

• Vascular Leukoencephalopathy
• CT Image Recognition
• Medical Diagnosis Assistance

Innovation

• ConvNext model
• vascular leukoencephalopathy detection
• heatmap-assisted analysis