LST-AI

LST-AI is an open-source, AI-driven extension of the LST tool for segmenting brain white matter lesions in multiple sclerosis (MS). It utilizes a deep learning ensemble of 3D U-Nets trained on 491 MS image pairs, improving lesion segmentation with a composite loss function. LST-AI outperformed existing models like LST and nnUNet in empirical tests, particularly excelling in the MSSEG-1 challenge dataset, and is available for use across multiple platforms, supporting its adoption in both research and clinical settings.