Automatic segmentation of gray matter multiple sclerosis lesions on DIR images

Multiple Sclerosis (MS) is a chronic inflammatorydemyelinating disease that affects both white and gray matter (GM). GM lesions have been demonstrated to play a major role in the physical and cognitive disability and in the disease progression. The diagnosis and monitoring of the disease is mainly based on magnetic resonance imaging (MRI). Lesions identification needs visual detection performed by experienced graders, a process that is always time consuming, error prone and operator dependent. We present a technique to automatically estimate GM lesion load from double inversion recovery (DIR) MRI sequences. We tested the proposed algorithm onDIR sequences acquired from 50 MS patients. Regions corresponding to probable GM lesions were manually labeled to provide a reference. The resulting automatic lesion load estimate provides a correlation of 98.5% with manual lesion number, and of 99.3% with manual lesion volume.