Knowledge-Driven Automated Segmentation of Cortical Lesions on MR Brain Images in MS

Introduction: Multiple sclerosis affects both white matter (WM) and gray matter (GM). Conventional MR imaging is widely used to assess T2 hyperintense lesions in WM. These techniques are suboptimal in detecting cortical lesions. Recently, double inversion recovery (DIR) and phase sensitive inversion recovery (PSIR) have been shown to increase the confidence with which cortical lesions can be visualized with higher confidence than the conventional sequences [1]. However objective quantification of cortical lesions is still challenging because of their locations, relatively poor lesion-to-tissue contrast ratio, shape and size. Here we report a knowledge-driven segmentation with minimal human intervention for quantification of cortical lesions.