Automatic detection of new lesions in patients with multiple sclerosis using the T1-w/T2-w magnetic resonance ratio

Pérez Sánchez, Laura

Automatic detection of new lesions in patients with multiple sclerosis using the T1-w/T2-w magnetic resonance ratio

Citation

Pérez Sánchez, L. Automatic detection of new lesions in patients with multiple sclerosis using the T1-w/T2-w magnetic resonance ratio. 2021. handle: http://hdl.handle.net/10230/48245

This citation was generated automatically.

Abstract

Multiple sclerosis is an autoimmune condition in which the immune system mistakenly attacks the myelin sheath, causing that messages travelling along the nerves become slowed or disrupted. Currently, the severity of the disease progression is given by the formation of new lesions. The aim of the project was to demonstrate that by using the ratio of T1-weighted to T2-weighted images, additional lesions can be measured, when compared to Fluid-attenuated inversion recovery (FLAIR) images. For this purpose, follow-up and baseline images were used in order to obtain subtraction maps. Then, the subtraction images were redefined using deformation field and local tissue information to detect disease activity. In this study, a database of 100 patients was employed, of which 34 present new lesions. The evaluation consisted of comparing the results with the ground truth of new lesions already established from the subtraction of FLAIR images and calculating metrics for detection. The developed method was influenced by registration inconsistencies and image intensity differences. Therefore, it requires a post-processing of the results to minimize false positives, leading to an increased difficulty when locating the smallest ones. Still, new lesions were detected employing this method that were not identified when using the subtraction of FLAIR images. In a 13.63% of the patients classified by the ground truth as if they did not present new lesions, utilizing the T1-w/T2-w magnetic resonance ratio, new lesions that went unnoticed, are now detected. Thereby, this ap- proach could be used in a real clinical environment to help on the detection of new lesions.