Brain lesion segmentation and atrophy measurement in MS patients: comparison between manual and automated methods (P3.351)

Objective: Compare accuracy and concordance between manual and automatic lesion and atrophy measurement on brain MRI of MS patients. Background: MRI is key in the diagnosis and management of MS patients. The presence of enlarging or new FLAIR/T2 hyperintense lesions and more recently brain atrophy, represent major endpoints in MS trials and guide therapeutic decisions. However, measurement of lesions and atrophy is still performed visually or in cumbersome-and time-consuming semi-manual processes. Recently, new automated methods have been developed to segment and quantify lesions and brain atrophy, simplifying the process. Nonetheless, little is known about the accuracy and concordance between manual and automated methods. We thus aimed to compare automated quantification to manual measurement in terms of accuracy and concordance. Design/Methods: Brain MRI data of 20 patients with relapsing-remitting MS were randomly selected from our MS Clinic cohort. Brain MRI were analyzed manually by 3 experienced neuroradiologists using Freesurfer and GCA atrophy scale. Automated analysis included algorithms implemented in Matlab (lesion growth algorithm [LGA], lesion prediction algorithm [LPA]) and Python. We then calculated inter-rater correlation and dice coefficients for all methods. Results: We found a significant correlation between lesion volume measured manually and automatized (intra-class correlation 0.985, P Conclusions: Automated lesion measurement in MS is reliable and shows good agreement with manual segmentation. Automated segmentation may be useful in clinical practice by reducing time of processing and inter-rater variability. Disclosure: Dr. Chaves has nothing to disclose. Dr. Varela has nothing to disclose. Dr. Serra has nothing to disclose. Dr. Osa Sanz has nothing to disclose. Dr. Stefanoff has nothing to disclose. Dr. Itzcovich has nothing to disclose. Dr. Correale has nothing to disclose. Dr. Fernandez Slezak has nothing to disclose. Dr. Farez has nothing to disclose.