Machine Learning-Based Classification of Acute versus Chronic Multiple Sclerosis Lesions using Radiomic Features from Unenhanced Cross-Sectional Brain MRI (4121)

Objective: To build a machine learning/artificial intelligence-based (ML/AI) tool to classify acute (T1 gadolinium-enhancing [T1Gd+] or new T2 hyperintense lesions) versus chronic T2 hyperintense multiple sclerosis (MS) lesions using only cross-sectional T1- and T2-weighted brain MRI without gadolinium contrast information. Background: MS lesions have substantial pathological heterogeneity, but typically appear quite similar on conventional T1- and T2-weighted images. Acute lesions, reflecting recent inflammatory disease activity, are identified with Gd enhancement, which underestimates recent activity given the short half-life of blood-brain barrier disruption, or through detection of new T2 lesions, which requires longitudinal MRI. Design/Methods: Brain T1- and T2-weighted MRIs from ADVANCE (1,512 patients with relapsing-remitting MS, NCT00906399) and ASCEND (886 patients with secondary progressive MS, NCT01416181) trials were retrospectively analyzed. Each lesion was sampled using a ball of radius 4mm with a 3mm periphery. Topology-agnostic radiomic features were extracted from each region, leading to 372 features in total. Linear and non-linear robust feature selection was performed. A variety of classifiers were trained using the reduced feature space and the 5 best-performing classifiers were combined using the stacking ensemble learning strategy. ADVANCE was used for training/validation and ASCEND for testing. Results: Feature selection yielded 50 features, among which 17 are core-based and 33 are periphery-based. The classification framework reached 74.9% balanced accuracy, 74.3% precision, 76.3% sensitivity, 73.6% specificity and 82.8% AUC on the validation set. On the testing data set, these classification metrics are 76.1%, 77.9%, 72.8%, 79.4%, and 83.9% respectively. Conclusions: The proposed ML/AI method demonstrated relatively high and generalizable performance across trial populations at different stages of MS to discriminate acute versus chronic MS lesions using cross-sectional, unenhanced MRI data, thus reducing the need for Gd contrast and/or prior reference MRI. Future ML/AI applications may allow further classification and staging of MS lesion subtypes based on conventional MRI. Study Supported by: Biogen Disclosure: Bastien Caba has received personal compensation for serving as an employee of Therapanacea. Bastien Caba has received personal compensation in the range of $10,000-$49,999 for serving as a Research Engineer with Therapanacea. Dawei Liu has received personal compensation for serving as an employee of Biogen. Dawei Liu has received stock or an ownership interest from Biogen. Aurelien Lombard has received personal compensation for serving as an employee of Therapanacea. Natasha Novikov has nothing to disclose. Alexandre Cafaro has received personal compensation for serving as an employee of TheraPanacea. Daniel Bradley has nothing to disclose. Enzo Battistella has nothing to disclose. Elizabeth Fisher has received personal compensation for serving as an employee of Biogen. Elizabeth Fisher has received stock or an ownership interest from Biogen. Elizabeth Fisher has received intellectual property interests from a discovery or technology relating to health care. Nathalie Franchimont has received personal compensation for serving as an employee of Biogen. An immediate family member of Nathalie Franchimont has received personal compensation for serving as an employee of Biogen. An immediate family member of Nathalie Franchimont has received personal compensation for serving as an employee of Akcea. Nathalie Franchimont has received personal compensation in the range of $10,000-$49,999 for serving as an officer or member of the Board of Directors for OMass Therapeutics. Nathalie Franchimont has received stock or an ownership interest from Biogen. An immediate family member of Nathalie Franchimont has received stock or an ownership interest from Biogen. Dr. Gafson has nothing to disclose. Parya MomayyezSiahkal has nothing to disclose. Zahra Karimaghaloo has received personal compensation for serving as an employee of NeuroRx. Dr. Arnold has received personal compensation in the range of $500-$4,999 for serving as a Consultant for Alexion. Dr. Arnold has received personal compensation in the range of $10,000-$49,999 for serving as a Consultant for Biogen. Dr. Arnold has received personal compensation in the range of $500-$4,999 for serving as a Consultant for Celgene. Dr. Arnold has received personal compensation in the range of $500-$4,999 for serving as a Consultant for Frequency Therapeutics. Dr. Arnold has received personal compensation in the range of $500-$4,999 for serving as a Consultant for GENeuro. Dr. Arnold has received personal compensation in the range of $5,000-$9,999 for serving as a Consultant for Genentech. Dr. Arnold has received personal compensation in the range of $500-$4,999 for serving as a Consultant for Merck. Dr. Arnold has received personal compensation in the range of $10,000-$49,999 for serving as a Consultant for Novartis. Dr. Arnold has received personal compensation in the range of $500-$4,999 for serving as a Consultant for Receptos/Celgene. Dr. Arnold has received personal compensation in the range of $10,000-$49,999 for serving as a Consultant for Roche. Dr. Arnold has received personal compensation in the range of $5,000-$9,999 for serving as a Consultant for Sanofi. Dr. Arnold has received stock or an ownership interest from NeuroRx. The institution of Dr. Arnold has received research support from Novartis. The institution of Dr. Arnold has received research support from Immunotec. Colm Elliott has received personal compensation for serving as an employee of NeuroRx Research. Nikos Paragios has received personal compensation in the range of $500-$4,999 for serving on a Speakers Bureau for AstraZeneca. Nikos Paragios has received personal compensation in the range of $500-$4,999 for serving on a Speakers Bureau for Ipsen. Nikos Paragios has received personal compensation in the range of $500-$4,999 for serving as an officer or member of the Board of Directors for ArteDrone. Nikos Paragios has received personal compensation in the range of $10,000-$49,999 for serving as an Editor, Associate Editor, or Editorial Advisory Board Member for Elsevier. Nikos Paragios has received intellectual property interests from a discovery or technology relating to health care. Dr. Belachew has received personal compensation for serving as an employee of Biogen Inc. Dr. Belachew has received stock or an ownership interest from Biogen Inc.