PHILADELPHIA—Magnetic resonance imaging (MRI) is considered crucial for in vivo detection and characterization of white matter lesions (WMLs) in multiple sclerosis.
Despite research on WMLs for at least the last 20 years using MRI, a report in the Journal of Neuroimaging noted that automated segmentation remains challenging.
“Although the majority of statistical techniques for the automated segmentation of WMLs are based on single imaging modalities, recent advances have used multimodal techniques for identifying WMLs,” wrote study authors led by University of Pennsylvania Perelman School of Medicine researchers and including participation from the Henry M. Jackson Foundation for the Advancement of Military Medicine. “Complementary modalities emphasize different tissue properties, which help identify interrelated features of lesions.
The study team proposed the Method for Inter-Modal Segmentation Analysis (MIMoSA), a fully automatic lesion segmentation algorithm that utilizes novel covariance features from intermodal coupling regression in addition to mean structure to model the probability lesion is contained in each voxel.
Using two datasets, the MIMoSA algorithm was validated by comparison with both expert manual and other automated segmentation methods. In the first at Johns Hopkins Hospital, in two datasets, bootstrap cross-validation was used to compare the performance of MIMoSA against OASIS and LesionTOADS, two popular automatic segmentation approaches, in 98 subjects. Publicly-available data from a segmentation challenge were used for performance benchmarking as a secondary validation, the authors stated.
Results of the Johns Hopkins study indicated that MIMoSA yielded average Sørensen-Dice coefficient (DSC) of 0.57 and partial AUC of 0.68 calculated with false positive rates up to 1%. In addition to being superior to performance using OASIS and LesionTOADS, the proposed method also performed competitively in the segmentation challenge dataset, researchers wrote.
“MIMoSA resulted in statistically significant improvements in lesion segmentation performance compared with LesionTOADS and OASIS and performed competitively in an additional validation study,” study authors concluded.
1. Valcarcel AM, Linn KA, Vandekar SN, Satterthwaite TD, Muschelli J, Calabresi PA, Pham DL, Martin ML, Shinohara RT. MIMoSA: An Automated Method for Intermodal Segmentation Analysis of Multiple Sclerosis Brain Lesions. J Neuroimaging. 2018 Jul;28(4):389-398. doi: 10.1111/jon.12506. Epub 2018 Mar 8. PubMed PMID: 29516669; PubMed Central PMCID: PMC6030441.