Unique MS Lesion Identification from MRI

TL;DR

The paper addresses the challenge of uniquely identifying MS white matter lesions (WMLs) and deriving meaningful per-lesion statistics that better relate to disability. It introduces a Hessian-based center detection on a lesion probability map to locate potential lesion seeds, followed by a Random Walker-based growth to assign tissue to distinct lesions and estimate their volumes. Through synthetic data and real MICCAI MS data, the method demonstrates improved lesion counting, robust separation of confluent lesions, and accurate total WML volume compared to baseline methods and prior approaches. This enables more informative biometrics for MS progression studies and could enhance longitudinal tracking and clinical decision-making.

Abstract

Unique identification of multiple sclerosis (MS) white matter lesions (WMLs) is important to help characterize MS progression. WMLs are routinely identified from magnetic resonance images (MRIs) but the resultant total lesion load does not correlate well with EDSS; whereas mean unique lesion volume has been shown to correlate with EDSS. Our approach builds on prior work by incorporating Hessian matrix computation from lesion probability maps before using the random walker algorithm to estimate the volume of each unique lesion. Synthetic images demonstrate our ability to accurately count the number of lesions present. The takeaways, are: 1) that our method correctly identifies all lesions including many that are missed by previous methods; 2) we can better separate confluent lesions; and 3) we can accurately capture the total volume of WMLs in a given probability map. This work will allow new more meaningful statistics to be computed from WMLs in brain MRIs

Figures (5)

• Figure 1: (a) Axial FLAIR image and its corresponding lesion probability map zhang2024isbi for an MS subject. (b) The same images as (a) with the centers of the unique lesions indicated by a color overlay. (c) The unique lesions after our growth step.
• Figure 2: From left to right: A slice of a synthetic lesion probability map without lesion confluence displayed with the jet colormap (Synthetic Image), the connected components from a binary threshold approach (Binary Threshold), the output of the Dworkin Method, and finally, our approach.
• Figure 3: From left to right: A slice of a synthetic image with six confluent lesions which are manually identified (Manual) and the results of identification via binary threshold (Binary), the Dworking Method, and our proposed method. Binary thresholding identified one lesion, Dworkin identified four lesions, whereas our method identified all six lesion centers.
• Figure 4: (a) Lesion counts from Binary threshold with connected components, the Dworkin method, and our method. (b) Total volume of captured lesions from binary thresholding, the Dworkin method, and after the growth phase of our method.
• Figure 5: (a) On the left is the sagittal MR image displayed with the lesion probability map above it. The middle images show the Dworkin-identified lesions overlaid, and the right images display the results of our method. (b) Same display order as (a), but on the axial MR image of another subject. The red-circled areas highlight lesions that the Dworkin method missed, while the green-circled areas show lesions that our method identified.