Quantitative MRI outcome measures in CMT1A using automated lower limb muscle segmentation

Area of Science:

• Biomedical Imaging
• Neurology
• Artificial Intelligence

Background:

• Lower limb muscle fat fraction (FF) MRI detects Charcot-Marie-Tooth disease 1A (CMT1A) progression.
• Manual segmentation for FF MRI analysis is time-consuming.

Purpose of the Study:

• To assess the responsiveness, efficiency, and accuracy of AI-enabled automated segmentation for FF MRI in CMT1A patients.

Main Methods:

• Recruited 20 CMT1A patients and 7 controls for 12-month assessments.
• Used three-point-Dixon technique for thigh and calf FF MRI.
• Employed Musclesense AI for automated lower limb muscle segmentation with quality control (QC).

Main Results:

• Mean calf FF significantly increased in CMT1A patients over 12.5 months (1.15%±1.77%, p=0.016).
• Automated segmentation with QC showed a standardized response mean (SRM) of 0.65.
• Without QC, FF change was nearly identical (1.15%±1.68%, p=0.01) with a similar SRM of 0.69.

Conclusions:

• AI-enabled automated segmentation is efficient and accurate for longitudinal FF MRI studies in CMT.
• Minimal QC time is required, and corrections are infrequent.
• This technique offers a viable method for assessing CMT1A progression.