Motivation: Manual measures of spinal lesion morphometry from MRI scans correlate with neurological prognosis in spinal cord injury (SCI) patients but are prone to intra- and inter-rater variability. Goal(s): Develop a software solution that automates the measurements of lesion morphometry. Approach: We developed a deep learning model that segments the spinal cord and intramedullary lesions, and that computes lesion length and width on the midsagittal slice. This method was compared against manual measurements in an SCI cohort. Results: The automatic approach showed good agreement with manual measurements. The method is open-source and will be released as part of the Spinal Cord Toolbox (v6.5+). Impact: Automatic computation of lesion morphometry can replace manual measurements, thus facilitating large multi-center studies in spinal cord injury patients by reducing intra- and inter-expert variability and saving time.
Valošek et al. (Tue,) studied this question.