Fig. 1.

Workflow from MRI segmentation to statistical parametric mapping analysis (SPM) for paraspinal muscle fat maps A) Location of center of rotation (CoR) used to center the radial ROIs was calculated and identified. B CoR coordinates were applied to matrix with paraspinal muscle segmentation. C FI% pixels within each ROI overlaying separate muscles were collected and transformed to columns (D) for each radial ROI per muscle. Each column entry represents the FI% of a single pixel. E For each radial ROI, FI pixels were averaged and the mean FI at every ROI was plotted for each muscle. F The radial fat distribution for separate muscles was graphed for all subjects (top) and then normalized based on individual radial muscle width so that the x-axis ranged from 0 to 100% of the radial width (middle). The mean distribution was then calculated for each subgroup. Then, regional differences in FI between groups were identified using SPM t-tests (bottom). Differences in FI were deemed significant between groups if the SPM{t} for a given region surpassed the critical threshold line (dotted line). As shown, multifidus FI was significantly larger in patients compared to controls in the 13.4–17.2% range of the distribution (p = 0.044) where the SPM{t} statistic surpassed the critical threshold (t* = 2.643)