Supervised segmentation framework for evaluation of diffusion tensor imaging indices in skeletal muscle

Abstract

Diffusion tensor imaging (DTI) is becoming a relevant diagnostic tool to understand muscle disease and map muscle recovery processes following physical activity or after injury. Segmenting all the individual leg muscles, necessary for quantification, is still a time-consuming manual process. The purpose of this study was to evaluate the impact of a supervised semi-automatic segmentation pipeline on the quantification of DTI indices in individual upper leg muscles. Longitudinally acquired MRI datasets (baseline, post-marathon and follow-up) of the upper legs of 11 subjects were used in this study. MR datasets consisted of a DTI and Dixon acquisition. Semi-automatic segmentations for the upper leg muscles were performed using a transversal propagation approach developed by Ogier et al on the out-of-phase Dixon images at baseline. These segmentations were longitudinally propagated for the post-marathon and follow-up time points. Manual segmentations were performed on the water image of the Dixon for each of the time points. Dice similarity coefficients (DSCs) were calculated to compare the manual and semi-automatic segmentations. Bland-Altman and regression analyses were performed, to evaluate the impact of the two segmentation methods on mean diffusivity (MD), fractional anisotropy (FA) and the third eigenvalue (λ₃ ). The average DSC for all analyzed muscles over all time points was 0.92 ± 0.01, ranging between 0.48 and 0.99. Bland-Altman analysis showed that the 95% limits of agreement for MD, FA and λ₃ ranged between 0.5% and 3.0% for the transversal propagation and between 0.7% and 3.0% for the longitudinal propagations. Similarly, regression analysis showed good correlation for MD, FA and λ₃ (r = 0.99, p < 60; 0.0001). In conclusion, the supervised semi-automatic segmentation framework successfully quantified DTI indices in the upper-leg muscles compared with manual segmentation while only requiring manual input of 30% of the slices, resulting in a threefold reduction in segmentation time.

Keywords: applications; diffusion tensor imaging (DTI); methods and engineering; muscle; musculoskeletal; post-acquisition processing; quantitation.

FIGURE 1

Schematic of the study set‐up. DTI measurements were performed at 3 time points, i.e. at baseline, 24‐48 hrs. post‐marathon and at a 2 weeks follow‐up. Manual segmentations were performed for 10 muscles in both upper legs at all 3 time points. At baseline, a supervised semi‐automatic segmentation was performed based on a selected set of manually segmented slices (transversal propagation). For the post‐marathon and the follo‐up time points the segmentation was automatically propagated without further manual input (longitudinal propagation)

FIGURE 2

Anterior view of upper leg muscle segmentations, where knees are at the bottom and hips at the top of the image. (A) Transversal slice, approximately halfway between knee and hip, showing segmentations of the 10 muscles considered in this study in both legs, for example, 21 slices for the BFLH muscle in this subject. (B) Longitudinal views of the upper legs with an example of input segmentations at baseline. Adjacent slices were selected as input when muscle shape changed. (C) Propagations resulting from the transversal propagation step. (D) The full volumes obtained with the transversal propagation step

FIGURE 3

Comparison of the manual and semi‐automatic segmentations of the mean values of λ ₃, MD and FA for the BFLH, RF and VL muscles of a representative subject at baseline, post‐marathon and follow‐up time points

FIGURE 4

DSCs for the supervised segmentations compared with the manual segmentations for the baseline (4A in blue), post‐marathon (4B in red) and follow‐up (4C in yellow) time points for all the segmented muscles. The numbers correspond to the muscles listed in Figure 2. Each dot reflects an individual subject and the group mean and standard deviation are shown in black

FIGURE 5

Percentage of segmented slices used for transversal propagation at baseline for the individual muscles. The BFSH of the right leg (blue) required the most manual segmented slices; the BFLH of the left leg (red) the fewest. Each point represents the mean and the range of the segmentation acceleration. The dotted line represents the average acceleration time

FIGURE 6

Linear regression analysis (top) and Bland‐Altman plot (bottom) at baseline of diffusion indices λ₃ (10⁻³ mm²/s), MD (10⁻³ mm²/s) and FA (—) comparing manual segmentation with those resulting from the transversal propagations

FIGURE 7

Linear regression analysis (top) and Bland‐Altman plot (bottom) post‐marathon of diffusion indices λ₃ (10⁻³ mm²/s), MD (10⁻³ mm²/s) and FA (—) comparing manual segmentation with those resulting from the longitudinal propagations

FIGURE 8

Linear regression analysis (top) and Bland‐Altman plot (bottom) at follow‐up of diffusion indices λ₃ (10⁻³ mm²/s), MD (10⁻³ mm²/s) and FA (—) comparing manual segmentation with those resulting from two longitudinal propagation steps