Large-scale analysis of iliopsoas muscle volumes in the UK Biobank

Abstract

Psoas muscle measurements are frequently used as markers of sarcopenia and predictors of health. Manually measured cross-sectional areas are most commonly used, but there is a lack of consistency regarding the position of the measurement and manual annotations are not practical for large population studies. We have developed a fully automated method to measure iliopsoas muscle volume (comprised of the psoas and iliacus muscles) using a convolutional neural network. Magnetic resonance images were obtained from the UK Biobank for 5000 participants, balanced for age, gender and BMI. Ninety manual annotations were available for model training and validation. The model showed excellent performance against out-of-sample data (average dice score coefficient of 0.9046 ± 0.0058 for six-fold cross-validation). Iliopsoas muscle volumes were successfully measured in all 5000 participants. Iliopsoas volume was greater in male compared with female subjects. There was a small but significant asymmetry between left and right iliopsoas muscle volumes. We also found that iliopsoas volume was significantly related to height, BMI and age, and that there was an acceleration in muscle volume decrease in men with age. Our method provides a robust technique for measuring iliopsoas muscle volume that can be applied to large cohorts.

Figure 1

Iliopsoas muscle manual annotations: (a) axial, (b) sagittal, and (c) coronal views, (d–f) showing of the segmentation (red) overlaid on the anatomical reference data, and (g) 3D rendering of manual segmentation.

Figure 2

Bland–Altman plot (a) of iliopsoas muscle volumes determined with CNN-based and manual segmentations ($n=90$), using a six-fold cross-validation experiment. Dotted lines represent the average bias ($-0.2$%) and the 95% limits of agreement. Overlays of the CNN-based and manual segmentations for two subjects (b,c), where the manual annotation is red, the CNN segmentation is green and the overlap is yellow.

Figure 3

CNN segmentations of the left and right iliopsoas muscles overlaid in purple (right) and blue (left) from a range of body types and iliopsoas muscle volumes: (a–c) small, (d–f) average, (g–i) large and (j–l) asymmetric. The top row for each subject displays the signal intensities without the segmentation result, the bottom row includes the iliopsoas muscle segmentations.

Figure 4

Difference in volume (ml) between the left and right iliopsoas muscles, separated by gender. Negative values indicate the right iliopsoas muscle is larger than the left.

Figure 5

Scatterplot of total iliopsoas muscle volume (ml) by height (cm), separated by gender.

Figure 6

Scatterplot of iliopsoas muscle index (${\text{ml/m}}^2$) by BMI (${\text{kg/m}}^2$), separated by gender.

Figure 7

Scatterplot of iliopsoas muscle index (${\text{ml/m}}^2$) by age at recruitment (years), separated by gender. The curves are fit to the data using a generalised additive model with cubic splines.