Development and validation of a reliable method for automated measurements of psoas muscle volume in CT scans using deep learning-based segmentation: a cross-sectional study

The study aimed to create an automated method for measuring psoas muscle volume using CT scans to enhance sarcopenia research. Utilizing a data set of 520 participants, the researchers developed a psoas muscle segmentation model through a three-step deep learning process based on the nnU-Net method. This model was implemented in AVIEW for automated segmentation, focusing on the relevant region of interest to improve efficiency. The method's accuracy was evaluated using the Dice similarity coefficient, achieving an average Dice score of 0.927 ± 0.019, comparable to manual segmentation. The automated system significantly reduced measurement time to an average of 2 minutes 20 seconds, 48 times faster than the manual method. The study concludes that the automated method provides consistent, unbiased psoas muscle volume measurements across various CT images, making it a valuable tool for sarcopenia research.