Development of an artificial intelligence powered software for automated analysis of skeletal muscle ultrasonography

Muscle ultrasound has high utility in clinical practice and research; however, the main challenges are the training and time required for manual analysis to achieve objective quantification of muscle size and quality. We aimed to develop and validate a software tool powered by artificial intelligence (AI) by measuring its consistency and comparability of expert manual analysis quantifying lower limb muscle ultrasound images. Quadriceps complex (QC) and tibialis anterior (TA) muscle images of healthy, intensive care unit, and/or lung cancer participants were captured with portable devices. Manual analyses of muscle size and quality were performed by experienced physiotherapists taking approximately 24 h to analyze all 180 images, while automated analyses were performed using a custom-built deep-learning model (MyoVision-US), taking 247 s (saving time = 99.8%). Consistency between the manual and automated analyses was good to excellent for all QC (ICC = 0.85–0.99) and TA (ICC = 0.93–0.99) measurements, even for critically ill (ICC = 0.91–0.98) and lung cancer (ICC = 0.85–0.99) images. The comparability of MyoVision-US was moderate to strong for QC (adj. R² = 0.56–0.94) and TA parameters (adj. R² = 0.81–0.97). The application of AI automating lower limb muscle ultrasound analyses showed excellent consistency and strong comparability compared with human analysis across healthy, acute, and chronic population.