Progressive changes in lower extremity skeletal muscle density were detected using automated imaging analysis in patients with PAD. In a study presented at the American Heart Association (AHA) 2025 Scientific Sessions, deep learning–guided CT analysis quantified regional calf muscle density decline over an 18-month follow-up period.

This prospective observational study enrolled 89 patients with PAD who underwent non-contrast lower extremity CT imaging. Calf muscle groups, including the gastrocnemius, soleus, and tibialis anterior, were manually segmented from axial images of the symptomatic limb. These segmentations were used to train a neural network–based deep learning model. The dataset was divided into training and testing sets using an 80/20 split. A subset of 45 patients underwent repeat imaging at 18 months to evaluate longitudinal changes.

The deep learning model demonstrated high agreement with manual segmentation. Dice coefficients reached 0.90 ± 0.02 for the gastrocnemius and soleus muscles and 0.89 ± 0.02 for the tibialis anterior muscle. Serial analysis identified significant reductions in muscle density across all three calf muscle groups at 18-month follow-up compared with baseline measurements, with all comparisons reaching statistical significance (P < 0.05).

These findings show that deep learning–based CT analysis enables efficient regional evaluation of skeletal muscle density in PAD. Automated segmentation supports longitudinal monitoring of muscle deterioration associated with disease progression and functional impairment.