AI-Powered Photoacoustic/Ultrasound System Quantifies Arthritis Activity in Fingers

Fully automated 3D photoacoustic/ultrasound system with deep learning (Dice 0.77, IoU 0.64) provided six imaging biomarkers; combined biomarkers correlated with disease activity (R²=0.52) in 43 rheumatoid arthritis finger joints.

• Pilot longitudinal study of 43 finger joints from inflammatory arthritis patients; manual segmentation served as reference standard.
• Deep learning model (DAF3D) achieved Dice score of 0.77±0.03 and IoU of 0.64±0.03 for joint space and synovium segmentation.
• Hyperemia on photoacoustic imaging showed the strongest association with clinical scores (R²=0.41); all six biomarkers combined yielded R²=0.52. Limitation: small sample, single-center, no external validation.

RadPigeon summaries are original and for information only. They are not clinical advice.