AI-assisted 3D screening estimates 3.8% scoliosis prevalence in Shanghai adolescents

An AI-assisted 3D back-scanning workflow screened 5,026 students and found an inverse probability weighted adolescent idiopathic scoliosis prevalence of 3.8%. The system achieved 83.2% sensitivity and 99.5% specificity, with modifiable factors like sitting posture and unilateral…

• Cross-sectional study using multistage stratified cluster sampling in 5,026 students (ages 9–15) with AI structured-light screening; standing full-spine radiography served as reference standard.
• IPW-adjusted diagnostic performance: sensitivity 0.832, specificity 0.995, PPV 0.865, NPV 0.993, overall accuracy 0.989.
• Modifiable risk associations included improper sitting posture in girls and frequent one-sided exertion sports; higher BMI and satisfactory sleep were protective. Limitation: single-city design may limit generalizability.

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