Artificial Intelligence in Scoliosis Screening: Clinical evidence and implementation in adolescent idiopathic scoliosis

Abstract

Introduction: Artificial intelligence is capable of carrying out scoliosis measurements and assisting with referral decisions by means of radiographs, photographs, surface scans, depth sensors, or ultrasound. The value of this approach in screening will depend on whether these methods are able to identify curves that are clinically relevant without leading to an excessive number of referrals or providing false reassurance.
Methods: A search of PubMed from January 2000 to 17 September 2026 was carried out to find out about the use of artificial intelligence and machine learning in the screening, diagnosis, measurement of the Cobb angle, and prediction of progression of adolescent idiopathic scoliosis. Thirty clinical studies and reviews were chosen for a structured narrative evaluation and were grouped according to clinical task, imaging modality, and validation design.
Results: The method involving automated radiographic Cobb angle measurement is supported by the most substantial body of evidence; a meta-analysis found the pooled mean absolute error to be approximately 3 degrees. Radiation-free methods have demonstrated high sensitivity in certain groups, but their performance depends on the quality of acquisition and the patients selected. A school-based study involving 5,026 pupils found that an artificial intelligence-assisted structured-light method—which included radiographic verification—had high specificity and a high negative predictive value. Although ultrasound and progression models may be useful for monitoring, concerns still exist regarding missed curves, incomplete calibration, and the lack of external validation.
Conclusion: The existing evidence indicates that task-specific artificial intelligence tools should be used under supervision. It is necessary to carry out checks on image quality, provide the option of withholding results that are uncertain, have a clinician review them, and carry out selective radiographic confirmation. However, the evidence does not support replacing clinical assessment or radiographs when treatment decisions are based on the Cobb angle.