MobileSpine: A Mobile-based 3D Reconstruction System for Remote Scoliosis Assessment

We present MobileSpine, an innovative system that enables remote scoliosis assessment through ordinary smartphone RGB image acquisition combined with state of the art three dimensional reconstruction. Conventional diagnostic pathways rely on two dimensional radiography, providing flattened anatomical projections and exposing patients to cumulative ionising radiation, while commercially available optical three dimensional scanners demand specialised hardware, scheduled clinical visits, and substantial financial investment, thereby restricting longitudinal surveillance. Our approach circumvents these barriers by implementing a lightweight multi view image generation protocol that guides users to capture a representation of the dorsal torso under natural illumination. Camera poses are estimated with a fine tuned COLMAP pipeline, after which photogrammetric depth cues are fused through an efficient meshing algorithm that yields watertight surface models. Quantitative evaluation on a heterogeneous cohort comprising adolescents and adults reveals a mean surface reconstruction error of 0.01 mm, a level of geometric fidelity that meets the stringent tolerances required for clinical decision making. Key spinal parameters—including Cobb angle, vertebral rotation, and sagittal balance—are automatically extracted from the reconstructed mesh and exhibit 99 percent agreement with annotations, underscoring diagnostic robustness. Consequently, MobileSpine lays the technical foundation for orthopaedic screening programmes, continuous at home monitoring during growth spurts, and data driven intervention planning, ultimately promoting equitable access to high quality musculoskeletal care.