Smartphone-based Structure-from-Motion for the remote assessment of trunk rotation in spine deformity.

BACKGROUND

In scoliosis care, telehealth typically focuses on virtual assessments of posture and flexibility as surrogates for physical examinations, with quantitative measurements limited to radiography. This study evaluates the feasibility of using smartphone-based structure-from-motion (SfM) for patients to capture reliable 3D representations of their torsos using accessible, low-cost tools and enable remote surface measurement of Axial Trunk Rotation (ATR) without physical scoliometer use. We assess the accuracy and reliability of ATR measurements derived from SfM models, identify recommended image capture conditions, and examine the influence of reconstruction parameters such as tie point number, depth map quality, manual background masking, and scaling.

METHODS

Twenty adolescents with idiopathic scoliosis were recruited from a single hospital and 3D surface scanning (3DSS) and SfM datasets were acquired during standard appointments. A structured protocol for image capture and reconstruction was developed. ATR measurements from SfM were compared with clinical (analog scoliometer) and technical (3DSS) gold standards using correlation and Bland-Altman plots. ATR from SfM models scaled in 2 software packages was compared with 3DSS. Intra and inter-user variability in manual image masking was assessed on a subset of 5 patients captured under varying background conditions.

RESULTS

A total of 60 to 80 photographs captured in a structured manner provided the best balance between time, patient fatigue, and reconstruction quality. Optimal reconstruction occurred with 15,000 tie points and medium-quality depth maps. ATR measured on SfM models showed strong positive correlation (>0.85, p < .001) with both the analog scoliometer and 3DSS, with proportional biases of 1.13° and 1.33° (≥90% points within agreement limits). Scaling comparisons showed mean biases of 0.78° (Agisoft Metashape) and 1.13° (Geomagic Wrap). Manual masking produced ATR variability consistently <2°.

CONCLUSIONS

Smartphone-based SfM offers a practical, low-cost alternative to 3DSS for evaluating trunk rotation. Its accuracy and reliability support its potential integration into telehealth workflows for scoliosis care.