Validity and Usability of an AI-Based Self-Screening Mobile Application for Scoliosis

The primary goal of this project was to develop a simple AI-based scoliosis self-screening application that enables self-screening for scoliosis using a smartphone. A secondary objective was to evaluate the app's validity in comparison to the widely used baseline scoliometer and assess its usability using the System Usability Scale (SUS).

The application utilizes Canny edge detection to analyze the rib cage outline, excluding the spinous process, and detects the highest rib point on one side and the lowest point on the opposite side. The angle of the rib hump is then calculated. Fifteen healthy participants underwent the Adams test to measure rib hump angles, and these results were compared with those from a scoliometer. A Spearman correlation analysis was performed to assess the validity of the app’s measurements. Usability was evaluated using the SUS.

The app showed a moderate correlation with the scoliometer, with a Spearman correlation coefficient of 0.67, indicating fair validity. The usability assessment using SUS yielded a score of 72, suggesting the application falls within the "good" range, demonstrating its practicality and potential user acceptance for home-based screening.

The AI-based scoliosis self-screening application showed moderate validity in comparison to the scoliometer, suggesting that while it holds promise, there is room for further improvement. The app's good usability score indicates that it has the potential to be widely adopted for self-screening, offering a non-invasive and accessible alternative to traditional scoliosis monitoring. Future studies should focus on improving the accuracy and reliability of the app and validating it with a larger, more diverse sample of users.

While the current version of the AI-based scoliosis self-screening application has not yet achieved a high level of validity, it shows great promise for future improvements. With advancements in technology and the inclusion of larger datasets, its accuracy and reliability could be significantly enhanced, resulting in better performance for scoliosis screening. The strength of this study lies in its innovative use of AI for self-screening, representing a crucial step toward the broader adoption of digital tools for early detection of musculoskeletal conditions. By incorporating this app into physiotherapy practices, it could enable earlier interventions and improve patient outcomes. Moreover, its potential integration into school health programs and public health initiatives presents a scalable solution for widespread scoliosis screening.