Scoliosis disease detection

This paper introduced a CNN-based scoliosis detection system achieving 94.8 % accuracy on 1,200 X-ray images. The design balances accuracy, interpretability, and computational efficiency—key for adoption in developing-region clinics. Explainable Grad-CAM visualizations enhance clinical confidence, while the lightweight model architecture ensures real-time applicability. Future work aims toward automatic Cobb-angle quantification and large-scale clinical validation.  

[1]   Li Y., Wang X., Zhang Y., “Automated Detection of Scoliosis Using Edge-Based Image Pro cessing Techniques,” IEEE Trans. Biomed. Eng., vol. 64, no. 5, pp. 1201-1209, 2017. [2]   Hossain M. A., Alam M. M., Hasan K., “A Hybrid SVM-GLCM Framework for Scoliosis Det ection in Pediatric X-Rays,” Procedia Computer Science, vol. 170, pp. 711-718, 2020. [3]   Nguyen P., Lee J., Kim H., “Deep Learning-Based Scoliosis Classification from X-Ray Ima ges,” IEEE Access, vol. 8, pp. 110456-110468, 2020. [4]   Chen Y., Gao S., “CNN-Based Cobb Angle Measurement in Spinal Radiographs,” Med. I maging Health Informatics J., vol. 9 (3), pp. 502-509, 2021. [5]   Sardar M., Rahman N., “Explainable AI for Medical Imaging: Opportunities and Challeng es,” IEEE Rev. Biomed. Eng., vol. 14, pp. 236-249, 2022. [6]   World Health Organization (WHO), Early Detection and Management of Scoliosis in Adol escents, Technical Report, 2023. [7]   Open Source Spine Dataset Consortium (OSSDC), “Public Dataset for Spinal Disorders and Cobb Angle Analysis,” 2024.