CNN Approach for Scoliosis Detection

Scoliosis is a complex three-dimensional deformity of the spinal column characterized by abnormal lateral curvature and vertebral rotation, primarily affecting adolescents during growth spurts. If left undetected, scoliosis can lead to severe postural imbalance, respiratory issues, and chronic pain. Conventional diagnostic procedures rely on manual Cobb angle measurement from spinal X-rays, which requires significant clinical expertise and is prone to human error and subjectivity. In modern healthcare systems, there is an urgent need for computer-aided diagnostic tools that can assist radiologists by providing consistent, fast, and accurate assessment of spinal curvature. In this project, we propose a machine learning–based scoliosis detection framework using X ray image analysis integrated with Artificial Intelligence (AI) and Machine Learning (ML) methodologies. The primary objective is to automate the identification of scoliosis patterns at an early stage by analyzing radiographic features without invasive procedures. The proposed system employs a Support Vector Machine (SVM) classifier trained on a curated dataset of spinal X-ray images. Prior to classification, several image pre-processing and enhancement techniques such as grayscale conversion, histogram equalization, and region-of-interest extraction are applied to improve contrast and visibility of the spinal structure. Feature extraction is performed using Histogram of Oriented Gradients (HOG) to capture spinal curvature shape information, and Gray-Level Co-occurrence Matrix (GLCM) to represent textural patterns of vertebrae and surrounding tissues.