Development of Scoliotic Spine Severity Detection using Deep Learning Algorithms

According to research conducted by Johns Hopkins' Division of Pediatric Orthopedic Surgery, around three million new instances of Scoliosis are identified each year, with the majority of cases affecting children between the ages of 10 and 12. The current method of diagnosing and treating Scoliosis, which includes spinal injections, back braces, and a variety of other types of surgery, may have resulted in inconsistencies and ineffective treatment by professionals. Other scoliosis diagnosis methods have been developed since the technology's invention. Using Convolutional Neural Network (CNN), this research will integrate an artificial intelligence-assisted method for detecting and classifying Scoliosis illness types. The software model will include an initialization phase, preprocessing the dataset, segmentation of features, performance measurement, and severity classification. The neural network used in this study is U-Net, which was developed specifically for biomedical picture segmentation. It has demonstrated reliable and accurate results, with prediction accuracy reaching 94.42%. As a result, it has been established that employing an algorithm helped by artificial intelligence provides a higher level of accuracy in detecting Scoliosis than manual diagnosis by professionals.