Early Stage Detection of Scoliosis Using Machine Learning Algorithm

Scoliosis is a spine disorder phenomenon that makes spine bend and it will form letter C or S. Manual method in determining the curvature of the spine results in a very low accuracy value. This occurs because of the noise present in the spinal x-ray images of patients with scoliosis. The noise can be either organ, blood, or bone that makes the calculation of curvature of the spine to be inaccurate. Therefore in this research the application of scoliosis classification based on the level of curvature of the spine will be made using images processing approach. The spinal curvature classification consists of several processes. The process begins by reprocessing the spinal bone image using the median filter and morphological watershed method. Then it proceeds with features extraction and then the process of classification begins with artificial neural network method. It is expected that the classification of curvature of the spine to be more accurate and can be handled in accordance with its classification.

Keywords: —Deep Neural Network, SVM, Convolutional Neural Network, Image Classification, Image Recognition, Transfer Learning, Machine learning.