Prediction of Spinal Abnormalities Using Machine Learning Techniques

Abstract

Lower back pain can be caused by many complications with any parts of the body in the lumbar spine. The compilation of a medical diagnosis is crucial to the medical practitioners in order for them to give a convenient treatment for the low back pain. The machine learning models that applied in the medical field for disease diagnosis assists medical experts in the diseases identification based on the symptoms at an early stage. This research aims to identify the most significant physical parameters that contribute to spinal abnormalities and also predict spinal abnormalities based on collected physical spine data by using unsupervised machine learning approaches such as Principal Component Analysis (PCA), and also using supervised machine learning approaches such as K-Nearest Neighbors (KNN) and Random Forest (RF). As a result, degree spondylolisthesis is the most significant parameter that contributes to spinal abnormalities. As a comparison of results between RF classifier and KNN classifier, KNN classifier performed better than RF classifier since the percentage of accuracy of KNN algorithm (85.32%) are higher compared to RF classifier (79.57%).