Diagnosis of Vertebral Column Disorders Using Machine Learning Classifiers

Medical science is characterized by the correct diagnosis of a disease and its accurate classification to avoid complexities at treatment/medication stage. This is often accomplished by a physician based on experience without much signal processing aids. It is envisioned that a sophisticated and intelligent medical diagnostic/classification system may be helpful in making right decisions especially at remote areas where specialist physicians are not present. With this in mind, this paper proposes diagnosis and classification of vertebral column disorders using machine learning classifiers including feed forward back propagation neural network, generalized regression neural network and support vector machine and evaluates their performance. The dataset is collected using the information from the magnetic resonance images (MRI) and is classified into three different classes which are disk hernia, spondylolisthesis and normal. The classifiers are trained using 50% ratio and 10 fold cross validation approaches and are comprehensively evaluated with different architectures, activation and kernel functions. Experimental results demonstrate that feed forward back propagation neural network is 93.87% accurate on unknown test cases and performs better than the other methods.