Computer aided diagnostic system based on SVM and K harmonic mean based attribute weighting method

Abstract

Listen

Translate article

IntroductionMachine learning techniques are popular tool adopted for medical diagnosis and one of the core component of medical diagnostic system. The objective of machine learning techniques is to provide accurate and timely diagnostic results during disease diagnosis phases. Further, it also helps the physicians and medical practitioner regarding disease diagnosis. The objective of this work is to improve the diagnostic accuracy of computer aided diagnostic system. MethodLarge number of machine learning techniques are integrated in the computer aided diagnostic system for the prediction of the diseases. These machine learning techniques consider different features of disease to diagnosis the disease. It is seen that all features are not equally important in diagnostic process and irrelevant features can lead to low prediction rate. Hence in medical field, identification of irrelevant features is warm area of research. To identify the relevant features for disease prediction, attribute weighting methods are adopted. It is observed relevant features can improve the diagnostic accuracy of computer aided systems. Hence, to improve the diagnostic accuracy rate, a k harmonic mean based attribute weighting method is developed, called KhmAW. Further, the proposed KhmAW method is integrated with SVM method, called KhmAW-SVM. In KhmAW-SVM, KhmAW method is used to identify the relevant features from dataset and SVM method is applied for diagnosis the disease. The proposed method classifies the datasets into healthy and non-healthy classes. ResultsFour datasets are used to validate the proposed KhmAW-SVM based computer aided diagnostic system. These datasets are Statlog heart disease, Parkinson disease, Liver disease and Pima Indian diabetes disease datasets and having non linearly separable data distribution. The simulation results of proposed KhmAW-SVM method are evaluated using accuracy rate. Further, the simulation results are assessed using 50-50 training-testing and 10 fold methods. It is stated that proposed KhmAW-SVM method achieves 94.28%, 99%, 89.93% and 92.38% accuracy rates for heart disease, Parkinson's disease, liver disease and diabetes disease respectively. ConclusionThe efficacy of the proposed method is evaluated using four well known diseases datasets and compared with large number of existing studies. It is stated that proposed KhmAW-SVM based computer aided diagnostic system achieves better quality results as compared to existing studies. Hence, it is concluded that proposed computer aided diagnostic system can improve the clinical decision making process and also help the physician and doctors regarding different diseases.