Efficient Diagnosis And Analysis Of Cardiovascular Disease Through Computational Intelligence

Abstract

The unhealthy life style and the dynamic conditions of environmental changes has rapidly increased the chances of heart diseases. An early diagnosis of heart diseases can minimize the future critical effects and fatal conditions. The need of automation in medical domain has increased in many folds in recent time. The automation systems are primarily targeted for early monitoring of diseases. The automation has a great help at the time of diagnosing and criticality in diagnosis of a disease. With advancement of new technologies in learning system, the processing and classification of observing data has attained speed and accuracy in it. However, the difficulty in observing data and its dependency on the classification process resulted into a large data processing. This limits the application of automation system in different critical usage. The objective of speedy processing and infallible accuracy with low processing overhead for early diagnosis of heart diseases is focused in the proposed research work. The presented approach developed a new data representation based on the characteristic representation of the monitoring parameters. Fourteen monitoring parameters referred for heart disease diagnosis from the standard Cleveland data set. The said parameters were used in the processing of heart disease diagnosis. A weighted clustering approach based on distance and gain parameters in clustering is presented. The proposed data sub clustering approach enhances the learning performance and it resulted into a faster and accurate decision system as compared to present approaches. In order to enhance the decision accuracy in addition to separate data monitoring, a continuous observation from ECG signal is proposed. Twelve descriptive features of ECG signal that defines the characteristic and variations related to heart operation are developed. The feature overhead is addressed to minimize by a fusion approach, where a selective approach of feature vector for a learning approach using neural network is presented. The proposed selective approach for fusion approach resulted in faster decision due to low feature counts and improved the accuracy of decision due to weighted method proposed.