Classification of Normal and Abnormal Heart Sounds Using The Combination of Metaheuristic Algorithms and Machine learning.

Abstract:

Nowadays, Cardiovascular diseases are known as one of the most important factors that threaten human health. They are among the most important causes of death in society. Many heart diseases and abnormalities can be diagnosed and evaluated using listening techniques. Listening to the sound of the heart has been one of the first practical methods for diagnosing heart diseases. This method has been developed as a cheap and non-invasive solution to check heart diseases by using computers. Phonocardiography (PCG) signals give information about the function of the heart valves during the heartbeat. These signals can be useful in the early diagnosis of heart diseases. Automatic heart sound classification has promising potential in the field of heart pathology. In this research, an automatic method for discriminating between normal and abnormal heart sounds is proposed. In this method, first, the heart sounds are segmented to 4 main parts: s 1   and s 2  sounds, systole and diastole segments. From these segments, statistical and time frequency features are extracted for classification. Before classification, we select the distinctive features using two approaches. In the first approach, the feature selection is  accomplished using particle swarm optimization algorithm (PSO).  In the second approach, we use Sequential Forward Feature Selection (SFFS) method. The proposed method was evaluated on the Physionet 2016 Challenge database using 10-fold  cross-validation method. In this database, the number of normal and abnormal PCG signals are not balanced therefore, in this paper, the synthetic minority over-sampling technique (SMOTE) is applied to produce balanced data. The evaluation results showed that the proposed method can distinguish the normal heart sounds from abnormal ones with accuracy of  98/03% and sensitivity and specificity of  97.64%, 98/43%respectively