Complexity analysis of heart sounds

• Main Author: Chew, Aaron Wai Kin
• Other Authors: Soh Cheong Boon
• Format: Final Year Project
• Language: English
• Published: 2016
• Subjects: DRNTU::Engineering
• Online Access: http://hdl.handle.net/10356/68274
• Institution: Nanyang Technological University

In the healthcare industry, it is a challenge of being able to determine clearly whether a heart signal is Normal or Abnormal. To solve this, it is essential to develop a system that is able to detect and define the difference between Normal and Abnormal heart signal. The purpose of this project is to work on existing researches and findings to integrate technics that can perform such diagnosis automatically to prevent any misdiagnosis. Phonocardiography (PCG) has shown its potential in recent years. A PCG is a diagnostic technique that produces graphical recording of the sound generate by the closing of the valves in a heart. This graphical recording is called phonocardiogram. One of the methods to capture these sounds into a phonocardiogram is by listening to a heart sounds with a stethoscope. Studies have shown that fractal structures can found in biomedical time series and multifractal nature of a biomedical signal has captured more attention among researchers. Most of the researches regarding cardio-signal analysis identifies the use of multifractals is oriented towards heart signals. In this report, a Multifrcatal detrended fluctuation analysis (MFDFA) algorithm was implement in MATlab to conduct experiments. This paper represents a step towards automatic detection of a human heart whether a heart sound produced is normal or abnormal with the use of phonocardiogram and multifractal analysis. Also, a system that is able to generate result that is accurate and efficient. A total of 23 phonocardiographic recording, consisting of normal and abnormal heart signal are used for this experiment. Results obtained at the end of the experiment, it has shown that the method used produced an 86.95% of accuracy on the dataset of 23 PCG recordings.