Cross-Domain Transfer Learning for PCG Diagnosis Algorithm

Area of Science:

• Biomedical Engineering
• Cardiology
• Artificial Intelligence in Medicine

Background:

• Cardiovascular diseases remain a leading cause of mortality worldwide.
• Traditional diagnostic methods for heart conditions can be subjective and time-consuming.
• Phonocardiogram (PCG) signals offer a non-invasive window into cardiac function.

Purpose of the Study:

• To develop an advanced deep learning framework for the automated diagnosis of cardiovascular disease.
• To improve the accuracy and efficiency of heart condition assessment using phonocardiogram (PCG) signals.
• To introduce novel signal segmentation techniques for enhanced diagnostic performance.

Main Methods:

• Utilized phonocardiogram (PCG) signals as the primary diagnostic data.
• Developed a novel deep learning architecture incorporating transfer learning and boosting techniques.
• Designed an improved signal segmentation method, moving beyond traditional R-R interval and fixed segmentation.

Main Results:

• The proposed deep learning framework demonstrated sustainable diagnostic performance.
• The enhanced architecture and segmentation method contributed to improved accuracy in heart condition identification.
• Evaluation was conducted using a public phonocardiogram (PCG) database, validating the model's effectiveness.

Conclusions:

• The developed deep learning framework offers a promising approach for objective and efficient cardiovascular disease diagnosis.
• The integration of transfer learning, boosting, and advanced segmentation significantly enhances PCG-based diagnostics.
• This methodology holds potential for widespread clinical application in early heart condition detection.