With the advent of artificial intelligence and medical big data, the medical field has seen tremendous groundbreaking developments. The automated classification of heart sounds is essential in diagnosing cardiovascular diseases. Cardiac auscultation is one of the most reliable and non-invasive methods to detect the presence of heart disease. However, a trained ear and considerable experience are required to diagnose heart diseases. The development of deep learning methodology for classifying heart diseases is one of the major breakthroughs in the field of medicine. This project aims to use a customized CNN model to detect heart diseases.
- Employed the CNN-based EfficientNet architecture with custom layers to detect cardiovascular anomalies in Phonocardiograms.
- Designed a 4-stage architecture involving denoising, segmentation, feature extraction, classification to classify the audio samples.
- Developed a user-friendly progressive web app using Gradio enabling users to effortlessly upload phonocardiograms for automated evaluation.