CCS599: DISSERTATION
SESSION 2023-2024__, SEMESTER 2
RESEARCH PROPOSAL FORM
A Hybrid Deep Learning Model for Diagnosing Heart
Research Synopsis (200-250 words):
Cardiovascular diseases are a major cause of death globally, and the increasing annual fatalities necessitate improved diagnostic methods. Current techniques have limitations in speed, accuracy, and invasiveness, making the development of new, accurate, rapid, and non-invasive diagnostics vital for timely treatment and prevention.
Traditional deep learning models struggle with image quality issues and fail to comprehensively analyze multi-source data like ECGs and ultrasound, limiting their ability to accurately diagnose heart diseases.
This study aims to enhance cardiac diagnosis by refining an existing CNN model through integration with alternative models or parameter adjustments, creating a hybrid deep learning model that can accurately and efficiently diagnose heart conditions.
The MIT-BIH ECG Database is a high-quality resource containing 48 half-hour dual-channel ECG recordings digitized at 360 samples per second and annotated by two experts, providing approximately 110,000 annotation points.
This research project expects to:
- Develop a groundbreaking diagnostic method for heart disease that's accurate, fast, and non-invasive.
- Validate the practical use of deep learning algorithms in diagnosing heart diseases.
- Propose a solution applicable to broader medical image recognition and analysis, potentially influencing advancements across multiple imaging fields.
This study's objective is to create a hybrid deep learning model to improve heart disease diagnosis accuracy and efficiency. By overcoming current deep learning limitations, it seeks to facilitate early detection and intervention, thereby enhancing patient quality of life, reducing societal and individual burdens, and showcasing significant research value and potential applications.