Deep Learning-Enabled Cardiac Abnormality Detection: A Comprehensive Pipeline for Early Prediction and Image Enhancement in Medical Imaging

Deep Learning-Enabled Cardiac Abnormality Detection: A Comprehensive Pipeline for Early Prediction and Image Enhancement in Medical Imaging

Publication Date : 2024-05-02
Author(s) :

S. Venkata Kiran, N. Mounika, S. Sasank Kumar Reddy, N. Roopa, K. Bharath Kumar Reddy
Conference Name :

Sri Venkatesa Perumal College of Engineering and Technology
Abstract :

Early diagnosis of cardiac abnormalities plays a crucial role in effective treatment and management of cardiovascular diseases. In this project, a novel approach for small object detection and classification in cardiac images to facilitate improved early diagnosis. Our method involves a comprehensive pipeline consisting of input image pre processing, edge detection, boundary extraction, KAZE feature extraction, region mapping, morphological analysis, ensemble learning, and convolutional neural network (CNN) classifier for decision making. The pre processing stage aims to enhance image quality and reduce noise, preparing the input for subsequent analysis. Edge detection techniques are employed to identify prominent edges within the images, followed by boundary extraction to isolate potential regions of interest. KAZE feature extraction is utilized to capture key features from these regions, facilitating robust representation for subsequent classification tasks. Through region mapping and morphological analysis, extracted features are further refined to delineate small objects of interest within the cardiac images. An ensemble learning approach is then applied to integrate information from multiple classifiers, enhancing classification accuracy and robustness. Finally, a CNN classifier is employed for decision making, leveraging deep learning techniques to classify detected objects into relevant categories indicative of cardiac abnormalities.

No. of Downloads :

5