Abstract :

Cardiovascular Disease Is A Leading Cause Of Global Mortality, Necessitating Precise Diagnostic Methods For Early Identification. Echocardiography Functions As A Non-invasive Imaging Technique To Assess Anatomical And Functional Anomalies Of The Heart. This Study Employed A Publicly Available EKG Dataset Of Echocardiography Pictures To Construct A Dual-branch Deep Learning System For Classification And Anomaly Detection. Transfer Learning Models Utilized For Classification Included VGG16, EfficientNetB, Proposed VGG16, Xception, And An Ensemble Of VGG16 With Xception. The YOLO Model Family (YOLO V5, V8, V9, And V11) Was Employed For Detection To Identify Aberrant Locations Using Bounding Box Annotations. Experimental Results Indicated That The Ensemble Model Attained Enhanced Classification Performance, With An Accuracy Of 93.3% And An F1-score Of 93.4%. In Contrast, YOLO V8 Produced Optimal Detection Outcomes, With A Mean Average Precision (mAP) Of 43.1%, Providing A Dependable Method For Localization. To Improve Interpretability, Explainable AI Methods Like Grad-CAM Were Utilized To Emphasize The Discriminative Areas Influencing The Model S Judgments, Hence Providing Transparency In Medical Analysis. Additionally, The Trained Models Were Incorporated Into A Flask-based Web Framework To Facilitate Real-time Inference, Visualization Of Predictions, And Smooth Interaction Between Doctors And The Automated System. The Suggested System Exhibits Considerable Potential In Assisting Clinicians With Automated, Precise, And Interpretable Diagnoses From Echocardiography Pictures, Thus Diminishing Dependence On Manual Evaluation And Enhancing Diagnostic Efficiency.

Published:

08-4-2026

Issue:

Vol. 26 No. 4 (2026)

Page Nos:

1720-1731

Section:

Articles

License:

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

How to Cite