International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 04 | Apr 2024
p-ISSN: 2395-0072
www.irjet.net
Cardiac Ultrasound Image Segmentation Using LU-Net Bharath B A1, Bharath M2, Vinanth U Aradhya3 , Sagar4 1 UG student Dept. of Computer Science and Engineering, Bangalore, Karnataka, India 2UG student Dept. of Computer Science and Engineering, Bangalore, Karnataka, India
UG student Dept. of Computer Science and Engineering, Bangalore, Karnataka, India UG student Dept. of Computer Science and Engineering, Bangalore, Karnataka, India ---------------------------------------------------------------------***--------------------------------------------------------------------3 4
Abstract - The detection and treatment of cardiac disorders
variability and improve diagnosis accuracy. This project aims to transform cardiac imaging by using a dataset of over 1000 echocardiographic pictures to provide a more reliable, accurate, and effective diagnostic tool for cardiology.
have greatly benefited by developments in the cardiac imaging technologies. This research presents a novel method for segmenting cardiac ultrasound images using the LU-Net model, a deep learning framework intended to improve the precision and effectiveness of cardiac diagnosis. Echocardiography, another name for cardiac ultrasound, is a crucial non-invasive diagnostic method in cardiology. But the subjective character of the analysis frequently limits the ability of highly trained clinicians to interpret echocardiographic images. The LU- Net model uses an advanced convolutional neural network design to overcome these difficulties. This architecture has shown impressive results in the automated segmentation of heart structures from ultrasound pictures. Over 10,000 echocardiogram pictures representing a variety of cardiac diseases and patient demographics were gathered and analyzed for the research. With a precision rate of 92 %, recall of 93 %, and segmentation accuracy of 94.5 %, the LUNet model was developed through rigorous training and validation. Comparing these performance measurements to more conventional approaches, which generally show 80–85% accuracy levels, shows a considerable improvement. Because of the LU- Net model's accuracy and speed, cardiac diagnostics workflow is streamlined, and earlier and more accurate diagnosis of cardiac anomalies leads to significantly better patient outcomes. Thus, this initiative represents a major advancement in cardiac imaging technology and provides physicians with an effective tool for the detection and treatment of heart ailments.
2. PROBLEM STATEMENT With the help of deep learning, this project seeks to automate Left Ventricular segmentation, build a reliable pipeline for data collecting and annotation, enhance annotation accuracy, and maximize GPU based model training. A dependable deep learning model, a variety of ultrasound datasets, a well-organized workstation, and customized models for different image kinds will all be part of the final product, which will improve cardiac ultrasound analysis and clinical decision support.
3. CONTRIBUTIONS With cutting-edge methods and sophisticated machine learning algorithms, the LU-Net project completely transforms the segmentation of cardiac ultrasonography images. LU-Net is essential to improving patient outcomes since it standardizes picture interpretation and improves diagnostic accuracy. Overall, LU-Net represents a revolutionary change in the direction of more dependable and easily available cardiac imaging technology, significantly advancing the area of medical imaging. The contributions of this project include:
Key Words: Convolution Neural Network (CNN), Machine Learning (ML), Artificial Intelligence (AI), Cardiac Disease, Echocardiogram, Electrocardiogram (ECG), Image Classification, Healthcare Technology.
Our e ort presents a state-of-the-art technique for cardiac ultrasound segmentation by utilizing LU-Net, an advanced deep learning model. The clarity and precision of ultrasonography pictures are much improved by this approach, which is essential for making an accurate diagnosis.
1.INTRODUCTION A vital component of cardiac diagnosis, cardiac ultrasonography is used in more than 20 million echocardiograms performed globally each year. Despite being widely used; cardiac ultrasonography accuracy is largely dependent on the operator's skill; studies have shown that practitioners can di er up to 20% in how they interpret images. By incorporating cutting-edge machine learning methods into the processing of cardiac ultrasound data, the LUNet project tackles this problem. By standardizing image interpretation, we hope to lower
© 2024, IRJET
|
Impact Factor value: 8.226
Innovative Segmentation Technique:
Standardization of picture Interpretation:
Among the most important contributions has been the standardization of picture interpretation for sonographers with different skill levels. The variability in cardiac ultrasound diagnosis that can reach 20% at present must be reduced, and this standardization is essential to that goal.
|
ISO 9001:2008 Certified Journal
|
Page 2517