Deep learning for prediction of cardiomegaly using chest X-rays

• Published in: Neural computing & applications 2024-11, Vol.36 (31), p.19383-19391
• Main Authors: Gupta, Mrigakshi, Singh, Akash, Kumar, Yatender
• Format: Article
• Language: English
• Subjects: Artificial Intelligence; Artificial neural networks; Chest; Computational Biology/Bioinformatics; Computational Science and Engineering; Computed tomography; Computer Science; Data Mining and Knowledge Discovery; Deep learning; Diagnosis; Heart diseases; Image Processing and Computer Vision; Medical imaging; Original Article; Probability and Statistics in Computer Science; X-rays
• ISSN: 0941-0643; 1433-3058
• DOI: 10.1007/s00521-024-10190-6

In the past decade, deep learning in biomedical imaging has exponentially increased the accuracy of disease detection and improved the health standards. This research paper introduces a novel approach for the early detection and diagnosis of cardiomegaly using the cardiothoracic ratio (CT ratio) measurement in chest X-ray scans. Cardiomegaly is a serious cardiac condition that can lead to life-threatening complications if left undiagnosed. The proposed method involves segmenting the heart from a chest CT scan using a convolutional neural network model, ResNet-18, and calculating the CT ratio, which is the ratio of the maximum width of the heart to the maximum width of the thoracic cage. Studies have shown that increasing CT ratio leads to an increasing risk of heart diseases. Hence, a need to monitor the CT ratio for every individual arises, for if the ratio changes, an alert to take precautions can be rang. The method is evaluated using a dataset of 490 chest X-ray scans, and it achieves an accuracy of 80% and a precision of 84%. The integration of CT ratio measurement in chest X-ray scan reports has the potential to aid in the early detection and diagnosis of cardiomegaly, allowing for prompt medical intervention and improving patient outcomes.