Explainable and Lightweight Machine Learning Model for Cardiomegaly Detection from Chest XRay Images

Cardiomegaly, marked by an abnormal enlargement of the heart, is a critical indicator of cardiovascular disease. Traditional methods for diagnosing this condition, such as calculating the cardiothoracic ratio (CTR) from radiographic images, often suffer from variability and subjectivity among clinicians. This paper introduces a novel automated approach that combines K-means clustering with a specialized CTR calculation algorithm to enhance accuracy. The approach addresses the limitations of existing techniques by segmenting cardiac and thoracic structures in radiographic images. The proposed method provides a reliable solution to improve diagnostic accuracy significantly. Comprehensive evaluation shows that the method can achieve diagnostic accuracy levels of up to 92%, closely aligning with clinical standards. This research represents a significant advancement in cardiomegaly monitoring by integrating machine learning into clinical practice. Automating the CTR calculation reduces variability in diagnosis and enhances precision. It also ensures consistency with clinical guidelines. This innovative approach has the potential to revolutionize the detection and management of cardiomegaly. It leads to better patient outcomes, improved care, and more efficient clinical workflows. Additionally, minimizing subjectivity associated with traditional CTR assessment fosters a more standardized diagnostic process across different clinical settings. The method is highly scalable, practical, and beneficial for widespread clinical use.