Designing a computer-assisted diagnosis system for cardiomegaly detection and radiology report generation

Chest X-ray (CXR) is a conventional diagnostic tool for cardiothoracic assessment, boasting a high degree of costeffectiveness and versatility. However, with an increasing number of scans to be evaluated by radiologists, they can suffer from fatigue which might impede diagnostic accuracy and slow down report generation. We describe a prototype computer-assisted diagnosis (CAD) pipeline employing computer vision (CV) and Natural Language Processing (NLP) trained on the publicly available MIMICCXR dataset. We perform image quality assessment, view labelling, segmentation-based cardiomegaly severity classification, and use the output of the severity classification for large language model-based report generation. Four certified radiologists assessed the output accuracy of the CAD pipeline. Across the dataset composed of 377,100 CXR images and 227,827 free-text radiology reports, our system identified 0.18% of cases with mixedsex mentions, 0.02% of poor quality images (F1=0.81), and 0.28% of wrongly labelled views (accuracy 99.4%), furthermore it assigned views for 4.18% of images which have unlabelled views. For binary cardiomegaly classification, we achieve state-of-the-art performance of 95.2% accuracy. The inter-radiologist agreement on evaluating the report’s semantics and correctness for radiologistMIMIC is 0.62 (strict agreement) and 0.85 (relaxed agreement) similar to the radiologist-CAD agreement of 0.55 (strict) and 0.93 (relaxed). Our work found and corrected several incorrect or missing metadata annotations for the MIMIC-CXR dataset, and the performance of our CAD system suggests performance on par with human radiologists. Future improvements revolve around improved text generation and the development of CV tools for other diseases.