Deep Learning Classification of Cardiomegaly Using Combined Imaging and Non-imaging ICU Data

In this paper, we investigate the classification of cardiomegaly using multimodal data, combining imaging data from chest radiography with routinely collected Intensive Care Unit (ICU) comprising vital sign values, laboratory measurements, and admission metadata. practice a clinician would assess for presence synthesis multiple sources however, prior machine learning approaches to task have focused on radiographs only. We show that non-imaging ICU can be used propose novel network trained simultaneously both data. compare predictive power single-mode joint network. use subset publicly available MIMIC-CXR MIMIC-IV datasets, which contain same patients. The approach alone achieves an AUC 0.684 standard 0.840. Our model 0.880. conclude value cardiomegaly, has potential improve performance patients, further work required demonstrate significant improvement.