COVID Comorbidity Prediction from Radiology Images

Researchers at the University of Illinois developed new software that enables users to predict overall health, outcomes, and cost. Also, it can find missing diagnosis codes utilizing medical imaging, such as a chest radiograph, without additional information from the medical record or manual review. This software not only improves the billing system by identifying missing diagnosis codes, but the deep learning techniques also allow making predictions regarding probabilities of comorbidities such as morbid obesity, diabetes, vascular disease, and COPD using a chest radiograph, which serves as a complex biomarker. Notably, this software can accurately predict the need for hospitalization of greater than two days' duration with COVID-19 infection, reducing patients' risk of potential complications and allow providing efficient treatment translates to fewer chronic conditions.