Natural Language Processing and Machine Learning for Detection of Respiratory Illness by Chest CT Imaging and Tracking of COVID-19 Pandemic in the United States
Ricardo C. Cury, István Megyeri, Tony Lindsey, Robson Macedo, Juan C. Batlle, Shwan Kim, Brian L. Baker, Robert J. Harris, Reese H. Clark
Abstract
Purpose To determine if natural language processing (NLP) algorithm assessment of thoracic CT imaging reports correlated with the incidence of official COVID-19 cases in the United States. Materials and Methods With the use of de-identified HIPAA compliant patient data from a common imaging platform interconnected with over 2100 facilities covering all 50 states, three NLP algorithms were developed to track positive CT imaging features of respiratory illness typical in SARS-CoV-2 viral infection. Findings were compared against the number of official COVID-19 daily, weekly, and state-wide. Results The NLP algorithms were applied to 450,114 patient chest CT comprehensive reports gathered from January 1 to October 3, 2020. The best performing NLP model exhibited strong correlation with daily official COVID-19 cases (r2 = 0.82, P < .005). The NLP models demonstrated an early rise in cases followed by the increase of official cases, suggesting the possibility of an early predictive marker, with strong correlation to official cases on a weekly basis (r2 = 0.91, P < .005). There was also substantial correlation between the NLP and official COVID-19 incidence by state (r2 = 0.92, P < .005). Conclusion With the use of big data, a machine learning–based NLP algorithm was developed that can track imaging findings of respiratory illness detected on chest CT imaging reports with strong correlation with the progression of the COVID-19 pandemic in the United States. Supplemental material is available for this article. Keywords: CT, Infection, Lung, Technology Assessment, Thorax © RSNA, 2020