Artificial intelligence for the detection of COVID-19 pneumonia on chest CT using multinational datasets

TLDR

Chest CT is emerging as a valuable diagnostic tool for managing COVID‑19 lung disease, and AI may rapidly differentiate COVID‑19 findings from other clinical entities. The study trained deep‑learning models on a multinational cohort of 1,280 patients, using CT scans from oncology, emergency, and pneumonia indications as normal controls, to localize pleura/lung parenchyma and classify COVID‑19 pneumonia. In an independent test set of 1,337 patients, the AI achieved 90.8% accuracy (84% sensitivity, 93% specificity), with a 10% false‑positive rate among 140 patients with other pneumonias, demonstrating high specificity in distinguishing COVID‑19 pneumonia from non‑COVID pneumonias.

Abstract

Chest CT is emerging as a valuable diagnostic tool for clinical management of COVID-19 associated lung disease. Artificial intelligence (AI) has the potential to aid in rapid evaluation of CT scans for differentiation of COVID-19 findings from other clinical entities. Here we show that a series of deep learning algorithms, trained in a diverse multinational cohort of 1280 patients to localize parietal pleura/lung parenchyma followed by classification of COVID-19 pneumonia, can achieve up to 90.8% accuracy, with 84% sensitivity and 93% specificity, as evaluated in an independent test set (not included in training and validation) of 1337 patients. Normal controls included chest CTs from oncology, emergency, and pneumonia-related indications. The false positive rate in 140 patients with laboratory confirmed other (non COVID-19) pneumonias was 10%. AI-based algorithms can readily identify CT scans with COVID-19 associated pneumonia, as well as distinguish non-COVID related pneumonias with high specificity in diverse patient populations.