Alliance for Pandemic Preparedness
May 20, 2020
Artificial Intelligence-Enabled Rapid Diagnosis of Patients with COVID-19
Category: Article Summary
Topic: Testing and Treatment
- Mei et al. developed an artificial intelligence (AI) algorithm that integrates chest CT findings with clinical history to provide rapid diagnosis of COVID-19. The algorithm achieved high discriminative performance (AUC=0.92) with equal sensitivity compared to a senior thoracic radiologist (84.3% vs 74.6%) on a test set of 279 patients.
- The algorithm also correctly identified 17 of 25 patients who presented with normal CT scans but had positive RT-PCR results, whereas radiologists classified all as COVID-19 negative
- While the generalizability of the algorithm is limited by sample size, it has useful implications as a potential screening tool.
Mei et al. (May 19, 2020). Artificial Intelligence-Enabled Rapid Diagnosis of Patients with COVID-19. Nature Medicine. https://doi.org/10.1038/s41591-020-0931-3