Artificial intelligence detects COVID-19 for the first time in US
Click Here to Manage Email Alerts
Researchers at the Icahn School of Medicine at Mount Sinai created the first algorithm of its kind in the United States to use artificial intelligence, combined with CT imaging of lungs and clinical data, to detect COVID-19, according to a press release.
“We have a collaboration with different hospitals in China,” Zahi Fayad, PhD, director of the BioMedical Engineering and Imaging Institute at the Icahn School of Medicine at Mount Sinai, told Healio Primary Care. “We integrated some of their patients’ CT data with clinical data to come up with a better diagnosis.”
The AI model — dubbed convolutional neural network, or CNN, by researchers — is based on CT scans and other clinical information such as abnormalities in white blood cell counts, age, gender and symptoms (fever, cough or cough with mucus) of 905 patients (mean age, 40.7 years; 488 men) admitted to 18 medical centers in 13 Chinese provinces between Jan. 17 and March 3, 2020. Among these cases, 419 confirmed were positive for COVID-19.
Fayad and colleagues randomly split the cohort into a training set comprised of 242 positive cases of COVID-19 and 292 negative cases; a test set of 134 positive and 145 negative cases; and a tuning set of 49 negative cases and 43 positive cases.
According to results published in Nature Medicine, the CNN had 84% sensitivity, whereas radiologists evaluating the images and clinical data had 75% sensitivity among the test cases. The AI system also improved the detection of patients who were positive for COVID-19 via reverse transcriptase polymerase chain reaction (RT-PCR) who presented with normal CT scans, correctly identifying 17 of 25 test patients.
The results suggest that CT could play an “important role” in the fight against COVID-19, Fayad said.
“Imaging can help give a rapid and accurate diagnosis,” he said in a press release. “Lab tests can take up to 2 days, and there is the possibility of false negatives — meaning imaging can help isolate patients immediately if needed and manage hospital resources effectively. The high sensitivity of our AI model can provide a ‘second opinion’ to physicians in cases where CT is either negative (in the early course of infection) or shows nonspecific findings, which can be common.”
Fayad said the robustness of the AI model is now being studied in more than 6,000 patients within the Mount Sinai Health System.
“I’m hoping that after that, we can deploy the model to our hospitals to see if it stands the test of time,” Fayad said in the interview. “If those study results are positive, we would love to collaborate with other health systems to use this tool as much as possible.”
References: