By Kat Jerich for Healthcare IT News
In a study published this week in Nature Medicine, researchers used AI algorithms in conjunction with chest CT scans and patient history to quickly diagnose patients who were positive for COVID-19 and improve the detection of patients who presented with normal CT scans.
“We were able to show that the AI model was as accurate as an experienced radiologist in diagnosing the disease, and even better in some cases where there was no clear sign of lung disease on CT,” said Dr. Zahi Fayad, director of the BioMedical Engineering and Imaging Institute at the Icahn School of Medicine at Mount Sinai, in a statement.
WHY IT MATTERS
Because the symptoms of COVID-19 are non-specific, it can be difficult to diagnose. Meanwhile, the SARS-CoV-2 virus-specific reverse transcriptase polymerase chain reaction (RT-PCR) test commonly used to identify COVID-positive patients can take up to two days to complete – and clinicians face the possibility of false negatives. RT-PCR test kits are also in short supply throughout many parts of the country.
This, researchers say, reiterates the need for other ways to quickly and accurately diagnose patients with COVID-19.
Researchers relied on CT scans of more than 900 patients that had been admitted to 18 medical centers in 13 Chinese provinces. They included 419 confirmed COVID-19-positive cases and 486 COVID-19-negative scans. The team also had access to patients’ clinical information, including blood test results, age, sex and symptoms.
Using patient data, Mount Sinai researchers developed an AI algorithm to produce separate probabilities of COVID-19 positivity based on CT images, clinical information and the two combined.
“In a test set of 279 patients, the AI system achieved an area under the curve of 0.92 and had equal sensitivity as compared to a senior thoracic radiologist,” researchers wrote.
In addition, the algorithm correctly identified 17 of 25 patients whose RT-PCR results had tested positive for COVID-19 but who presented with normal CT scans; for comparison, radiologists had classified all the patients as COVID-negative.
Although clinicians in the United States do not frequently use CT scans to diagnose COVID-19, researchers say imaging can play a vital role in conserving hospital resources and treating patients quickly.
“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,” said Fayad.
“It’s something that should be considered on a wider scale, especially in the United States, where currently we have more spare capacity for CT scanning than in labs for genetic tests,” Fayad continued.
THE LARGER TREND
Researchers have increasingly relied on AI to diagnose and treat patients with the novel coronavirus.
In March, cognitive computing platform vendor behold.ai announced it had developed an AI-based algorithm to flag chest X-rays from COVID-19.
Calling its platform “instant triage,” behold.ai predicted it could help speed COVID-19 diagnosis.
“As we evaluate further positive cases from across the world, our results will be further validated,” said behold.ai Chief Medical Officer Dr. Tom Naunton Morgan.
“This will increase the utility of our instant triage and potentially help reduce the burden on healthcare systems as more and more cases of pneumonia present and require rapid diagnosis,” Morgan said.
Other technology vendors have adapted existing tuberculosis-detecting AI technology to help indicate COVID-affected lung tissue in chest X-rays.
ON THE RECORD
Mount Sinai researchers say their next steps will be to further develop the model to forecast patient outcomes and to share their results with other healthcare facilities.
“This study is important because it shows that an artificial intelligence algorithm can be trained to help with early identification of COVID-19, and this can be used in the clinical setting to triage or prioritize the evaluation of sick patients early in their admission to the emergency room,” said Dr. Matthew Levin, director of the Mount Sinai Health System’s clinical data science team.
“This is an early proof [of] concept that we can apply to our own patient data to further develop algorithms that are more specific to our region and diverse populations,” said Levin.
“This toolkit can easily be deployed worldwide to other hospitals, either online or integrated into their own systems,” said Fayad.