Researchers have developed an artificial intelligence method that can identify a range of acute neurological illnesses in CT scans within a few seconds, when time is essential in assessing the life-threatening conditions.
Conditions such as stroke, hemorrhage and hydrocephalus were identified much faster with deep learning than through human diagnosis, according to a study conducted at the Icahn School of Medicine at Mount Sinai and published Monday in the journal Nature Medicine.
“With a total processing and interpretation time of 1.2 seconds, such a triage system can alert physicians to a critical finding that may otherwise remain in a queue for minutes to hours,” senior author Dr. Eric Oermann, an instructor in the Department of Neurosurgery at Mount Sinai, said in a press release. “We’re executing on the vision to develop artificial intelligence in medicine that will solve clinical problems and improve patient care.”
The Mount Sinai AI Consortium, known as “AISINAI,” first developed first AI method to assess the neurological illnesses. The consortium is a group of scientists, physicians and researchers dedicated to developing artificial intelligence for practical uses in medicine.
RELATED AI system can predict side effects of drug combinations
“The application of deep learning and computer vision techniques to radiological imaging is a clear imperative for 21st century medical care,” said study author Dr. Burton Drayer, chairman of radiology at the Mount Sinai Health System.
Researchers programmed the AI system by using 37,236 head CT scans to identify whether an image contained critical or non-critical findings. It used “weakly supervised learning approaches” using natural language processing and large clinical datasets from the Mount Sinai Health System.
In all, 96,303 reports were analyzed. Images marked urgent — or STAT — took an average time of 174 minutes from when the test was ordered until a preliminary report is published, while routine ones took 241 minutes among radiologists. That includes the gap between after the test and when a radiologist looks at the scans.
RELATED New retina model may eventually predict outcome of eye diseases
The computer software was tested for how quickly it could recognize and provide notification compared with the time it took a radiologist to determine a disease.
It took the physician 150 times longer — three minutes — to assess the image.
Within the next two years, the researchers expect to have enhanced computer labeling of CT scans and a shift to “strongly supervised learning approaches” and novel techniques for increasing data efficiency.
RELATED AI detects stroke, dementia from brain scans
“The expression ‘time is brain’ signifies that rapid response is critical in the treatment of acute neurological illnesses, so any tools that decrease time to diagnosis may lead to improved patient outcomes,” study co-author Dr. Joshua Bederson, chairman for the Department of Neurosurgery at Mount Sinai.