A team of Wexner Medical Center researchers have developed a program that can distinguish which patients need immediate attention.

Luciano Prevedello, a radiologist at Ohio State, led research that helped build artificial intelligence that analyzes hundreds of CT brain scans. It was created to decrease the time it takes for patients in critical condition to receive care.

The program can read one scan in about six seconds.

Four critical conditions were selected to design the algorithm to detect hemorrhages, masses, hydrocephalus and strokes. Because strokes are traditionally very difficult to detect, a separate algorithm was developed.

Luciano Prevedello, a radiologist at Ohio State, led research on artificial intelligence programs that can speed up the process of analyzing patient scans. Credit: Courtesy of Ohio State

Richard White, chairman of the Department of Radiology at Ohio State, said the practicality of the artificial intelligence program has always been an area of importance, most recently with the screening of strokes.

“We really said, ‘How can we actually get to a patient in need quicker?’ by the virtue of this alerting us to the presence of something,” White said.

In addition to identifying one of the four abnormalities, the algorithm program also can prioritize health issues based on how important they are.

The researchers used 2,583 images from 246 patient scans to create the algorithms. Of those, 100 scans were tested to determine how critical a patient’s needs were. According to the research, one algorithm was over 90 percent accurate in recognizing brain abnormalities such as tumors. The other algorithm was over 80 percent accurate in detecting strokes.

The issue with current medical treatment is that many scans are marked as urgent by doctors who feel a patient’s condition could be critical. This creates a large pool of scans, all with similar levels of urgency. For scans to then be prioritized by severity, a doctor must be looking at the group of scans. That is why the complete accuracy of the algorithm is not entirely important, as long as the program is working in real-time to alert medical professionals, Prevedello said.

“The whole data set can say, ‘OK, based on the analysis there is a lot more urgent finding here,’ and it can grade that urgency,” he said. “And then we can use that information to restructure the workload in the way that the most critical findings go to the top of the list.”

White has been in charge of providing the support and securing financial and philanthropic aid for the project.

“I think this will become the identity for [the radiology] department,” White said. “It’s an opportunity for us to allow, for the first time in my recollection, Ohio State’s radiology program to jump out in the ranks, which is exciting.”

Prevedello said the artificial intelligence, although not free of error, is useful because of its consistency.

“Computers also don’t sleep, so that consistency and the fact that it’s always there and can be executed in almost real-time really improves patient care overall,” he said. “It’s like having another set of eyes on your studies and making sure nothing will fall through the cracks.”

Artificial intelligence within the medical field is not a concept out of reach anymore with advances in technology happening each day.

“The software that we’ve developed might not be immediately implemented, but the concept of having artificial intelligence in our practices is not entirely new.” Prevedello said. “It’s just getting more sophisticated.”