VIDEO: AI for stroke detection on CT imaging

"A very important aspect of radiological care now has been stroke detection and stroke intervention," Allen explained. "And a number of the AI software developers have tools that will analyze the perfusion data from the CT angiogram and make a fairly rapid determination as to who might benefit from a neuro-interventional procedure."

He said this technology is automated, reading imaging, as it comes off the scanner before a radiologist sees it, flags the exam and sends an automated alert to the stroke critical care team. The radiologist on the team can then make a final confirmation, and, if needed, the neurologist-interventional cath lab can be activated and the patient moved there for an immediate thrombectomy. 

However, the vendors offering this technology have marketed it to hospital administration, emergency departments and neurology departments, so radiologists are not always a primary consideration when the AI is adopted. According to Allen, it is important hospitals include radiologists as part of the on-call stroke care teams. Allen notes the U.S. Food and Drug Administration (FDA) cleared these technologies as a triage tool, not as a stand-alone, autonomous diagnostic technology. He said the AI is also not always correct, so it still requires a radiologist to make the final call on a diagnosis. 

"We worry about the risk that someone is getting pushed down the hall to the cath lab and they are going to end up with a negative angiogram because nobody waited for the radiologists interpretation," Allen said. 

He said it is important to include radiologists in the workflow and there is some sort of check off that the radiologist has reviewed the case.