“It is impressive the amount of effort and focus that has already been invested in research into improving image acquisition and optimization - I would say, without any question in my mind, that Canon Medical is an industry leader in this specific area,” he remarked.

In parallel with its Automation Platform (AP), Canon Medical has developed the ^AUTOStroke solution*, a suite of AI-based applications designed to streamline stroke-related workflow. By automatically consolidating results into a single summary and alerting for abnormalities, the AP helps to support fast triage and tailored treatment decisions. Powered by Altivity, ^AUTOStroke* helps the specialist swiftly analyze and categorize images, and detect the signs of ischemic and hemorrhagic stroke in minutes. This ‘one-stop’ solution provides access to information required to administer life-saving treatment for patients.

Dr. Chang has worked closely with the Canon Medical team in the development of the new tool.

“In the beginning, it was a challenge and there were many things that we had to solve and think critically about, but over the months and the years we've seen ^AUTOStroke* evolve into a very practical and useful tool in our day-to-day work," he said.

En route to development of the new tool, Canon Medical supported Dr. Chang’s award-winning research into a deep learning AI system for hemorrhage detection in non-contrast head CT^1,2.

“The study is something we’re proud of, and the technology has been integrated into the stroke triage pipeline in the ^AUTOStroke solution* at Canon Medical,” said Dr. Chang.

Now, ^AUTOStroke* is used at UCI to automatically triage and evaluate every patient with suspected stroke that comes to the emergency room.

“It really helps us in our workflow to augment and maximize the time that we, as physicians, have to focus on and take care of patients,” added Dr. Chang.

As a clinician, Dr. Chang’s priority has been to ensure that the eventual tool offers true value in clinical practice. According to him, performance across two metrics is important: time and accuracy.

“Firstly, we want to make sure that the algorithm can actually get us the result in a time-efficient manner. Fortunately, AI systems that leverage deep learning technology are very fast. Almost invariably, a prediction can be made within 30 seconds to a minute of receiving images from the scanner,” he explained. “In several of our own internal studies, we've seen that the interpretation of hemorrhage requires on average 21 seconds per case and the detection of large vessel occlusion requires just over 35 seconds per case. That was an important key metric to validate from day one to confirm that the responses generated by the AI system are, in fact, speedy.”

“Secondly, we wanted to ensure that the algorithm is accurate. Not only must the AI interpretate exams quickly, but the response itself must be something that you as a clinician can trust,” he continued. “And in multiple internal and external benchmarks, we've seen that this is also true --- the performance of hemorrhage detection has routinely exceeded 95% in accuracy, and the detection of large vessel occlusion has been accurate in over 97% of cases. These are benchmarks are key to help facilitate widespread adoption. Importantly, it helps the clinician gain trust in ensuring that the AI system is a tool that they can work and interact with in a very reliable and robust way.”

In stroke, the ubiquitous phrase ‘time is brain’ dominates developments in new diagnostic solutions.

“While we have approximately 90 minutes to treat a patient with stroke, up to 30 minutes of that time might be spent on imaging alone,” noted Dr, Chang. “An AI solution like ^AUTOStroke* enables rapid interpretation within half a minute from when images are acquired, facilitating rapid and robust treatment and in turnaround time in ways that would be very difficult without that type of system.”

1. https://www.researchgate.net/publication/351141208_Validation_of_a_Deep_Learning_Tool_in_the_Detection_of_Intracranial_Hemorrhage_and_Large_Vessel_Occlusion
2. https://us.medical.canon/news/press-releases/2018/06/07/2946/