Outcomes driven innovation in stroke:

AUTOEmergency – Stroke CT Package Used in Clinical Practice at Gates Vascular Institute (GVI)

Prof. Kenneth V. Snyder MD, PhD, FAANS
Watch a YouTube video of the full discussion and see more topics in this webinar.
I will provide a brief update on the American heart / American stroke guideline criteria, building from Professor Chang’s article… Then I’ll talk about our experience in clinical practice with automated programs and algorithms. And I’ll finish with the incredible workflow opportunities gained by using Canon’s advanced applications and Stroke CT Package.

An update on American heart / American stroke guideline criteria

Very interestingly, our American heart / American stroke guideline criteria in the US had a modification to it a few years ago. The use of perfusion imaging in less than six hours is no longer a recommendation or requirement.

And the reason for that… This slide lists five randomized controlled trials less than six hours out from the onset of stroke. Trials that used perfusion for the selection of patients had the most benefit. But for the trials that did not have perfusion imaging and used CT scan alone, there was still a significant benefit based on the ASPECTS score. For that reason, we didn't want to rule out the ability for people to have the benefits of thrombectomy in less than six hours.
Randomized controlled trial results
We had been using perfusion imaging to select patients for thrombectomy since 2010. So it was hard for us to recognize the importance of doing a randomized controlled trial. A 20-year-old individual with an NIH score of 10 or higher and a viable brain on a perfusion study would have to be randomized, and that was an ethical dilemma for us. But two randomized trials did need to occur. With no surprise, the trial was stopped with tremendous benefit for the thrombectomy arm.

The imaging that was used in many of these trials used a custom platform that has changed its thresholding over the years. The Diffuse 3 trial that was stopped early with tremendous benefit used a core of less than 70cc with a penumbra-to-core mismatch ratio of greater than 1.2. This has become level one evidence in the United States.

Current state-of-the-art treatment can – if within four and a half hours – identify whether someone's a TPA candidate, identify if there's an LVO and then, using the DAWN and Diffuse criteria, enroll patients for thrombectomy. We're fortunate to be in a prospective study center where IRB allows us to study well beyond this range. As Dr. Chang mentioned, we have both RAPID AI and Viz AI available to us and we've been studying on Canon’s AI based Automation Platform as well. So with the same raw DICOM dataset, we're able to post-process with all vendors that are currently available and ask questions on thresholding as well.

Experience in clinical practice with automated programs and algorithms

Acute thrombectomy has radically changed our learning curve on what is core and what isn't. This was a 29-year-old NIH of 17 that showed up with a very large DWI region. We emergently took the patient for thrombectomy and this is the post-op image on the right side. The take-home message here is DWI imaging is not core imaging. The ATP pumps are shut down, and the cell may be swollen, but if you're able to recover that tissue you can get full resolution and that's where the perfusion imaging really helps us.

We took the Canon AI LVO algorithm, which was trained on 3,000 cases, and did our own accuracy assessments and studies on those and published those as well in our radiology journal. We also saw tremendous benefits with the Bayesian CTP algorithm, improvements in ventricle segmentation, thresholding improvements, summary map improvements, and an improved ability to pick the arterial and venous input function.

I would like to bring up a case that I think teaches us that we have so much to learn on about stroke and stroke imaging. This 28 year -year-old showed up, likely from a carotid dissection, just within four and a half hours with a high NIH. He was given tPA and had an ASPECTS score lower than most criteria for treatment. We saw sluggish flow in the carotid and an intracranial M1 occlusion. The cerebral blood volume score was extremely low. It would have been considered potential core based on all of those imaging criteria.
Algorithm improvement with Bayesian CTP
28 year old with likely carotid dissection
We brought this young patient in for a thrombectomy. The carotid head completely occluded itself. We went up with an aspiration catheter and opened up that carotid. There was a lot of thrombus remaining in the carotid. We then addressed the intracranial M1 occlusion and went up with aspiration there. The minute we opened up that vessel, the patient screamed out in pain and started vomiting. Of course, knowing that was potential core, we were worried this patient bled into their head. We intubated the patient, and did an on-table LCI image. No hemorrhage on the scan.

We put the patient back under angio and noticed that what before was open was now occluded again and throwing clot likely from the carotid. We gave the patient dual anti-platelets on the table and stented the carotid artery. We noticed that on the petrous end, the stenting was necessary right at the transition to the skull base, which required another wingspan stent. Then we went up intracranially and took a look. Within a few minutes, we were concerned about the wingspan stent filling with clots. We ended up seeing that they threw a clot and ended up with a superior trunk occlusion. We had to give two half doses of Integrillin to reopen that backup on this patient.

So here's a 28 year -year-old with what was likely dead tissue that got TPA, Heparin, Ticagrelor, aspirin and Integrillin. We really didn’t expect this patient to survive. The CT showed some basic angular staining the next day. He had an NIH of 4 and has been neurologically perfect since. I think there's a lot to learn as we're applying these algorithms.
Initially not expected to survive, this 28-year-old patient has been neurologically perfect since the day following treatment
Core infarct and hemorrhage

Workflow opportunities gained by Canon’s advanced applications and AUTOEmergency - Stroke CT Package

Currently, we use the CT perfusion algorithm to tailor our treatment to the patient. This is an example of a core infarct in the cerebellum and hemorrhage in the PCA territory with a basilar occlusion. In that case, we go up and do a coil occlusion of the PCA territory that's hemorrhaged and open the basilar back up.
We can't make the patient perfect, but we can make them as good as their perfusion scan. We're very excited about the entire workflow.
Flow diverter with high-def detector
A few months ago, we installed the 4DCT from Canon, allowing us to get perfusion imaging and angiography in the same suite. Our patients don't even come into the emergency room.
They come in through a stroke triage door right onto the scanner. If the scanner suggests a viable brain, we bring the team in and do the stroke treatment right there on the table.

This algorithm also allows me to do single vessel distal perfusion imaging to start to confirm what we think our physiologic parameters are on X-ray and actually measure them in three-dimensional space.

Canon’s entire AI workflow is built into this system. So we're able to take advantage of all the dose imaging reduction and imaging enhancements of the CT scanner, as well as take advantage of the dose reductions and angiography table in real real-time. This allows us to achieve extremely low radiation doses for our patients and staff.

We've co-built with the Canon team an FDA-approved high-resolution imaging detector that lets us see images at a resolution that most places aren't able to achieve in our flow diverters or stent retrievers

We are again publishing the ability to use angiography to develop those perfusion algorithms on table-automated TICI scoring and the ability to look at the CT scan and determine with its Hounsfield units what we're calling a perviousness or an ability of the dye to go through the clot and try and predict first pass or choice of the device due to what we expect the clot to potentially be.
Prof. Kenneth V. Snyder MD, PhD, FAANS
Neurosurgeon State University of New York Buffalo, New York, USA
Contact Us