Boosting Acute Stroke Care Through AI-Supported Technology

Shiga University of Medical Science Hospital is a 603-bed core hospital located in Shiga Prefecture, Japan. It provides a wide range of highly advanced medical services, including emergency and disaster medical care. The hospital has established a 24-hour, 365-day consultation and treatment system for stroke patients, which has been certified by Japan’s Stroke Society as a Primary Stroke Center (PSC), and a Core PSC Facility. It has also been selected by the Society for certification to provide endovascular mechanical thrombectomy therapy 24-hours a day, 365 days a year.

It has worked together with Canon Medical in evaluating its new technologies and techniques over many years. We interviewed Professor Yoshiyuki Watanabe of the Department of Radiology to find out more about acute stroke treatment at the hospital and to gather his feedback on Abierto Reading Support Solution, which is currently in clinical operation at the hospital.

Abierto Reading Support Solution consists of the Abierto Automation Platform, which uses AI to automate reception and transmission of image data and execution of various applications, and Findings Workflow, which supports centralized management of analysis and results.

“In the diagnosis of acute stroke, it's essential to minimize delay and obtain the images needed for treatment planning quickly,” said Prof. Watanabe. “There have been great advances in the treatment of acute stroke. Our hospital has developed its own treatment protocols in accordance with the guidelines of Japan’s Stroke Society.”

Could you describe the clinical workflow for acute stroke patients at your facility?
When a patient is brought to our hospital, we conduct a brief medical examination and immediately perform diagnostic imaging. For patients with suspected acute stroke within six-hours of onset, we first perform a simple CT scan of the head, neck, and chest. The chest scan is used to check whether the patient has a pacemaker or an aortic dissection, which is a contraindication for rt-PA treatment. If Medium vessel occlusion (MeVO)/large vessel occlusion (LVO) is suspected, CTA is also performed. If the results show no intracranial bleeding and the time window is within four-and-a-half hours of onset, rt-PA is administered according to the guidelines.

If the time window is within four-and-a-half- to six-hours of onset, the ASPECTS score is used to determine if Endovascular treatment (EVT) should be performed, taking into account the extent of early ischemic changes, the occlusion status of the cerebral vessels as determined by CTA or MRA, and the patient's age.

If the time window is estimated to be within six- to 24-hours of onset, a plain chest X-ray is obtained to identify any metal in the patient's body, followed by MRI of the head. If DWI/FLAIR mismatch is confirmed by MRI, we assume that the patient is within four-and-a-half hours of onset and that rt-PA administration is indicated.

The same is true for patients with an unknown time of onset, which make up roughly 30% of the patients brought to our hospital. If DWI/FLAIR mismatch is not seen, the clinical-DWI mismatch and DWI ASPECTS score are evaluated to determine if EVT should be performed.

Could you tell us about examination times and therapeutic outcomes?
As for the time spent in the examination room, a simple CT scan takes about ten minutes, with an additional 20 minutes for contrast-enhanced CT and 20-30 minutes for MRI. Our CT and MRI systems are in the same area, so even when both examinations are needed, we can proceed to treatment within one hour from the time of arrival. At our hospital, the percentage of patients who receive rt-PA treatment is 16.7%, and the percentage of patients who undergo acute recanalization therapy (including angiographic procedures) is 33%. Overall, approximately 70% of our patients show an improvement in symptoms.

Why did you decide to start using Abierto Reading Support Solution?

When we were considering updating our PACS, we learned about Abierto Reading Support Solution and decided to install the system in 2021 because we thought it would provide valuable support to the physicians who interpret the image data. In the imaging of acute stroke patients, the important points are to identify patients with hemorrhage, to identify patients with extensive early ischemic changes, to evaluate occlusion of the main arteries, to assess areas of hypoperfusion using perfusion imaging, to identify mismatch between ischemic core volume and neurological symptoms by DWI, and to estimate the time since onset.

The Abierto Reading Support Solution includes useful applications that address all these points. Currently, Abierto Reading Support Solution for Neuro (‘Hemorrhage Analysis’, ‘Ischemia Analysis’, ‘Brain Vessel Occlusion’, and ‘Brain Perfusion’) are being introduced to support diagnosis and the interpretation of CT scans of the brain and Central Nervous System.

What do you see as the benefits of Abierto Reading Support Solution?
With Abierto Reading Support Solution, all we need to do is transfer the data acquired by a modality to an application server known as the Automation Platform, which identifies the images, starts up the appropriate application, performs analysis, and sends the analysis results to the PACS. All of this is done automatically.

In addition, the Automation Platform is an on-site platform, with the server installed in the hospital. So there is no need to upload data externally, as is the case for cloud-based applications. This helps to ensure security.

What are your thoughts on the accuracy of analysis?
After the introduction of Abierto Reading Support Solution, we conducted evaluations of hemorrhage analysis in simple CT head scans at our hospital and verified its diagnostic accuracy.

First, we reviewed 48 cases of intracerebral hemorrhage (ICH) among 142 cases diagnosed with ICH based on head CT performed at our hospital from January to December 2020. The detection results of Abierto Reading Support Solution were compared against the detection results of the diagnostic radiologists based on visual assessment. Abierto Reading Support Solution detected high-absorption areas in 44 of the 48 cases, for a detection rate of 91.4%.

We then compared the detection results of Abierto Reading Support Solution (Hemorrhage Analysis) against those of the diagnostic radiologists for all simple head scans performed at our hospital from May to August 2021. The results showed a sensitivity of 0.897 and a specificity of 0.896.

Could you tell us what your expectations are for Abierto Reading Support Solution?
We're currently preparing for the application of 4D-CTA combined with perfusion analysis using 320-row area detector CT (ADCT) for the diagnosis of stroke. If it becomes possible to accurately evaluate intracranial perfusion, we may be able to establish a CT-first diagnostic approach that's applicable to a variety of clinical situations, which would simplify our diagnostic workflow.

A new emergency center is scheduled to be completed in about three years, and a new CT system will be installed there. If it becomes possible to make the diagnosis with CT alone, the entire diagnostic process could be completed at the center, reducing the burden on both patients and medical staff and enabling more efficient examinations by reducing the total time required.

How is MRI used for the diagnosis of acute stroke and how can AI provide support in this?
MRI has better tissue contrast than CT. For example, DWI provides far superior visualization for detecting the presence of cerebral infarction. It's also considered to be more accurate than CT in identifying lesions referred to as ‘stroke mimics’, such as brain tumors. For these reasons, MRI is employed at many facilities in Japan for the diagnosis of cerebral infarction.

However, there are facilities in Japan at which stroke specialists and radiologists aren't always available, and at some facilities, diagnosis is performed by physicians who aren't specialists in stroke diagnosis. This is especially true for CT and MRI examinations performed at night. At such facilities, the use of AI-assisted diagnostic imaging would be a great benefit to both the diagnosing physicians and the patients.

Are there any particular challenges in the diagnosis of acute stroke using MRI?
A number of validation studies on the detection of early ischemic changes, including AI-assisted interpretation, are currently in progress. On the other hand, DWI/FLAIR mismatch determination, which is thought to be helpful for estimating the onset time of cerebral infarction, has not been discussed much, although it's expected to be useful in clinical practice. We feel that further research and discussion will lead to its appropriate application in the clinical setting.

DWI/FLAIR Measurement and DWI/PWI Mismatch can be installed in Abierto Reading Support Solution as applications to support MR image interpretation and diagnosis in patients with cerebral infarction. We're currently conducting joint research with Canon Medical Systems Corporation to evaluate these applications.

How do you see the future of AI imaging?
Another promising future application of AI will be to help predict the patient's prognosis after treatment. Current diagnostic imaging techniques suffer from the limitation of only being able to identify the presence of various lesions. Many other types of clinical information and imaging features have not yet been incorporated into the diagnostic process. Some examples would be: ‘post-treatment bleeding is likely when this pattern is observed’, ‘a good post-treatment prognosis can be expected for this pattern even though the ischemic area is extensive’, and so on.

Currently, treatment policies are determined mainly in accordance with established guidelines, with some degree of flexibility based on the clinical experience of the medical professionals involved. However, the ability to identify various imaging patterns using AI would allow the optimal treatment to be selected for each individual patient. This can be expected to improve treatment efficiency and help reduce medical costs. We believe that this will provide great benefits in many different aspects of healthcare.

Shiga University of Medical Science Hospital opened its Stroke Care Unit (SCU) in May 2021. A highly trained team comprised of neurosurgeons, neurologists, nurses, rehabilitation staff, and pharmacists specializing in stroke care, is available to provide intensive healthcare services 24 hours a day. The number of patients transferred to the hospital’s emergency room for suspected acute stroke was 178 in 2021 and 222 in 2022. For the diagnosis and treatment of these patients, three clinical departments (Emergency Medicine, Neurology, and Neurosurgery) work together to actively conduct thrombolytic therapy (tPA therapy) and catheter-based mechanical thrombectomy in the hyperacute stage as well as ‘drip and ship’ thrombolytic therapy, in cooperation with affiliated hospitals.