Altivity, Intelligent Healthcare Made Easy

Canon Medical works in close collaboration with some of the world’s leading experts in all modalities to develop relevant technologies. Innovations in Artificial Intelligence (AI) have become a key focus in this in recent years. The Satellite Lunch Symposium held at ECR 2022 provided some important insights into how the progress that Canon Medical has made in deep learning is already of benefit in clinical practice.

The symposium, which was entitled: ‘Altivity, Intelligent Healthcare Made Easy’, was moderated by Prof. Mathias Prokop. Prof. Prokop is a Radiologist and Chairman of the Departments of Radiology at Radboud University Medical Center in Nijmegen, and as well at the University Medical Center Groningen, in the Netherlands. Prof. Prokop is recognized the world over for his achievements in new imaging technologies. Following a short introduction by Prof. Prokop, the audience listened to three presentations from eminent speakers about some of the most innovative applications of Canon Medical’s deep learning technology in clinical practice.

“We have some exciting innovations in CT and also MR that leverage AI in clinical routine but are not what we usually expect when we hear about AI, namely that AI is being used to help find these diseases,” said Prof. Prokop. “Today, we are looking at something completely different, which is the use of AI to make images better, and in that way, help us to achieve the right diagnosis.”

“We wanted to find a way to get the best of our two high-end CT scanners (the Aquilion ONE and Aquilion Precision) together, without huge cost, so as to provide the best value to our customers,” he said. “We think the solution is a particular technique called Super Resolution. Super Resolution is a type of algorithm that has been worked on for decades, but with the introduction of deep learning in recent years, it has advanced a great deal. It has, for example, been successfully applied in satellite imaging and in natural image processing on cell phones. So we wanted to bring this technology to CT and see what it could do. That's how we came up with the idea of a Super Resolution Deep Learning. And our resultant product is called Precise IQ Engine (PIQE).”

To train any neural network three ingredients are required: A training target that represents the ideal case, training input that represents the current system, and the neural network that can do the job.

Canon Medical has invested in all three components to optimize the algorithm and its performance. For the training target, Canon Medical has had the unique advantage of being able to access high resolution clinical data from the Aquilion Precision, which was introduced five years ago. This represents a wealth of high-quality clinical cases at a very good dose to present what the ground truth will look like. To create a training input the image is reconstructed at the highest resolution achievable. Denoising is implemented after, to allow us to preserve the best resolution possible. The neural network used is three-dimensional neural network.

“In clinical cases, we have seen the benefits of PIQE, including improved visualization of stents, calcium, small vessels, aortic valves, and reduced blooming artifacts from calcium and stents,” remarked Dr. Yu. “All the doctors who have used PIQE agree that for the small structures that I have mentioned earlier, the conspicuity, the diagnostic confidence, has been greatly improved. I hope that this technology gets into the hands of more customers to enhance the care of patients.”

The Diagnostic Image Analysis Group (DIAG) of Radboud University Medical Center in Nijmegen, the Netherlands, has been involved in the development and implementation of advanced imaging techniques that feature AI-solutions, in collaboration with Canon Medical for some time. One key focus is on the use of advanced neuroimaging tools to improve detection, diagnosis and treatment of stroke. In acute stroke, imaging needs to be performed and interpreted immediately, so that the right treatment for each patient can be determined and implemented fast. Recently the Medical Center has evaluated Canon Medical’s Automation Platform. Dr. Anton Meijer, Neuroradiologist provided an update on the applications around stroke, and workflow in particular.

“Time is of essence especially in stroke imaging so optimize CT workflow is crucial,” emphasized Dr. Meijer. “There are different points in the workflow which can be optimized with deep learning and AI.”

He explained that AI can also support the technician in selecting the correct scanning protocol to have image reconstruction either for optimized image quality but also to reduce radiation dose and can also support the physician in making the correct diagnosis, in stroke, with AI tools to instruct imaging for detection of hemorrhage or large vessel occlusions. He recognized that it can also support specialists in treatment decision-making.

“It is really important that we collaborate with the software engineers because the algorithm must be well trained and optimized,” he added. “ We must have adequate reference standards that should not only rely on the density of the abnormalities, especially in the skull base.”

“In stroke, it is important to have quick processing of perfusion CT. Deep learning really speeds up our workflow to make our diagnosis quick and fast and it is expected that also this CT perfusion acquisition will improve with deep learning processing, especially if we take clinical characteristics into account to have a more accurate prediction of outcome,” he said.

However, Dr. Meijer recognizes the challenges for the tools in clinical practice.

“The AI tools used in clinical practice must serve the clinicians and the radiologists,” he reiterated. “The ultimate goal is to improve the outcome of the patient, but for this we need real life outcome measures in order to have an idea what the added value of such tools are in clinical practice and often those are lacking in general, as is study of the effectiveness and value in clinical practice.”

“AI can facilitate the workflow which is really important for stroke imaging it can aid detection of relevant pathology and treatment decision-making, but it is really important that you know your AI tool that you manage expectations, and also that you evaluate your AI tool and collaborate with your with your vendors,” he concluded.