From AI-assisted imaging to AI-assisted diagnosis

Mark Hitchman, managing director of Canon Medical Systems UK explains why it is an exciting era of innovation and practical application of AI into UK diagnostic imaging services.

Artificial Intelligence (AI) in imaging is about solving many of the problems in the modern healthcare environment. This includes providing the tools to help clinicians make confident decisions faster when faced with growing backlogs; to help simplify workflows that can optimize staffing and equipment resource deployment; and about reducing the stress and exhaustion on health professionals.

AI in clinical diagnostic imaging practice began with the introduction of AI-assisted imaging, firstly through CT, and now MRI, using a Deep Learning reconstruction AI algorithm called Advanced intelligent Clear-IQ Engine (AiCE). It differentiates ‘noise’ from true signal to clean up images resulting in high-quality scans free from distortion. This helps to preserve edges, improve textures, and maintain details to assist with clearer clinical interpretation. It also reduces the need for image retakes and is at a much lower patient radiation dose than offered before.

It is being used by dozens of hospitals across the UK bringing benefits to patients. But AI has far more potential to demonstrate in imaging. More recently, we have started rolling out AI and automation applications for imaging that drill into specific disease conditions.

Prioritizing patient cases with AI-assisted diagnosis

Strokes occur once every five minutes in the UK and are the fourth leading single cause of death. Great results have been achieved in bringing these figures down over the years due to expanding research, medical innovations, and awareness campaigns but ‘time is brain’ is still incredibly important – the quicker clinicians can identify and treat a patient with stroke, the better the outcome from death or long-term neurological damage.

The automation of stroke diagnosis using AI in diagnostic imaging has the potential to streamline stroke-related workflow by automatically consolidating results into a single summary and alerting for abnormalities. Our triage tool AUTO Stroke can help to swiftly analyze and categorize images to detect signs of ischemic and haemorrhagic stroke in minutes. This has the potential to provide access to information that can speed up the administration of life-saving treatment.

UK based innovation in AI for the future of healthcare

Great strides are also being made on the visionary future of AI. Behind every deep-learning innovation there is lots of data. This is what builds the algorithms and feeds the machines with knowledge. Only a few years ago it was estimated that 97% of the 50 petabytes of data produced by hospitals per year went unused. The big data sources include patient medical records, CT or MRI scan images, pharmacy records, laboratory results and all the other sub-sectors of the healthcare ecosystem.

It is the decanting of this relevant healthcare data safely, which is pseudonymised to ensure protection of patients’ personal details, that is key to accelerating the pace of turning great AI ideas into daily clinical practice reality. Each AI application needs up to 100,000 data sets or even more to learn from along with a development process involving human clinical evaluation.

We are very proud to have a home-grown hub of AI research and development via Canon Medical Research Europe in Edinburgh. This group of software engineers and architects collaborates with 15 partners in the UK across academia, the NHS and industry via its Safe-Haven Artificial Intelligence Platform (SHAIP). This means, in essence, that the data being used to develop future AI innovations is gathered from UK specific data sources, making the development accurate and specific to our patient populations, at the same time as working closely with colleagues in Japan, China, Europe. , and the USA.

The team is responsible for building a set of services and web applications that include useful tools for clinicians to select and annotate patient data for machine learning, together with infrastructure for data scientists to develop, train and validate algorithms within the hospital environment. Once an algorithm has been created using SHAIP, it can be deployed into a Clinical Cockpit to allow effortless demonstration to clinicians. Many projects including stroke and Covid are fully underway and wider work in progress includes oncology, congenital heart disease, ultrasound, interventional angiography, and diagnostic support.

We all have a part to play in embracing AI’s potential

AI is only as good as the data it is built from, which means that collaboration is key. We are committed to consolidating machine learning and deep learning technologies and delivering dedicated, uncompromised quality and value across the entire care pathway.

We’re excited by the era of AI in the future of modern healthcare. From accepting AI’s place within imaging departments, to innovating future ideas and delivering market ready solutions. Every day we can all step forward to leverage data and transform it into essential imaging insights via AI.


[i] Brain Research UK, https://www.brainresearchuk.org.uk/neurological-conditions/stroke accessed January 2022.

.

Leave a Comment