Focus on stroke
Specifically, their research is focused on new technologies for stroke diagnosis through collaboration with the Industrial Centre for AI Research in Digital Diagnostics (iCAIRD). The objective is to reduce ‘door-to-needle’ time – the time elapsed between admission and treatment in ischemic stroke. In this condition, a blockage in blood flow reduces oxygen supply to brain tissue. The sooner the clot or occlusion can be removed, the faster blood flow to deprived regions can be restored and the better the chances of recovery. Current clinical guidelines indicate that if four-and-a-half hours lapses before treatment for an ischemic stroke, it fails to have an impact – so door-to-needle time is crucial.
Elements that can slow down time to specific diagnosis and treatment include patient transfer and language barriers, as well as painstakingly reading through many medical notes. NLP works on the basis of algorithms that identify, extract and convert human language into an analyzable format. It is already used in applications, such as personal computer assistants. In the context of medical applications, the specialized language, which is often full of idiosyncrasies, such as specific abbreviations as well as misspellings and typos (from hastily recording information in urgent situations), and confidentiality of patent records mean that AI applications require special adaptations and training to be effective.
Canon Medical’s AI scientists have already created a state-of-the-art algorithm for classifying medical reports and have partnered with Prof. Sotirios Tsaftaris of Edinburgh University (Canon Medical/Royal Academy of Engineering Research Chair in Healthcare AI) to develop a feasible model. This includes a breakthrough algorithm that can incorporate medical knowledge – in this case, information from the International Classification of Diseases, into the design of the model to obtain higher accuracy.
The AI model under development aims to ‘turbo charge clinicians’, and make their job as easy as possible by supporting their decision-making processes. They will be able to specifically source the AI algorithm output, providing reassurance and context. For example, when they click on a sentence it could show exactly where the information was derived from, the clinician can either agree or disagree with it. It will present the facts, but the clinician still makes the decisions.
Over the next year, the team will continue their research and further focus on text algorithms that specifically support clinical decisions in the treatment of acute stroke.
Please click HERE
if you would like to read the complete article.