Artificial Intelligence (AI) Research
- Artificial Intelligence Research
The Artificial Intelligence (AI) Research team is the foundation of the Canon Medical AI Centre of Excellence. It is made up of scientists, engineers, and clinical researchers, and hosts a thriving cohort of Canon-sponsored MSc, PhD and EngD students.
Canon Medical’s vision for AI in healthcare is the application of machine learning techniques to rich medical datasets to positively impact patient outcomes. They are developing algorithms that drive healthcare technologies, from diagnostic AI-enabled computer-aided detection (CADe) and computer-aided diagnosis (CADx), to predictive and ultimately prescriptive analytics. In pursuit of this vision, the AI Research team works closely with global academic partners and clinical collaborators, as well as Canon Medical Group colleagues in Edinburgh, Japan, Europe and the USA, in order to translate the AI state of the art into effective clinical decision support. Within Edinburgh, they partner predominately with the Clinical Cockpits, Image Analysis and Safe-Haven Artificial Intelligence Platform (SHAIP) teams.
To maintain their skills and research edge, AI Research scientists attend and present at training courses and international conferences in artificial intelligence, healthcare informatics, and clinical sciences.
Here’s what our employees have to say:
"I really enjoy working with motivated colleagues on challenging technical projects that contribute to the future of AI developments in healthcare. It’s especially rewarding to work at the interface of industry, academia and clinical practice. Through our involvement with ‘Industrial Centre for Artificial Intelligence Research for Digital Diagnostics (iCAIRD)’ www.gla.ac.uk/news/headline_620272_en.html, I’m looking forward to new opportunities to work directly with NHS partners within the hospital environment." - Alison, Senior Scientist
"As a recent graduate, interning in the AI Research team at Canon medical was an incredible opportunity to develop my skills in machine learning in a supportive, motivating and warm environment. I was given the freedom to work on meaningful and innovative state-of-the-art AI solutions. The experience has inspired me to pursue a career in AI." - Lucas, Intern
"As a team, we are incredibility fortunate to be working on technology that is truly transformative at a time when we are directly contributing to the coming of age of AI in healthcare. It’s hugely motivating and exciting." - Sandy, Technical Manager
See below the latest innovation from the Artificial Intelligence (AI) Research team
Fall detection from accelerometer data on wearable devices
This research  examined how to detect falls in elderly patients from accelerometer data collected with wearable devices, using different machine learning methods. There were ethical and practical difficulties associated with obtaining data from real incidents. Hence, we contrasted algorithms trained on simulated data from a younger volunteer population, with the detection of outliers after training on data taken from normal daily activities.
Figure 1: We thresholded the accelerometer signal at 1.6g, and then classified the identified peaks as “falls” v.s. “activities of daily living” such as sitting heavily or descending stairs. The figure on the left shows an example of each. The video on the right shows an example accelerometer signal, with 3 peaks identified and classified.
Anatomical landmark detection in CT volumes
This research  looked at the application of CNNs for efficient semantic parsing of a CT volume. The output is a set of key anatomical landmarks which we defined in an atlas, which we can then use in follow-on image analysis tasks such as visualisation, registration, segmentation and vessel tracking.
Figure 2: The figure on the left shows our set of 22 landmarks in the anatomical atlas. The video on the right shows an example CT scan (using the Visualisation team’s Global Illumination algorithm) with the detected landmarks. The landmarks are marked as magenta squares or cubes. Scan courtesy of Prof. Keith Muir, Queen Elizabeth University Hospital (Glasgow University).
 Lisowska, A., O'Neil, A. and Poole, I., 2018. Cross-cohort Evaluation of Machine Learning Approaches to Fall Detection from Accelerometer Data. In HEALTHINF (pp. 77-82).
 O’Neil, A.Q., Kascenas, A., Henry, J., Wyeth, D., Shepherd, M., Beveridge, E., Clunie, L., Sansom, C., Šeduikytė, E., Muir, K. and Poole, I., 2018, September. Attaining Human-Level Performance with Atlas Location Autocontext for Anatomical Landmark Detection in 3D CT Data. In European Conference on Computer Vision (pp. 470-484). Springer, Cham.