Medical Machine Learning
Medical machine learning involves the development of new algorithms and models which can then interpret medical data and improve clinical diagnosis and prognosis.
Currently, we apply medical machine learning to the following healthcare areas:
- Cancer (colorectal and breast cancer)
- Obstetrics and gynaecology
- Orthopaedics (arthroscopy, hip replacement)
- Neurology (transient ischaemic attack, stroke, vascular dementia)
- Public health.
We are rapidly expanding into other areas, check back soon for more details.
Machine learning and healthcare
Machine learning has been applied to the analysis of medical data for many years. Until recently, this was usually based on carefully handcrafted algorithms that performed limited, well-described tasks in imitation of the human process for analysing data, in particular medical images.
Medical machine learning focuses on how computers learn from medical data, and is at the intersection of statistics, computer science and medicine. The marriage between the different fields is driven by the unique computational challenges of building statistical models from massive data sets, which can include billions or trillions of data points. Using these datasets, we can not only learn those tasks that healthcare professionals can already do well, but also learning those where physicians have had only limited success.
Medical machine learning algorithms are likely to comprehend an image or a complex medical dataset as well as humans and potentially lead to improved or new diagnoses. This is driven by the fact that many current systems assume that each condition exists in isolation and is identifiable in one data source. We believe that most conditions are only visible by bringing together multiple sources of information or specific patterns in this information.
The AIML is connecting with the medical community, since we have an unparalleled opportunity to use complex data to develop solutions that help healthcare professionals and patients. Using world-leading expertise within AIML, we will focus on diagnosis, prognosis and treatment monitoring, and also on lowering healthcare costs. Our successes in commercialisation reflects how our technology could have a rapid impact on medical applications. In fact, this happens much quicker than medical devices. We are also aiming to tackle the biggest problems in methods in medical machine learning by creating tools to curate and annotate data, by making annotation of data more efficient or needed, and by creating tools that could interact with data and algorithms
Medical machine learning can be applied to the following diseases: cardiology, colorectal cancer screening, obstetrics and gynaecology, arthroscopy, public health, transient ischaemic attack, stroke, vascular dementia, and breast cancer screening.
Our current projects and areas of work include:
AIML and SAHMRI
The AIML has partnered with SAHMRI to explore how machine learning can be applied to methods and diseases that are being investigated at SAHMRI. The current focus is on image analysis in cardiovascular problems and neurological diseases, such as transient ischaemic attack, stroke and vascular dementia.
We are working together to develop new methods to improve diagnosis accuracy with the use of medical images, -omics datasets and patient demographics.
By working in conjunction with other institutions, AIML can use data to improve research outcomes with a focus on improving patient outcomes.
How to partner with us
- Professor Anton van den Hengel (Director, Centre for Augmented Reasoning)
- Professor Gustavo Carneiro (Director, Medical ML)
- Dr Johan Verjans (Senior Lecturer, Deputy Director Medical ML)
- Professor Mark Jenkinson
- Dr Qi Wu
- Dr Lingqiao Liu
- Professor Javen Shi
- Renato Hermoza
- Yu Tian
AIML medical machine learning experts Dr Johan Verjans and Dr Lauren Okaden-Rayner have co-authored chapters in Artificial Intelligence in Medical Imaging, which is now free to download from Springer Publishing.
The book explores opportunities, applications and risks of AI in medical imaging.