Image Description: Human brain made out of coding.
We live in an era of rapidly evolving science and technology in which fairly new concepts are emerging, such as Big Data, Machine Learning (ML), Artificial Intelligence (AI), among other fancy terms. In this article, I am going to focus on one of these terms and discuss some of the world’s leading projects in ML/AI at Oxford.
ML is an emerging subset of AI in which programmers enable computers to ‘learn’ for themselves how to make decisions and carry out specific tasks based on algorithms and statistical modelling. ML doesn’t necessarily require explicit instructions to perform these tasks. Instead, ML relies on recognising patterns and features from data. The use of ML has been adopted in many subject areas ranging from biomedical sciences to astronomy. There are many applications for ML in biomedical sciences including healthcare diagnostics and disease prognosis, cancer research, bioinformatics and DNA sequencing.
Professor David Clifton of the University of Oxford said, “In healthcare, we have seen an astounding level of hype surrounding the use of AI, but there is real promise for helping people”. He added “Digital technology is now part of everyday life.”
There are many applications for ML in biomedical sciences including healthcare diagnostics and disease prognosis, cancer research, bioinformatics and DNA sequencing.
At the University of Oxford, scientists and researchers led by Professor Paul Leeson have been using ML to improve diagnostic accuracy of heart scans used to assess patients complaining of chest pain. The expected diagnostic accuracy will increase from 80% using conventional imaging to more than 90% using this new technology. It works by extracting more than 80,000 data points from a single echocardiogram image to overcome variability in detection abnormalities. With the introduction of ML to echocardiogram, nurses and doctors will be able to detect subtle changes in the image that cannot be seen by the naked eye. Therefore, it will avoid misdiagnoses of fatal causes of chest pain and save many patients’ lives. In addition to having a direct impact on patient health and outcome, this new technology will save the NHS millions of pounds by decreasing the number of further tests and procedures.
Similarly, Professor Charalambos Antoniades and his colleagues from Department of Cardiovascular Medicine has been using ML to analyse state-of-the-art coronary computed tomography (CT) images to predict the risk of having fatal heart attacks many years before they occur. The idea is to detect certain imaging biomarkers in the fatty tissue and degree of inflammation around the coronary arteries. In addition to predicting the occurrence of heart attacks, this new technology can help in predicting the risk of death from these heart attacks. The aim is to introduce this service to the standard CT scans in the NHS hospitals and other hospitals around the world.
On the other hand, Professor Alison Noble from the Department of Engineering Sciences has been working on improving the quality of ultrasound imaging in various clinical uses. In particular, she has been analysing ultrasound scans of pregnant women using ML to improve the detection of anomalies in fetuses.
At the Computational Health Informatics lab at Oxford, researchers have been using ML to provide useful information about critically ill patients admitted to the Intensive Care Unit (ICU). They were able to closely monitor vital signs of ICU patients using advanced sensors and analysing the resulting data to stratify patients according to their risk of death or getting serious adverse events.
At the Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences (NDORMS), Dr Sara Khalid from the epidemiology group applies ML methods to build clinical risk prediction models for musculoskeletal conditions. She said, “These models are designed to predict risk of a patient developing a certain condition over a certain period of time”. She went on to say “One of the potential advantages of ML methods is that they sometimes (although not always!) outperform traditional statistical methods, for instance when there is a large number of complex and inter-related features to choose for the prediction”.
Discover ML & AI at Oxford by visiting this hashtag: #OxfordAI