The AI healthcare revolution

Our researchers in The London Medical Imaging and AI Centre for Value-Based Healthcare have been working on many uses for artificial intelligence (AI) in health and are pushing boundaries of technological advancement in imaging equipment and the capabilities of modern software, resulting in both improvements in care and the wealth of data.

Making earlier and more accurate diagnoses

Medical imaging has been an integral part of healthcare for decades due to its ability to offer non-invasive ways of diagnosing and monitoring patients for signs of disease.

Each hospital Trust hold millions of these images – Guy’s and St Thomas’ NHS Foundation Trust holds 5.5 million – but AI can be used as a decision support tool for the radiology teams, who currently have to analyse each record manually.

Through the use of neural networks, algorithms can be trained to process thousands of imaging files to a strict classification system. In addition, as the algorithms process the visual information at pixel scale, they can flag very incremental differences to volume and structure which are not necessarily visible to the naked eye.

Indeed, in work published earlier this year, teams from King’s & Guy’s and St Thomas’ NHS Foundation Trust cut the reporting time for X-rays to receive expert radiologist opinion, from 11 to less than 3 days.

The same concept is also being applied to antenatal ultrasound scans, where AI can be used to identify abnormalities in the fetal heart and lungs.

Improving surgical outcomes

Machine learning, another subset of AI, has been shown to be highly effective in the operating room through the creation of 3D computer models of patients’ own organs, which mean surgeons can plan more effectively for complex surgery. When combined with advances in image and sensor guidance technologies these same models can be used in real-time during the procedure itself.

A team that includes researchers from the School of Biomedical Engineering & Imaging Sciences are using this to improve cardiac resynchronisation therapy – a common procedure for patients with heart failure. Success rates are highly patient-specific and the software helps to find the best plan for each procedure.  

The EpiNav project, a collaboration between King’s, UCL and the National Hospital for Neurology and Neurosurgery (NHNN,) have used a similar concept for neurosurgery that treats a certain type of epilepsy. Here, it is used to create a 3D map of the patient's brain that helps surgeons calculate the routes to take that reduce the risks of damaging the healthy parts of the brain. Already, the EpiNav software has been used in over 150 epilepsy surgeries at NHNN.

Using hospital data to target resources more effectively

In addition to specific clinical challenges, King’s is also looking at the potential of AI across the whole patient pathway; from admission in A&E, diagnostic imaging scans, interventions and aftercare through to discharge.

The TOHETI programme has been working in similar areas with the potential for change. One such part of this, is work looking at combining CT scanning and rapid AI reporting to diagnose patients with acute chest pain. This is something that could save time and money and lead to a much more efficient triage system and reassurance for patients.

The new London Medical Imaging and AI Centre for Value-based Healthcare aims to look at whether systems like this could be used across 12 patient pathways including stroke, heart failure and breast cancer using a powerful new data infrastructure.

By using the data processing potential of AI on such a large scale, researchers will be able to monitor how patients move through each stage of triage. Over time this will help them gain new insight for improvements in care and eventually offer predictive guidance for hospitals to optimise resources.