Artificial intelligence and medical imaging: the potential to transform healthcare

The Academy, together with the Japan Society for the Promotion of Science (JSPS), recently held a one-day symposium on artificial intellingence and medical imaging. The co-Chairs of the event, Professor David Hawkes and Professor Kensaku Mori, offer some reflections of the day below. 

 

Artificial Intelligence (AI) and medical imaging are exciting areas of medical science and technology, and the advances being made at the interface of these fields have the potential to transform healthcare delivery.

We were therefore delighted to have been asked to co-Chair a recent symposium on the application of AI to medical imaging, held by the Academy of Medical Sciences (AMS) and the Japan Society for the Promotion of Science (JSPS).

The one-day event brought together a group of experts to discuss the current research landscape, and explore what the future impacts for healthcare might be as a result of the advances being made in the application of AI algorithms to medical imaging research. Sharing current status and difficulties in medical imaging AI will be a key precursor to the establishment of collaboration between UK and Japanese researchers. We heard from eight researchers working on applications as varied as the use of AI in fetal scans to neuroimaging as a biomarker for mental illness. An overview of each talk can be found below, and you can view the PowerPoint slides on our dedicated policy page.  

These conversations led us to reflect that there may be times where our roles as researchers should extend to advocacy and communication – as those involved in the development and application of AI, we are well placed to explain the technology, how it works, and what the limitations may be. We also need to listen to the concerns of the public and address these in the way we design and build these technologies.

Another helpful focus was that to increase public understanding and trustworthiness of AI for patient use, we must not forget that the research and innovations based on it should have clear clinical utility. It was suggested that while there are many fascinating algorithms being developed, not all of them have a clearly identifiable need in the healthcare system. In order to optimise the use of AI in healthcare, we should be doing patient-centered research to ensure that the technology is in line with actual need. An understanding of the ultimate value could come from working together with multiple organisations in this field, and we’re delighted this symposium therefore also provided an opportunity to strengthen – and hopefully catalyse - collaborations between UK-based and Japan-based researchers.

The discussions also recognized that it is not just patients and the public that are apprehensive about AI, but also healthcare professionals who might question how their roles will change as AI applications become more mainstream. Although it is popular to imagine a future where artificial intelligence has wholly replaced healthcare workers, attendees were more measured and agreed that such a future is unlikely - healthcare providers have compassion and wider expertise, and patients will still want to relate to another person during diagnosis and treatment. What we really need to be asking is what skills clinicians and other healthcare providers need in the future to help facilitate the effective and appropriate translation of AI into clinical practice to improve patient care.

Currently, this field is being driven technologically by the big technology companies, but as many of those who attended the symposium work in universities, there was a discussion about the roles academia, healthcare, small tech companies, and global medical engineering companies can also have. Universities and healthcare systems are well placed to foster new innovation due to the large amounts of data that they have access to. On the other hand, big companies have a role to play in providing secure computer platforms for deploying AI and machine learning. Recognizing the role of each sector there was a feeling that relationships should be built not only between researchers, but also with industry and other sectors to enable the technology to be pushed forward to create useful applications within healthcare.

An interesting perspective on this was given by Dr Pearse Keane, who presented the successful collaboration he led between Moorfields Eye Hospital and DeepMind Health, which aims to enable eye health professionals to increase the accuracy and speed of the diagnosis of the most common eye conditions when reviewing scans. However, despite the real world application, Dr Keane highlighted that even with full institutional support, it took a long time to get the correct arrangements in place to able to share the data. This is a clearly a barrier that needs to be overcome.

It was a delight to hear more about the exciting research currently taking place and we are encouraged that this symposium highlighted the breadth of shared interests between the UK and Japan, hopefully creating some opportunities to work together in this area. This one day symposium has enabled UK and Japanese researchers to understand the current achievement and the issues that should be solved in medical imaging AI in the future. These include not only technical issues but also regulatory hurdles and ethics.

Although this was only a one day event there were many ideas generated to enable closer UK-Japan collaboration in medical AI, and we hope that sharing ideas and solutions will continue to stimulate collaborations between the UK and Japan.

Overview of presentations

The day started with a series of scientific presentations from both Japan-based and UK-based researchers. Following an introduction and brief history of AI by Professor Hawkes FMedSci, the morning’s talks focused on how AI applications – including algorithms and machine learning approaches - can support healthcare provision and improve patient care. First, Professor Daniel Rueckert discussed the potential applications of AI in foetal scans, showing how it could help in the training of sonographers, create automatic checklists of visited planes in the image, and reduce the variability between operators. Ultimately, all such applications will have tangible benefits in ensuring that such scans are performed accurately and consistently. Next, Professor Kensaku Mori outlined the many ways that medical procedures can be assisted using machine learning - from diagnosis and detection to surgery aid. Professor Mori also went on to discuss what challenges need to be considered to progress developments in this field.

Dr Miyuki Uematsu demonstrated how AI could assist surgeons, using the case study of a navigation system to support surgeons during graft replacement for aortic aneurysm and preventing complications after surgery.

We also heard from Professor Hideaki Haneishi about how an interesting method - Robust Principal Component Analysis – can be used as an image-processing tool to reduce artefacts in medical images. Having algorithms such as this support healthcare professionals to interpret images, is likely to improve accuracy and efficiency, and in doing so could be cost-saving and shorten the time to diagnosis, with an overall positive outcome of patient experience and care.

Later, Professor Janaina Mourao-Miranda presented her fascinating work investigating neuroimaging as a potential biomarker for mental health disorders and how machine learning could be used to predict these disorders. In the future she would like to combine her imaging work with other information such as genetics and life-style parameters to see if this increases the accuracy of predictions, potentially leading to personalised medicine and treatment stratification – a potentially huge step-change in how we currently diagnose and treat mental health conditions.

Dr Jorge Cardoso also reflected on the added value of combining various sources of data, noting that large imaging data sets, integrated with clinical reporting data, can be used to streamline clinical workflow, improve the quality of care, and help make progress towards long-discussed ambitions for precision medicine. Demonstrating the incredible impact this work could have, he gave the example of an automated system for analysing brain scans that provides the clinician with an easily interpreted report that can aid dementia diagnosis.

While the morning’s scientific presentations highlighted the exciting value that can be afforded by applying AI to medical imaging, such emerging technologies inevitably raise important ethical questions, and ultimately regulatory issues. As such, the afternoon talks and panel session focused on patient concerns, and how best to integrate AI into existing healthcare systems.

During his presentation, Professor Makoto Hashizume was the first to suggest that it is perhaps the ‘black-box’ of the processing (i.e. ‘the unknown’), that worries the public and patients most – both in terms of how algorithms work and their accuracy, but also in terms of how their data is being used in the large data sets required to develop and optimise future AI applications. In response, it was suggested that the research community needs to be better engaged with patients and the public to explain how AI works in order to increase public trust and confidence.

Professor David Hawkes FMedSci is Founder and Director of the Centre for Medical Image Computing at University College London, and Professor Kensaku Mori is Director of the Information Technology Centre at Nagoya University. Professor Hawkes and Professor Mori co-chaired the symposium.

To find out more about this symposium, a previous news article can be found here and the presentations can be downloaded from our dedicated policy page.

Key contacts


Mentoring Masterclass - April 2024

Book Now

Clinical Academics in Training Annual Conference 2024

Book Now

Mentoring Masterclass - June 2024

Book Now
View more

Mentoring Masterclass - April 2024

Book Now

Clinical Academics in Training Annual Conference 2024

Book Now

Mentoring Masterclass - June 2024

Book Now
View more
 
 
 
 
 
 
FB Twitter Instagram Youtube