ss

Adoption of AI Technologies in Diagnostics

Share

AXREM is the UK trade association representing the interests of suppliers of diagnostic medical imaging, radiotherapy, healthcare IT and care equipment in the UK. Our group is comprised of most of the industry supply companies, complemented by the services of a secretariat.

AXREM members supply the majority of diagnostic medical imaging and radiotherapy equipment installed in UK hospitals. In doing so, our member companies and their employees work side by side with Radiologists, Radiographers and Practitioners, Oncologists and a wide range of healthcare professionals in delivering healthcare to patients using our technologies. Our members therefore have unique knowledge, experience and insight into the workflow and challenges faced by healthcare professionals on a day-to-day basis, which enables us to develop and offer innovative solutions to improve the speed and quality of diagnostic procedures and treatments with our ultimate aim of improving patient care.

Professionals and patients are very well aware of the difficult situation that the NHS (and other healthcare providers) are in at the current time and the global scarcity of available healthcare professionals to cope with the demands being placed upon them. Many people have suggested AI and automation is one answer to this challenge, but there are many obstacles to introducing new technology into the diagnostic process which place substantial constraints on the speed of innovation.  Even given these constraints, the adoption of AI within the UK health and care system is likely to accelerate as providers look to rollout new innovations to help improve outcomes and resolve the increasing backlogs that have grown substantially during the COVID pandemic.

The purpose of the AXREM AI SFG is to promote the adoption of beneficial AI technology and to identify and seek to overcome the hurdles to adoption encountered by suppliers while attempting to develop and deploy AI solutions into both public and private providers in the UK.

Most recent AI solutions are based on a process of ‘deep learning’ in which an artificial neural network is presented with hundreds, thousands or even millions of annotated examples of a particular kind of pathology and gradually ‘learns’ to recognise what that pathology looks like. This process is very data dependent, and so the work of data scientists and curators is as important as the innovation coming from the AI scientists and software engineers who develop the actual algorithms.

A key part of developing safe and reliable AI solutions is the use of a representative data set while performing the AI training. There are many well publicised examples of where bias as been introduced during the training process leading to highly inaccurate results.  The answer to this problem is to use data drawn from the actual target population, which means that a generic AI solution needs to be trained on a representative subset of the entire population.  This objective seems daunting or even impossible, but it can be done using a network of ‘trusted research environments’ or TREs coupled to a technology called ‘federated learning’ that can facilitate building a generic model safely and effectively.

There is a lot of ‘hype’ and excitement in the market around the use of AI in medical imaging, much of it warranted some of it not. For example, the potential gains that clinicians (and managers) could expect from applying AI to the process of triaging a backlog of unread X-ray images to prioritise the work of the radiologist, are real and the technology has recently matured to a level at which that is a realistic task for AI. However, it’s not realistic to think that AI will ‘replace radiologist’ anytime soon as some disruptors in the field have claimed.

There are many examples emerging where AI could effectively ‘turbo charge’ radiologists in their day-to-day work allowing them to spend more time where their expertise is needed most. In more of a research environment, it’s now possible to use AI within clinical trials to speed up the laborious process of measuring tumour volumes during a drug trial for a potential cancer cure to track changes in the cancer pathology.

The adoption of new technology in healthcare can occasionally bypass, or even ignore, the needs or desires of the patient, but this cannot be allowed to happen with the use of AI in diagnostics. The AXREM AI SFG is committed to engaging with relevant stakeholders in the sector to ensure that when AI is adopted, it’s used appropriately and effectively which is the main reason why AXREM has recently adopted the Department of Health and Social Care’s guidance on good practice for digital and data-driven health technologies as our code of conduct.