This site is optimised for modern browsers. For the best experience, please use Google Chrome, Mozilla Firefox, or Microsoft Edge.

Beta Site: This is a new site which is still under development. We welcome your feedback, which will help improve it.

Feedback form

Automated deep learning may allow doctors to greatly extend AI systems for healthcare

 
Automated deep learning may allow doctors to greatly extend AI systems for healthcare

Automated deep learning may allow medical professionals who aren’t experts in artificial intelligence (AI) to be able to build AI algorithms.

The ‘AI that can build AI’ could allow a much wider range of people to develop healthcare applications for AI, supporting earlier detection and treatment of disease.

Researchers at the NIHR Moorfields Biomedical Research Centre entered data from publicly available medical image datasets, including two types of eye scan, into the Google AutoML programme - a machine-learning algorithm that learns to build other machine-learning systems. 

The programme created an automated algorithm that was used to classify the diseases in the images using AI.
For simple classification tasks, they were able to demonstrate a diagnostic performance similar to that of many state-of-the-art AI systems.
However, further work will be required before these approaches can be applied in real world clinical practice. Writing in Lancet Digital Health, the researchers expressed the need for deep learning experts and clinicians to collaborate to ensure AI is used correctly in clinical practice and highlighted the need for regulatory guidelines in this area.
Pearse Keane, NIHR clinician scientist and Consultant Ophthalmologist at Moorfields Eye Hospital, said:
“At present, the development of AI systems requires highly specialised technical expertise. If this technology can be used more widely – in particular by healthcare professionals without computer programming experience – it will really speed up the development of these systems with the potential for significant patient benefits."
“While the derivation of classification models without requiring a deep understanding of the mathematical, statistical and programming principles is attractive, comparable performance to expertly-designed models is currently limited to more simple classification tasks. The process needs refining and regulation, but our results show promise for the future expansion of AI in medical diagnosis.”

Read more