United Kingdom and United States
An FDA-cleared algorithm analyses chest x-rays to quickly distinguish whether the image is normal or indicates an abnormality.
Behold.ai is applying artificial intelligence to medical imaging to distinguish normal x-rays from those with abnormalities accurately. Their FDA-cleared red dot algorithm utilises machine learning and was trained on over 30 000 chest x-rays.
They take an approach of “ruling out normal”: if the algorithm does not find an abnormality, the healthcare team can immediately move on to look for other causes of symptoms. Behold.ai estimates the timely, accurate analysis of ruling out normal could save the NHS over £100m per year.
The red dot platform integrates seamlessly with existing IT setups within hospitals. The algorithm will analyse the chest x-ray as soon as it’s taken. Using secure technology, the image is analysed on the platform. If there is an area of concern, it’s highlighted with a heatmap consisting of a red area. Regions of urgent concern can be flagged more quickly, and radiologists notified. Behold.ai is working with Wellbeing Software to fast track the red dot platform application in Covid-19 diagnoses.
“We continue to see more and more digital health startups utilising artificial intelligence for image analysis. Back in 2017, I spoke with Jeroen, co-founder of Aidence, about using deep learning in CT scan images for early lung cancer diagnosis. At this time, they had already inputted 45k annotated images into their algorithm.
With time, as these systems being more and more accurate, there is an opportunity to detect conditions more easily and support healthcare professionals and systems under increasing strain. Recently, we have seen AI image analysis used in heart scans, stroke and moles. 42.7 million imaging tests were undertaken on the NHS in England in 17/18, the opportunity for optimisation is huge.” – Dr Vinay Shankar