Over 400 New Jersey Auctions End Tomorrow 04/25 - Bid Now Over 1650 Total Lots Up For Auction at Four Locations - MA 04/30, NJ Cleansweep 05/02, TX 05/06, NJ 05/08
OXFORD, 21 September 2017 – Mirada Medical Ltd, a leading global medical imaging brand, is announcing a breakthrough technology at ASTRO 2017 by revealing the results of its research into the use of deep learning to develop next generation autocontouring solutions. The research has illustrated that contours drawn using deep learning trained software are mostly indistinguishable from human-drawn contours. They represent a substantial improvement on conventional autocontouring technologies for Radiation Treatment Planning available in the market at present.
Adapting the classic Turing Test*, Mirada has conducted a blind survey of oncologists and dosimetrists, asking them to identify whether a contour was drawn by a human or a computer. While current state-of-the-art atlas-based autocontouring was easily identified as computer generated, Mirada's new Deep Learning Contouring (DLC) technology performed substantially better and its contours were found to be mostly indistinguishable from human-drawn contours. When asked to assess whether contours were suitable for treatment planning purposes, clinicians found the Deep Learning Contours to be suitable as often as they did human-drawn contours.
To enable clinicians to experience this for themselves, Mirada Medical is challenging ASTRO attendees to participate in its Turing Test and attempt to identify which contours have been created by the new technology and which by industry professionals.
Hugh Bettesworth, Chief Executive, at Mirada Medical explains:
"We have taken advantage of cutting edge AI technology to address one of the everyday clinical needs in Radiation Oncology. Autocontouring technology that closely imitates clinical experts can deliver consistently high levels of contouring quality while potentially saving time and resources. The Turing Test is a natural way to validate this approach and, as part of it, we've been challenging radiation oncologists and clinical experts to see if they can tell the difference".
"Our new technology is set to revolutionize autocontouring and, while we have already undertaken a wide range of research and validation of our own, we don't want the attendees of ASTRO 2017 to take our word for it - we invite them to take our Turing Test".
The Maastricht Radiation Oncology (Maastro) clinic in the Netherlands has already formed a productive collaboration with Mirada Medical. Andre Dekker, Head of Department of Information & Service comments: "The contours are the closest to clinically acceptable contours we've seen from any autocontouring system we have evaluated. For some organs, our clinicians found it very hard to distinguish between their own contours and those that were automatically generated".
Professor Wouter van Elmpt, Ph.D., Assistant Professor and Senior Scientist, also at the Maastro clinic, adds: "We've been excited to be collaborating with Mirada Medical on this project. After 12 months' work interrogating the software and testing its capabilities, we are clear that this kind of deep learning technology will probably change the way autocontouring is done in future".
*Mirada Medical has applied a modified Turing Test to autocontouring technology. This is the first time the Turing Test, a classic test of AI capability, has been applied in this domain.
