Search This Blog

Friday, December 27, 2019

Medical AI Can Save Patients From Excessive Exposure To Radiation

AI’s potential in medicine has already attracted large amounts of media and public attention. However, some of the specific uses and consequences of AI in the context of health care aren’t particularly well known, at least not compared to awareness of the general utility of artificial intelligence.
Some of these uses may be life-saving, and in more ways than one. In recent months, a small number of companies and researchers have begun using AI for the purposes of medical imaging, harnessing machine learning algorithms in order to construct 3D models out of 2D images.
This may seem like a fairly standard employment of AI at first glance, but what’s special about this emerging use is that it’s focused on avoiding the need for MRI, CT and PET scans. In the case of CT and PET scans, this means that patients aren’t subjected to abnormal levels of radiation, which with CT scans at least can result in an increased risk of cancer.
One company offering this kind of technology is Zebra Medical Vision. Based in Israel, its main product is an AI-based “Imaging Analytics Engine” which can analyze medical scans in various formats and identify a wide variety of diseases. This month, it announced a partnership with orthopaedic and neurosurgery firm DePuy Synthes, through which it would bring its AI-powered analytics technology to orthopaedic images.
In particular, Zebra Medical Vision’s machine learning algorithms will be used within this partnership to create 3D models of patients from X-ray images. By doing this, the company hopes to provide cost-effective surgical planning to hospitals and surgeons worldwide, helping them to avoid or at least cut down on MRI or CT scans.
“The algorithms were trained on very large volumes of CT scans and the inference is from 2D images of the same patients. Zebra-Med’s strategic access to hundreds of millions of images enables these types of breakthroughs,” explains Eyal Gura, the firm’s co-founder and CEO.
Unsurprisingly, Zebra have used this announcement to highlight the effectiveness and versatility of its technology. That said, one of the most positive ramifications of using AI to create 3D medical images is that, by cutting down on the need for CT scans, it will save thousands (if not many more) patients from being subjected to relatively high does of radiation. In the long run, this could be a life saver for many people.
On average, a CT scan produces anything from 1 to 10 millisieverts (mSv) of radiation. To put this in some perspective, one chest CT scan delivers the same amount of radiation as 100 to 200 X-rays. This can leave patients with a 1 in 2,000 chance of developing cancer later in life. This is still a low probability, but compared to the chance of developing cancer after a low-dose X-ray–which has not been linked to an increased risk of cancer by studies–it’s certainly not negligible.
As such, Zebra Medical Vision’s use of machine learning to avoid unnecessary scans is a very positive development. And it’s not the only company working in this area. Munich-based startup ImFusion is another company leveraging AI, although in its case it’s transforming 2D ultrasound images into 3D models, rather than X-rays. This once again reduces the need for CT scans, while in ImFusion’s case it also provides surgeons with 3D models in real-time during surgery, something which can’t be done with MRI or CT scans.
There’s also plenty of research being conducted in the area of creating 3D models out of 2D images. Publishing their research in Nature last month, researchers at the University of California, Los Angeles have successfully used deep learning to create 3D models out of images obtained from microscopes. Likewise, in 2018 the Allen Institute announced that it had used AI to create 3D models of cells, potentially enabling doctors to track how cancer changes them.
The work above doesn’t specifically relate to creating replacements for CT and MRI scans. Nonetheless, it underlines that a growing body of research and development in this area is taking shape, one that may in the future largely eliminate the use of scans involving medium or higher doses of radiation. And when the day comes when AI is ubiquitously used to create 3D models and reduce our exposure to radiation, it will be providing a double boost to our collective health.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.