DSC-RX100M3
ƒ/2
8.9 mm
1/30
125

Good stuff happening in the application of machine intelligence to radiology, resulting in superhuman performance out of the gate. Photo from my last visit to Imagen.AI with their Head of Machine Intelligence, Sumit Chopra, and cofounders Tom Hotchkiss and Alex Dresner.

Pioneering FDA approval of their diagnostic AI at the point of care: “Artificial intelligence algorithms have tremendous potential to help health care providers diagnose and treat medical conditions,” said Robert Ochs, PhD, deputy director for radiological health in the FDA’s Center for Devices and Radiological Health. “This software can help providers detect wrist fractures more quickly and aid in the diagnosis of fractures.”

Machine Intelligence for medical imaging is a canonical case of tech inevitability: is there any chance humans will read X-rays, MRI scans, pathology sections, or any visual diagnostic 50 years from now?

3 responses to “Imagen AI for Radiology”

  1. sbove Avatar
    sbove

    All depends on whether humans can keep discovering/learning new things that they have to teach the machines 😉 Long way to go in most areas of medicine… My father is a Sr. pediatric pathologist who reads ~thousands of slides and electron micrographs each year…He has been a med school prof since 30 yrs old and still conducts much primary research even while helping run a major clinical lab at Cincinnati Children’s Hospital. He is still learning like an 18 year old at 83+ (at a Liver convention in SF right now, actually). Perspective granted me through observing his world > I’d bet it will take 20 or 30 years for Sr. Pathologists to be completely displaced in complex visual diagnostic work for simple/common diseases. At the cutting edge and for more complex conditions, human primary researchers/diagnosticians will be required for many more decades. But AI diagnostics will play an increasingly large role at the microscope/X-ray/Radiograph-viewer/MRI/PETscan screen, if nothing else as a guide and aide (augmentative decision support) to med-students, new MDs, and even Sr. MDs. So this tech is massively valuable and important – especially outside of top research hospitals.

    Reply
  2. jurvetson Avatar
    jurvetson

    Very good point. The research will continue for some time, but the automation of superhuman routine performance will allow the learning to touch 10x as many people in need (think rural clinics in the developing world)

    Reply
  3. jurvetson Avatar
    jurvetson

    This just in… My former colleague Jared Dunnmon has just published new Deep Learning Radiology results…. go radiology superheroes!

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *