will Watson replace radiologists?

This forum made possible through the generous support of SDN members, donors, and sponsors. Thank you.

badasshairday

Vascular and Interventional Radiology
15+ Year Member
Joined
Apr 6, 2007
Messages
3,923
Reaction score
361
http://www.diagnosticimaging.com/radblog/display/article/113619/1806004

Interesting article. There is a lot of talk about incorporating AI into medical practice these days. What do you guys think?

Nobody can predict 50 years from now, but it reminds me of the times of robotic surgery which was very hot in early 2000s.

It may or may not replace radiology. But not in the foreseeable future, at least next 20-30 years or even more.

There is much more to medical imaging than this. A lot of attempts were made in 90s to quantify many imaging findings or standardize many findings, but it was not successful.

The only modality that we use computer for is CAD for breast imaging and even in that case, I have found in horrible. Nevertheless, I myself look at it on every case, but most breast imagers completely ignore it. In one word it is useless.

Back to medical imaging, if you want to do it right, the final impression is a combination of medical history (not available these days, you have to look it up in EMR), physical exam (not available), Labs and MOST important comparison of different modalities. There is more to it than just images. You will understand it clearly when working with Ultrasound techs. They are pretty good at finding the abnormality, but not making the final impression.

One aspect of these software designs is marketing. They design it and want to market it, so they claim that one day it may replace the most intellectual advanced aspect of medicine which is radiology.

Let me tell you sth: For many reasons people always talk about doom and gloom in radiology more than any other field. In 70s they were talking how endoscopy will replace Upper and lower GI barium studies so radiologists will be out of work. In 80s many were talking about how MR will replace CT. In 90s there were talk about turf issues and also over supply of radiologists. In early 2000s they were talking about outsourcing to India. In 2010s they are talking again about over supply and also tons of other BS.

When I was in training esp intern or first year resident I used to take some of these serious. They may happen one day, one day we may become jobless. But the likelihood is less than most other fields. You can say the same for many fields. For example once a medical student asked me that he wants to GI, but what if they replace screening colonoscopy with CTC.

Though the future may go in any direction, but more than 90% of assumptions in the last 40 years have been wrong. Nobody can predict the future. So o whatever you enjoy, whether DR, ER, IR, GS, NS or pathology.
 
Members don't see this ad :)
Image recognition is far behind diagnosing and treating based on algorithms. Take those annoying captchas for example, those are used to fool bots.
 

There is something definitely wrong with the way they did that mini-study, considering it was written 1998, and the last I saw computer EKG reads (2010), they were still wrong in obvious cases half the time.

In radiology, every study is inherently SO different from another with technique, anatomy (basic and iatrogenically altered), subQ fat differences, that I cannot see a computer algorithm accounting for everything. You would think that by now a software would have come out pointing out areas of opacity on routine chest x rays (which any 3rd year medical student could do). The fact that even this has taken so long to do shows that the machines are nowhere close
 
Top