Started By
Message

re: Which medical field is most resistance to automation and AI?

Posted on 1/31/19 at 3:25 pm to
Posted by Gaston
Dirty Coast
Member since Aug 2008
41163 posts
Posted on 1/31/19 at 3:25 pm to
The AI better be the Dr too, because good luck getting the oncologist, or whoever, to understand how the AI describes the case.
Posted by greenwave
Member since Oct 2011
3878 posts
Posted on 1/31/19 at 3:29 pm to
quote:

OB/GYN


Males are vanishing in this field.
Posted by flyAU
Scottsdale
Member since Dec 2010
24899 posts
Posted on 1/31/19 at 3:30 pm to
I deal in cancer diagnostics and we already have algorithms that score patients. Pathologists want it as an aide, but it would technically do their job for them in most cases.
Posted by Korkstand
Member since Nov 2003
29002 posts
Posted on 1/31/19 at 3:30 pm to
quote:

Wow, that’s incredibly wrong. It’s damn near impossible to give a yes or no answer in that field.
For a human, maybe.
quote:

One of my friends does it, and there’s a reason he makes an incredible amount of money.
All the more reason many people are determined to automate it.
quote:

It’s incredibly difficult.
It's difficult mostly because you have to spend a lot of time looking at every little detail, and know what you're looking at. These are things that computers are getting very, very good at doing.


I hope your friend's school loans are paid off, because he might be out of a job in the next few years.
Posted by Korkstand
Member since Nov 2003
29002 posts
Posted on 1/31/19 at 3:33 pm to
quote:

The AI better be the Dr too, because good luck getting the oncologist, or whoever, to understand how the AI describes the case.
This is the easiest part to automate.
Posted by Gaston
Dirty Coast
Member since Aug 2008
41163 posts
Posted on 1/31/19 at 3:36 pm to
He started at $700k/yr a long damn time ago. He says that so many damn PAs order shite they don’t need, and the DRs want yes/no answers to shite there is no yes/no for. I think he reads ~150 cases a shift...I’m sure he’d be fine with a pay cut in exchange for the 100 easy ones being read by a computer.
Posted by Gaston
Dirty Coast
Member since Aug 2008
41163 posts
Posted on 1/31/19 at 3:38 pm to
quote:

the easiest part


My friend graduated, Davidson, with an English Lit degree and describe things incredibly well...he says the Drs don’t ‘get’ his descriptions...how’s this going to be easy?
Posted by flyAU
Scottsdale
Member since Dec 2010
24899 posts
Posted on 1/31/19 at 3:39 pm to
quote:

The AI better be the Dr too, because good luck getting the oncologist, or whoever, to understand how the AI describes the case.



The tests that we do point to what treatment would be effective. The actual information going to the oncologist would be very simple.
Posted by flyAU
Scottsdale
Member since Dec 2010
24899 posts
Posted on 1/31/19 at 3:44 pm to
quote:

My friend graduated, Davidson, with an English Lit degree and describe things incredibly well...he says the Drs don’t ‘get’ his descriptions...how’s this going to be easy?



Wish I could go into more detail, but you have a pathologist looking at tissue counting cells and looking at the reactivity to the cells with reagents. In the past it has always been "positive/negative". Advanced diagnostics uses antibodies which expresses in specific types of cancer. The report that the Path puts together is this simple information. Same thing my software is doing.
Posted by Gaston
Dirty Coast
Member since Aug 2008
41163 posts
Posted on 1/31/19 at 3:46 pm to
Not arguing that. Just saying that to my radiologist friend things are getting more difficult for them not easier.
Posted by Ronaldo Burgundiaz
NWA
Member since Jan 2012
6692 posts
Posted on 1/31/19 at 3:48 pm to
quote:


I always get a kick out of a board full of engineers, salesmen, and "consultants" cheering on AI

Engineering will be the dead last thing that AI will be able to replicate. The creativity needed for proper engineering will be incredibly difficult to replicate.

AI is good and will get better at reading MRI data. But it will be incredibly difficult for AI to build a better MRI machine/alternative. We will all be dead before that comes around. Maybe our children or grandchildren will see that day.
Posted by MorbidTheClown
Baton Rouge
Member since Jan 2015
71508 posts
Posted on 1/31/19 at 3:49 pm to
proctology
Posted by flyAU
Scottsdale
Member since Dec 2010
24899 posts
Posted on 1/31/19 at 3:49 pm to
quote:

Not arguing that. Just saying that to my radiologist friend things are getting more difficult for them not easier.



Which actually bodes well for an AI crunching the information and reporting it to a doctor in easily digestible information. In the end, an X-ray is 1's and 0's in their Pacs system. Computer aided (eventual replacement) will happen if we want more advanced care.
Posted by cajuncarguy
On the road...Again!
Member since Jun 2013
3135 posts
Posted on 1/31/19 at 3:50 pm to
Anesthesia
Posted by Upperdecker
St. George, LA
Member since Nov 2014
31815 posts
Posted on 1/31/19 at 3:50 pm to
Psychiatry
Posted by Korkstand
Member since Nov 2003
29002 posts
Posted on 1/31/19 at 3:54 pm to
quote:

I think he reads ~150 cases a shift...I’m sure he’d be fine with a pay cut in exchange for the 100 easy ones being read by a computer.
I don't think you quite grasp how dramatic the difference is between a person reading 150 images per shift, and a computer reading 150 images per second. One way costs thousands of dollars and 12 hours, and the other costs a fraction of a penny and one second. And the cheaper way is more accurate, to boot.

A computer can look at an image, find any issues that it has been "trained" to find, and then store a perfect copy of the entire image in its memory for later. There might be something innocuous in that image that showed up on other images from other patients, and those patients later on developed a problem. No human can do the type of analysis that a computer can. And once you've trained one, you've trained them all. You don't have to send each and every computer through expensive medical school.

Posted by Korkstand
Member since Nov 2003
29002 posts
Posted on 1/31/19 at 3:56 pm to
quote:

My friend graduated, Davidson, with an English Lit degree and describe things incredibly well...he says the Drs don’t ‘get’ his descriptions...how’s this going to be easy?
I'm not sure you understand the point of computers.

The same computer can describe the same image to a thousand different doctors in a manner that each doctor can best understand.
Posted by flash
NOLA
Member since Sep 2005
514 posts
Posted on 1/31/19 at 3:58 pm to
Did you even read your links?
AI can be helpful, but lacks human judgement for critical decision making.

quote:

AI offers a major opportunity to enhance and augment radiology reading, not to replace radiologists.
Posted by Korkstand
Member since Nov 2003
29002 posts
Posted on 1/31/19 at 3:58 pm to
quote:

Just saying that to my radiologist friend things are getting more difficult for them not easier.
That's because he's an imperfect person dealing with imperfect people.
Posted by flyAU
Scottsdale
Member since Dec 2010
24899 posts
Posted on 1/31/19 at 4:00 pm to
quote:

There might be something innocuous in that image that showed up on other images from other patients, and those patients later on developed a problem.


You know what you are talking about.

End goal is to find the patterns in patients amid the myriad of tests performed. This has started out with algorithm's and will build to specific AI's that correlate data from all patients to find specific traits/factors which can point to effective treatment as well as the holy grail: identifying who is going to have these problems and attack them before they happen. Heart medicine before heart problems etc. The goal is negate the illness as much as possible.
first pageprev pagePage 2 of 5Next pagelast page

Back to top
logoFollow TigerDroppings for LSU Football News
Follow us on X, Facebook and Instagram to get the latest updates on LSU Football and Recruiting.

FacebookXInstagram