Impact of AI on Derm

[krby10]

Really interesting article about the (potential) future of dermatology: The End of Human Doctors – The Bleeding Edge of Medical AI Research (Part 2). Looks like the golden age of derm is coming to an end.

---

Members don't see this ad.

 

[NSight8]

The guy writing this article seems like he knows nothing about what dermatologists do .... I'm glad that he thinks that al derms do is decide whether something is benign or malignant and then do biopsies. You do a lot of management of all kinds of skin disease, both medically and surgically in derm... I think this would have an impact on the field but not nearly as much of an impact as it would on purely diagnostic fields like diagnostic pathology and diagnostic radiology. The deep learning system is completely agnostic to the type of images you feed it, it just happens that they fed it derm images, radiology or path imaging deep learning just required enough imaging


Sent from my iPhone using SDN mobile

 

[NSight8]

And if you want to look at the other ROAD specialties, every one is facing challenges: Radiology may be significantly impacted by AI, deep learning, automated diagnoses, Ophtho: optometrists in the mid west just won rights to perform eye surgeries, etc, which is fairly ridiculous but obviously poses a very significant threat, anesthesia: impact of CRNAs continues to be an issue ..... so in doing predictions of where a field will go, just know that all fields (in medicine and outside of it), will be impacted by automation, competing practitioners, disruption by tech.... so just be good at what you do, be innovative, don't be a dinosaur, and you'll be fine


Sent from my iPhone using SDN mobile

 

[jdomm]

The guy writing this article seems like he knows nothing about what dermatologists do .... I'm glad that he thinks that al derms do is decide whether something is benign or malignant and then do biopsies.

Umm ...

What of dermatologists then? Well, this system can only identify 9 subgroups as well as dermatologists, out of 2032 diagnoses. For now, dermatologists have the edge in something like 2023 diagnostic tasks.

I don't think he is being unreasonable. Mole clinics are a thing, and make up a pretty big amount of practice for many derms. Not only that, this looks at non-melanocytic lesions too.

 

[doctalaughs]

Really interesting article about the (potential) future of dermatology: The End of Human Doctors – The Bleeding Edge of Medical AI Research (Part 2). Looks like the golden age of derm is coming to an end.

These types of articles make me laugh. Even if the algorithms were perfect - ie 100% match with malignant versus benign which is far from truth- it would not replace a single dermatologist.

These types of algorithms might help me do my job in the same way a vacuum cleaner or washing machine might help a cleaning person. But if you haven't replaced a cleaning person with a robot yet, why in the hell would a software program that assists in one aspect of a job replace a doctor who does a million things?

Look at it this way. Lets just say we eliminate everything a dermatologist does except a skin check. Lets say we then magically eliminate liability and trust issues and take a high schooler who is just trained to do a biopsy only, relying only on the algorithm.

The first patient walks in with 300 lesions. 200 are moles, 50 AKs, 25 SKs, 10 warts, 1 embedded tick and 1 piece of stuck-on lint. Will it be logistically feasible for the program to image all these in a reasonable time-frame (probably not?). What will the program say when encountering something unexpected (lint or tick)? Will the patient be happy if they get a biopsy of lint? Will thw patient be happy with 25 recommended biopsies? Is the computer going to triage atypical nevi? Are you trusting the high schooler to correlate path result with clinical and to do facial biopsies? Again, that is a tiny part of what dermatologists do and even that is almost impossible to replace.

Sent from my iPhone using SDN mobile

 

[krby10]

These types of articles make me laugh. Even if the algorithms were perfect - ie 100% match with malignant versus benign - it would not replace a single dermatologist.

These types of algorithms might help me do my job in the same way a vacuum cleaner or washing machine might help a cleaning person. But if you haven't replaced a cleaning person with a robot yet, why in the hell would a software program that assists in one aspect of a job replace a doctor who does a million things?

You don't need to automate all the tasks a dermatologist does to significantly reduce the number of dermatologists needed. If you are able to automate the simple bread and butter cases like this that comprise the majority of what dermatologists do on a day to day basis, then you will need fewer dermatologists in total. It's basic economics.

 

[ActinicKeratosis]

This article goes along with the "sky is always falling" propaganda. I'm not a dermatologist yet, but based on my limited dermatology experience technology is a very very long way off from replacing a physician. McDonald's isn't even fully automated yet...it can't be too hard to teach robots to serve a Big Mac with fries and a drink right? I'm hopefully going into the field and I am not worried at all.

 

[doctalaughs]

You don't need to automate all the tasks a dermatologist does to significantly reduce the number of dermatologists needed. If you are able to automate the simple bread and butter cases like this that comprise the majority of what dermatologists do on a day to day basis, then you will need fewer dermatologists in total. It's basic economics.

Ok tell me HOW with your vast knowledge of how dermatology is practiced. Specifically. You have a software program that distinguishes 5-10 different lesion types (hand picked images by dermatologists) with reasonable accuracy particularly on the benign/malignant distinction

Tell me how you would integrate this to "need fewer dermatologists in total." Hint: many answers would generate the need for more dermatologists


Sent from my iPhone using SDN mobile

 

[NSight8]

Krby10 and jdomm, learn more about the paper, it didn't even compare malignant melanomas to seborrheic keratosis which is the more difficult comparison... and there was a significant overlap between these groups from the algorithm. Also these were preselected images... telederm is wrong approximately 30% of the time compared to real derm... this is far from actually imaging a real person. So instead of trying to bash a field you don't understand, go crawl back under the rocks you came from. Ignorant trolls


Sent from my iPhone using SDN mobile

 

[doctalaughs]

It truly boggles the mind how people think AI will replace a doctor when it hasn't replaced a nanny, a teacher or even a janitor or a cleaning person. Sure, AI has great potential to change/ improve /supplement physician work but wont replace even a PA or nurse for likely centuries (at which time robots will likely be our overlords anyway so who cares).

Sent from my iPhone using SDN mobile

 

[asmallchild]

This thread is going to be on a short leash

I like articles like the one the OP posted from an intellectual curiousity viewpoint. These are almost worthless (IMO) from a practical viewpoint. Researchers have been looking for the magic bullet for skin examination for decades now (hint: we still haven't found it despite researchers looking into total body photography, Melafind, dermoscopy, confocal microscopy, etc, etc, etc)

I like articles like these and it may very well be possible that our ability to practice medicine will be enhanced by AI in the coming years

Like most others with actual derm experience, I have no fears that AI will actually take my job within my working career (not to mention full body skin exams currently comprise 0% of my practice.....)

 

[krby10]

I don't think AI will "take anyone's job" either. AI won't automate the job of a dermatologist, but in the near future it will automate many of the tasks dermatologists do, particularly those that are not that complex and for cases that are common. Some of these tasks will increase the demand for other tasks completed by dermatologists (as in this case with biopsies), but others won't. In the long run it is undeniable that we will need far fewer dermatologists 20 years from now than we do today. Not a problem for those currently practicing, but something for those in training to keep in mind.

 

[asmallchild]

I don't think AI will "take anyone's job" either. AI won't automate the job of a dermatologist, but in the near future it will automate many of the tasks dermatologists do, particularly those that are not that complex and for cases that are common. Some of these tasks will increase the demand for other tasks completed by dermatologists (as in this case with biopsies), but others won't. In the long run it is undeniable that we will need far fewer dermatologists 20 years from now than we do today. Not a problem for those currently practicing, but something for those in training to keep in mind.

It doesn't sound like you have much experience in dermatology. Would you care to elaborate on which mundane tasks within dermatology are susceptible to automation by AI?

I don't think anyone who has secured a coveted dermatology residency slot is going to lose any sleep over these articles / doomsday prophecies. In the end, there will always be a need for well-trained dermatologists.

 

[doctalaughs]

In the long run it is undeniable that we will need far fewer dermatologists 20 years from now than we do today. Not a problem for those currently practicing, but something for those in training to keep in mind.

Your post makes sense until this statement.

"Undeniable" is ridiculous.

We may need less dermatologists in 20 years, we may need more but the tech developments in that article have no bearing on that at all. Demographics, population, medicolegal and insurance/ healthcare reform have way, way more influence.

Would you say it's "undeniable" we'll need fewer teachers? Surgeons? Internists? Software programmers? Daycare providers? Firefighters?

I can imagine parts of each of these jobs that may be automated in 20 years yet it would be asinine for me to say it's "undeniable" we will have fewer people in each of these professions.

You clearly have very little insight into what dermatologists (or other doctors) do. I have no doubt radiologists and pathologists will continue to be needed as well.




Sent from my iPhone using SDN mobile

 

[jdomm]

I'm not really sure why there are several people are claiming the author thinks that derm is gonna disappear because of this AI. The piece goes out of its way to say that this is only a single task (albeit an important one).

Researchers have been looking for the magic bullet for skin examination for decades now (hint: we still haven't found it despite researchers looking into total body photography, Melafind, dermoscopy, confocal microscopy, etc, etc, etc)

The same could be said for self-driving cars, or Go playing systems, or machine translation. We have new technology that is only 5 years old, and it is far more successful in automating expert tasks than anything that has come before.

 

[doctalaughs]

I'm not really sure why there are several people are claiming the author thinks that derm is gonna disappear because of this AI. The piece goes out of its way to say that this is only a single task (albeit an important one).



The same could be said for self-driving cars, or Go playing systems, or machine translation. We have new technology that is only 5 years old, and it is far more successful in automating expert tasks than anything that has come before.

I don't really have any problem with the author's article and think this technology is "neat" and "interesting" (although not currently practical - ie I would not use it in my practice now but might be useful to me in the future).

I think everyone is more responding to the OP who makes outlandish claims about his interpretation of the article predicting this technology will "undeniably" reduce the need for dermatologists and the "golden age of dermatology is over."


Sent from my iPhone using SDN mobile

 

[username456789]

I don't really have any problem with the author's article and think this technology is "neat" and "interesting" (although not currently practical - ie I would not use it in my practice now but might be useful to me in the future).

I think everyone is more responding to the OP who makes outlandish claims about his interpretation of the article predicting this technology will "undeniably" reduce the need for dermatologists and the "golden age of dermatology is over."


Sent from my iPhone using SDN mobile

Bingo.

Based on the account details (member for two years, never posted until this thread), it smacks of an alt account for someone who didn't match Derm.

 

[krby10]

Premed actually, although I admit I doubt I could match into Derm even if I got into med school. More so I was just interested in hearing the thoughts of actual dermatologists. Apologies if I came off as inflammatory.