Which area of pathology is most immune from AI?

[Spikebd]

With a lot of talk about AI replacing radiologists and pathologists, it has me wondering which field is most protected from image analysis or other forms of AI.
I think a lot of CP work would be difficult to replace with AI. Being a lab manager will be a relatively secure gig if you can find a job. Forensics is obviously safe from these developments as well.

My AI proof fields of pathology
1. Forensics
2. Blood banking and transfusion med
3. Molecular/lab manager
4. At the bench grossing

Most worrisome areas
1. Paps
2. GI biopsies
3. Cyto
4. Other AP fields?

I’ve wondered about hemepath, but it seems difficult for an AI to integrate flow/smear/aspirate/marrow/molecular all in one coherent report.

I think a lot of people outside medicine or even med students think pathology is easier than it really is. I’m slightly concerned about it impacting my job and salary prospects, but think AI still has a long ways to go.

---

Members don't see this ad.

 

[octopusprime]

Be careful, Skynet is watching you.

 

[jupiterianvibe]

Dermpath is immune. There is too much clinicopathologic correlation required.

 

[dukeresident]

Dermpath is immune. There is too much clinicopathologic correlation required.

None are immune, some are just safer than others.

 

[wholeheartedly]

So, as a tech we were going to get this fancy new specimen sorter that the lab assistants could put samples into and it would separate out the different tube types and send them to the automated processing line to be spun, aliquoted, or tested or whatever was needed. It was supposed to be a huge time and labor saver. Send the EDTA tubes to the heme line, coags to the coag machines, aliquot and run chemistries. Yeah, not so much in reality.

If you can't automate machines to perform these kinds of tasks accurately, I'm not really too worried about AI replacing pathologists in our lifetimes. I think it'll be a supplemental tool.

 

[gbwillner]

The only things we do that I believe are truly immune from tech are:
1. FNA
2. Lab management/Lab Director roles
3. Forensics/autopsy
4. Clinical consults/tumor boards

... that's about it. Everything else has some risk.

 

[LADoc00]

Why are people still even getting pap smears read?

Guardasil 9 is insanely effective and viral testing has, as predicted long ago, been show to be equivalent and even more sensitive than glass monkeys making a billion slides from vagina brushes...how the heck that is still a "thing" in 2019??

Are cytologists out there still writing reports with parchment and quills as well???

Forensics is useless FYI as with a wave of a hand by a sheriff, you dont even need an medical degree to investigate death so not sure why that is on the list.

In summary, AI ISNT what you folks need to worry about. You need worry about the collision of common sense with declining access to healthcare/LE dollars.

 

[yaah]

For the life of me, I can't figure out the sturm und drang about AI and pathology slides. Pathology slides are essentially a 19th century technology that we have optimized over the years and are currently an effective and cheap way to analyze things. But there are other ways to analyze tissue. I don't understand why there is such intense emphasis on AI for H&E review. Sure, it might be useful for certain things, it might improve efficiency, but it ain't cheap yet! And it doesn't really add a ton of value at present for the capital required to make it useful (computer power, digital storage, software, hardware, none of which adds any reimburseable value). Where AI is probably going to have more value is a different sort of tissue analysis possibly based on molecular techniques. I don't know. maybe I'm wrong. But follow the money. Replacing one pathologist by efficiency gains from AI and all the expense that goes into that might end up costing more than it costs to just pay the pathologist. We can't even get the hospital to pay for a decent IT system, let alone AI technology that doesn't benefit the hospital at all.

 

[WEBB PINKERTON]

Why are people still even getting pap smears read?

Guardasil 9 is insanely effective and viral testing has, as predicted long ago, been show to be equivalent and even more sensitive than glass monkeys making a billion slides from vagina brushes...how the heck that is still a "thing" in 2019??

Are cytologists out there still writing reports with parchment and quills as well???

Forensics is useless FYI as with a wave of a hand by a sheriff, you dont even need an medical degree to investigate death so not sure why that is on the list.

In summary, AI ISNT what you folks need to worry about. You need worry about the collision of common sense with declining access to healthcare/LE dollars.

Pap-Tissue correlation is still vitally important. Many significant lesions will be missed without a Pap (at least in my experience). Co-testing isn't wasting very many health care dollars. My god, cytopathology accession numbers have probably dropped 60 percent in the last 5 years! Meanwhile CAP and ASCP haven't dropped the price of their proficiency testing any. A national duopoly will do that. If they cared about their members, they would acknowledge the changing landscape of Pap screening and lobby to reduce the burden so those of us that want to continue to offer the service can. It is one of the most important roles we serve. Seems like the cases of Cervical cancer are some of the saddest we see as it is typically mid to upper 30s female with young children.

Anti-vaxxers aren't exactly helping either. Still a lot of unnecessary HPV infections out there.

 

[rirriri]

Dermpath is immune. There is too much clinicopathologic correlation required.

I regularly talk to AI companies and am working on my own AI project.

Dermpath is one of THE first targets for AI companies, FYI----along with GI and Prostate biopsies.

As with any and all technology in medicine; don't feel threatened by it. Better to accept the reality it is coming...and adjust accordingly.

In realistic terms we are years away from any meaningful incorporation of AI into any workflow. You will hear hype-filled articles from time to time but ignore it. The vast majority of Path departments haven’t even embraced digital pathology yet—-so AI is a few notches beyond that at this point in terms of value and efficiency in the current state of Path workflow.

 

[Napoleon1801]

From what I've seen/heard of AI, it likely can't replace pathologists, but will likely act as significant aid as digital pathology begins to develop. Look at current technology like the Iris automated urinalysis and Sysmex Cellavision. These two software platforms do a decent job at classifying basic ID's, but requires a tech to oversee the call. Something similar might come to AP, but it will be awhile given the expense to switch. If image analysis software could highlight microinvasion, mitotic figures, and such, it could be a helpful adjunct.

My bet would be that places like Quest would be the first ones to double down on the technology.

 

[dukeresident]

Note...https://www.nature.com/articles/s41591-019-0539-7.epdf?author_access_token=BI9AOTsesmNoV2lSdpucn9RgN0jAjWel9jnR3ZoTv0PDGU3ZwysZtsN41a2fOgaoj4PRxjTvAHjSFrKF_S_mq4QNNV8dNoxAjytIQuVz9vdjplLQHUSEPiIo392MzIJY8fqxLKHC5vIwNpLLEoXMnA%3D%3D&fbclid=IwAR1y8UNZtbBLr5Xlty7fwfV5iTgiBiN-Lf4sc8QFbzYydUQAgU1WxqSpwDk

 

[Adrian Cocot]

I predict that AI will be a disruptive innovation. The low hanging fruit is going to be things like polyps, SCC/BCC, PIN/prostate CA, LSIL/HSIL, endometrial dating -- basically any kind of high-volume biopsy specimen that require little brainpower to interpret, i.e. 85% of GYN, GI, GU and skin biopsies. Probably IHC reading as well (think PD-L1, MMR, ER/PR, etc). All the difficult, litigation-prone stuff will be left for humans, so that attorneys will know whose name they can put on the lawsuit.

Oh, and as LADoc says, automatization will probably lead to downward pressure on reimbursement for the biggest billing codes (305 and 342). So with the traditional income streams drying up, pathologists will have to get either very creative (or lucky) to retain an independent practice -- or get used to an inpatient job that's more administration, middle management and fighting for your little slice of the DRGs.

The saving grace here is that even when the tech will be mature (~10 years, my guess), regulatory issues will delay full-scale adoption by another decade or thereabouts. So you millennial and genZ med students and residents should still be able to squeeze out a career, but you won't be able to overstay your welcome like all the boomers have.

 

[Torsed]

Forensics?

 

[WEBB PINKERTON]

Article about AI and digital pathology in the new CAP today

Better than glass slides? Digital pathology’s challenge - CAP TODAY

September 2019—The FDA’s clearance this year of a second digital pathology system for primary diagnosis is likely to lead to lower costs and greater innovation and interoperability, say digital pathology advocates. A multicenter study led the FDA in May to clear Leica Biosystems’ Aperio AT2 DX...

www.captodayonline.com