Paige Receives First Ever FDA Approval for AI Product in Digital Pathology

[KeratinPearls]

Paige Receives First Ever FDA Approval for AI Product in Digital Pathology

Paige, the global leader in AI-based diagnostic software in pathology, today announced that the U.S. Food and Drug Administration (FDA) has granted de

www.businesswire.com

---

Members don't see this ad.

 

[tiredguy]

So how will this be sold to the urologists for their labs? To be marketed as a service to their patients or is there (will there be) a code for AI-assisted diagnosis?

 

[WEBB PINKERTON]

If it is like what happened to cytopathology, there will be a new code for it. That was what happened to cervical cancer screening. Then of course you will need less pathology labor.

 

[deleted314957]

I really have to wonder about their pathologists with the 70% and 24% figures as related to sensitivity/specificity.
Anybody who has successfully practiced path for more than thirty years and has had no complaints from ANY urologist
knows this is bs. I am sure I have made mistakes in many areas of pathology but if I was that sh***y I don’t think they would have kept me around. They talk as if we are not trained to identify subtle and small foci of cancer. What the hell do they thing we do all day. It might be a nice screening tool if it is cheap.
Problem is, uros will only send to folks who use it (prediction), paths will have to buy it, it will be uber expensive and it will not add to compensation.
Time to retire folks.

 

[caffeinegirl]

The barrier to entry here is the adoption digital pathology (getting a scanner in, validating it, etc). Don't think that most pod labs will invest in that in order to implement AI (think of the cost of scanners+AI and compare to labor savings for pathologist time). Might be a consideration for the Bostwick-types.
A CPT code for that would be nice - but doubt that will happen for a long time.

 

[LADoc00]

This has been around in Radiology for a long time and has gone nowhere.

The biggest issue is cost vs. increased efficiency, there isnt a shortage of pathologists who will work on the cheap so this is an AI solution looking for a problem that isnt there.

We need AI to deliver packages, pour beers at a tap house, make pancakes generally work in food service, assist with customer support as we have plenty of folks able to read the now decreased caseload of glass....

 

[octopusprime]

This has been around in Radiology for a long time and has gone nowhere.

The biggest issue is cost vs. increased efficiency, there isnt a shortage of pathologists who will work on the cheap so this is an AI solution looking for a problem that isnt there.

We need AI to deliver packages, pour beers at a tap house, make pancakes generally work in food service, assist with customer support as we have plenty of folks able to read the now decreased caseload of glass....

agree.

 

[WEBB PINKERTON]

So we are now pro pathologist surplus to keep AI away?

AI took off in cytopath but you can bill more for using it and cut labor costs immensely.

 

[Med Director New England]

I don’t see this changing to much. You still have to review carefully a biopsy that the machine tells you is negative & you still have to look carefully at positive biopsies to quantify the cancers in each part, Gleason grade the cores, etc.

Would you rush thru a case just b/c this technology tells you the case is negative ?

 

[Vicinal]

Yeah it seems like BS. Another academic derived useless device. When you look at the companies’ “advisors” it is academic people that have no idea what pathologists actually do.

Then slickly marketed to clinicians that have no idea what pathologists do. It’s mostly our fault because of the way our field is presented in medical school.

Exposure to path during pre-clinical years encompasses seeing representation of a disease state in a picture or on definitively diagnosed slide. The nuances are left out.

Leaders are commonly CP derived (serving as admin or chair because they can hang out all day because techs and instruments do the labor) . To these individuals laboratory tests are provided by whiz bang machines cranking out millions of results and conflate that to AP.

No one really spends the time to understand what a surgical pathologist actually does. Later exposure to path in tumor board presentations, clinicians hear phrases like hyperchromatic of atypical with IHC profiles and the PULM doc or CTS thinks a challenging diagnosis is obvious.

The momentum seems to be the field‘s demise imminent. Academic programs are filling with trainees and attendings that due to poor management are designed to decimate the field. This is compounded by diminishing reimbursement.

It is unfortunate that such an amazing inexpensive, commonly definitive, and reproducible test is being pilfered. Almost any clinician can be fooled to believe they can read and MRI, CT, or an X-ray but put an H and E slide or a marrow aspirate that is remotely complex in front of them and they freak out.

I hate digital images. You can promise generations of new trainees that they can sign out from their hot tub on an iPad. Digital images seem so inefficient to work with.

How long does it take a well trained pathologist to scan an adequately prepared slide. Seconds..

‘Maybe AI can help with inter-observer variable 3+4 versus 4+3 challenges but is that really the problem that needs to solved in medicine?

 

[Path hall]

Yeah it seems like BS. Another academic derived useless device. When you look at the companies’ “advisors” it is academic people that have no idea what pathologists actually do.

Then slickly marketed to clinicians that have no idea what pathologists do. It’s mostly our fault because of the way our field is presented in medical school.

Exposure to path during pre-clinical years encompasses seeing representation of a disease state in a picture or on definitively diagnosed slide. The nuances are left out.

Leaders are commonly CP derived (serving as admin or chair because they can hang out all day because techs and instruments do the labor) . To these individuals laboratory tests are provided by whiz bang machines cranking out millions of results and conflate that to AP.

No one really spends the time to understand what a surgical pathologist actually does. Later exposure to path in tumor board presentations, clinicians heat phrases like hyperchromatic of atypical with IHC profiles and the PULM doc or CTS thinks a challenging diagnosis is obvious.

The momentum seems to be the field‘s demise imminent. Academic programs are filling with trainees and attendings that due to poor management are designed to decimate the field. This is compounded by diminishing reimbursement.

It is unfortunate that such an amazing inexpensive, commonly definitive, and reproducible test is being pilfered. Almost any clinician can be fooled to believe they can read and MRI, CT, or an X-ray but put an H and E slide or a marrow aspirate that is remotely complex in front of them and they freak out.

I hate digital images. You can promise generations of new trainees that they can sign out from their hot tub on an iPad. Digital images seem so inefficient to work with.

How long does it take a well trained pathologist to scan an adequately prepared slide. Seconds..

‘Maybe AI can help with inter-observer variable 3+4 versus 4+3 challenges but is that really the problem that needs to solved in medicine?

The hospital I trained at would operate on 4+3, but 3+4 was a candidate for active surveillance. So it matters, at least at some institutions. At any rate, the reproducibility of Gleason scoring is around 50%. The reproducibly of most pathologic diagnoses is between 80% to 60% for 2 observers. Reproducibility drops when you add more observers. The likelihood of 6 or more pathologists agreeing on a single case is like <20%. That’s just for the histologic diagnosis. Grading and staging also have their own reproducibility issues. No one really talks about it, but it’s pretty bad. Pathology really should be automated from a treatment stand point.

 

[Vicinal]

The hospital I trained at would operate on 4+3, but 3+4 was a candidate for active surveillance. So it matters, at least at some institutions. At any rate, the reproducibility of Gleason scoring is around 50%. The reproducibly of most pathologic diagnoses is between 80% to 60% for 2 observers. Reproducibility drops when you add more observers. The likelihood of 6 or more pathologists agreeing on a single case is like <20%. That’s just for the histologic diagnosis. Grading and staging also have their own reproducibility issues. No one really talks about it, but it’s pretty bad. Pathology really should be automated from a treatment stand point.

General reproducibility? So just because a treatment decision is made on grade 2 versus 3 on a screening test based on PSA we need AI? Maybe in this case using patterns where you are definitively under sampling an organ unless you use fusion with definitive target is the issue. Wouldn’t it make more sense to have a real test to gauge propensity. A marker in urine using mass spec. So is 45% 3 and 55%4 really group 2 or group 3 in the organ. It’s just a needle core. Why not have ancillary tests in areas that are known to be generally interobserver variable for better classification rather than using AI based morphology. The AI patterns have to be programmed into the solution. Who is doing this programming and is this truly representative of the best way figure out propensity? I don’t think it’s helpful to shove every single diagnostic problem into the interobserver variable equation. Path has these areas just like every other field. Grading dysplasia, distinguishing reactive atypia from dysplasia etc. it’s far more reasonable to drive ancillary tests that stratify these areas than to use morphology based AI. Maybe there will be solutions outside of the visible spectrum that can detect signatures of malignant transformation and dysplasia rather than morphology based AI. I’m sure people are working on this. You still have to cut a slide, stain it, and scan it. Propensity should be the focus. A test for propensity should move beyond a 5 micron FFPE slide section and interrogate fluids tissues etc. using ancillary developing technologies.

 

[Vicinal]

The hospital I trained at would operate on 4+3, but 3+4 was a candidate for active surveillance. So it matters, at least at some institutions. At any rate, the reproducibility of Gleason scoring is around 50%. The reproducibly of most pathologic diagnoses is between 80% to 60% for 2 observers. Reproducibility drops when you add more observers. The likelihood of 6 or more pathologists agreeing on a single case is like <20%. That’s just for the histologic diagnosis. Grading and staging also have their own reproducibility issues. No one really talks about it, but it’s pretty bad. Pathology really should be automated from a treatment stand point.

Where is the data to support these numbers and what are the metrics of this presumed variability? Is is dysplasia in Barrets? Serrated polyps? reactive atypia versus dysplasia in urothelial biopsies? ADH? Maybe AI can help with this stuff. For GYN cytology it’s easier because liquid based technology interrogates the entire cell. The procurement and preparation is very mechanized and standardized. FFPE sections are a whole different world. I am always suspicious of studies concerning reproducibility in pathology as they are horrendously designed. Also there is always some diagnostic company run by a non surgical pathologist touting these perceived discrepancies and getting buy in from some luminary eminence based academic. Is pathology the only field that has so much eminence based medicine? Write a book and be the go to person at medical center ultimately driven by surgery and oncotherapy reimbursement and that opinion is the truth?

 

[futuredoc331]

I spent a few months reading about this stuff for a rotation back in med school. There are many places in Europe that are almost all digital. The AI assistance hasn't fully taken over yet, but it's coming.
Even if it never amounts to more than a screening tool that ranks cases/slides in order of importance, it will definitely increase efficiency. Soon 3 pathologists will do the work of 4. Then 2 will do the same.
The robots also don't need benefits, time off, or partnership.

The below link is to the Camelyon challenge which is a competition that was held for a couple of years. The results were impressive.

Home - Grand Challenge

 

[deleted314957]

The hospital I trained at would operate on 4+3, but 3+4 was a candidate for active surveillance. So it matters, at least at some institutions. At any rate, the reproducibility of Gleason scoring is around 50%. The reproducibly of most pathologic diagnoses is between 80% to 60% for 2 observers. Reproducibility drops when you add more observers. The likelihood of 6 or more pathologists agreeing on a single case is like QUOTE]

I don’t believe it and it think is is BS. I DO believe you have figures, studies, experts, etc. which and who do believe and publish it. I don’t believe their studies or figures.

Now for my reasoning. How could/can so many of my cohort have done this for thirty plus years and been so sh***y?
We would have been out on our asses. We generally had excellent agreement between bx and excision. I never had that many pathologists disagree with me. Zero uro complaints. 35 years! It is nuts.

Do you practice pathology and, if so, how long? This study doesn’t seem foolish to you?

 

[Vicinal]

I spent a few months reading about this stuff for a rotation back in med school. There are many places in Europe that are almost all digital. The AI assistance hasn't fully taken over yet, but it's coming.
Even if it never amounts to more than a screening tool that ranks cases/slides in order of importance, it will definitely increase efficiency. Soon 3 pathologists will do the work of 4. Then 2 will do the same.
The robots also don't need benefits, time off, or partnership.

The below link is to the Camelyon challenge which is a competition that was held for a couple of years. The results were impressive.

Home - Grand Challenge

Many places in Europe? For primary diagnosis and not IHC? Please name a few. Would love to read all about it.

 

[caffeinegirl]

Many places in Europe? For primary diagnosis and not IHC? Please name a few. Would love to read all about it.

There are several players in this space in Europe - Ibex has I think the edge. Here is just a snippet of their adoption so far

Unilabs signs deal with Ibex to deploy AI-powered cancer diagnostics

Unilabs, a leading international provider of diagnostic services, and Ibex Medical Analytics, the pioneer in artificial intelligence-powered cancer diagnostics, have agreed a deal to implement Ibex’ AI platform in Europe.

unilabs.com

UK expands funding to trial Ibex’s AI-based prostate cancer diagnostic

The UK has announced funding for six new NHS trusts that will be part of the clinical trial of Ibex’s AI technology to detect prostate cancer.

www.medicaldevice-network.com

 

[deleted314957]

I spent a few months reading about this stuff for a rotation back in med school. There are many places in Europe that are almost all digital. The AI assistance hasn't fully taken over yet, but it's coming.
Even if it never amounts to more than a screening tool that ranks cases/slides in order of importance, it will definitely increase efficiency. Soon 3 pathologists will do the work of 4. Then 2 will do the same.
The robots also don't need benefits, time off, or partnership.

The below link is to the Camelyon challenge which is a competition that was held for a couple of years. The results were impressive.

Home - Grand Challenge

I got news for all you younger folks. From a financial and job security aspect y’all better hope that every path has to do the work of 2-3 paths. They have cut reimbursement so much that is what it will take to get back to 500k/yr.
Used to do in the neighborhood of 10-12k surgical/yr ( admittedly gi heavy) but it also included all hospital work (180 beds suburban)and I was medical director.

Today it seems lots of paths complain at the 5k level without having any significant other responsibilities. Or won’t do marrows or fna’s or read this or that and did not adequately learn clin path.

There are too many of us and those that are working,frankly, are underworked. Just look at academia. There are places with 12 or 16 residents and there are 35 staff pathologists.
The whole place does 40k surgical/yr. That is ideal for 7 foursomes for bridge and the rest can work/teach.

 

[Vicinal]

I got news for all you younger folks. From a financial and job security aspect y’all better hope that every path has to do the work of 2-3 paths. They have cut reimbursement so much that is what it will take to get back to 500k/yr.
Used to do in the neighborhood of 10-12k surgical/yr ( admittedly gi heavy) but it also included all hospital work (180 beds suburban)and I was medical director.

Today it seems lots of paths complain at the 5k level without having any significant other responsibilities. Or won’t do marrows or fna’s or read this or that and did not adequately learn clin path.

There are too many of us and those that are working,frankly, are underworked. Just look at academia. There are places with 12 or 16 residents and there are 35 staff pathologists.
The whole place does 40k surgical/yr. That is ideal for 7 foursomes for bridge and the rest can work/teach.

Exactly. It is unbelievable and the academics keep complaining that they are over worked. I can’t figure it out. Is it inability to focus, faulty time management, subspecialization, or all of the above.

 

[LADoc00]

I got news for all you younger folks. From a financial and job security aspect y’all better hope that every path has to do the work of 2-3 paths. They have cut reimbursement so much that is what it will take to get back to 500k/yr.
Used to do in the neighborhood of 10-12k surgical/yr ( admittedly gi heavy) but it also included all hospital work (180 beds suburban)and I was medical director.

Today it seems lots of paths complain at the 5k level without having any significant other responsibilities. Or won’t do marrows or fna’s or read this or that and did not adequately learn clin path.

There are too many of us and those that are working,frankly, are underworked. Just look at academia. There are places with 12 or 16 residents and there are 35 staff pathologists.
The whole place does 40k surgical/yr. That is ideal for 7 foursomes for bridge and the rest can work/teach.

Yah, its incredible to me how younger folks complain with what I consider almost hobby-like levels of work. 5K surgicals a year is a like a 3-hour a day job.

WTF are folks doing with the other 21 hours/day??

 

[KeratinPearls]

I got news for all you younger folks. From a financial and job security aspect y’all better hope that every path has to do the work of 2-3 paths. They have cut reimbursement so much that is what it will take to get back to 500k/yr.
Used to do in the neighborhood of 10-12k surgical/yr ( admittedly gi heavy) but it also included all hospital work (180 beds suburban)and I was medical director.

Today it seems lots of paths complain at the 5k level without having any significant other responsibilities. Or won’t do marrows or fna’s or read this or that and did not adequately learn clin path.

There are too many of us and those that are working,frankly, are underworked. Just look at academia. There are places with 12 or 16 residents and there are 35 staff pathologists.
The whole place does 40k surgical/yr. That is ideal for 7 foursomes for bridge and the rest can work/teach.

Pathologists aren’t trained to become general pathologists nowadays after graduating residency. Hell I talked with an academic from MD Anderson who told me graduates aren’t capable of independent signout by the time they graduate residency.

Ive met attendings who aren’t comfortable signing out cyto or heme. I know of people who are AP/CP trained but dropped their CP because they failed it and had no hope in passing it.

 

[KeratinPearls]

Yah, its incredible to me how younger folks complain with what I consider almost hobby-like levels of work. 5K surgicals a year is a like a 3-hour a day job.

WTF are folks doing with the other 21 hours/day??

Drinking coffee and eating snacks.

 

[WEBB PINKERTON]

Use those extra hours to run other side projects. Make money, create an empire.

 

[KeratinPearls]

Use those extra hours to run other side projects. Make money, create an empire.

Laundromats and tree cutting businesses. Any other business ideas Webb?

 

[WEBB PINKERTON]

Laundromats and tree cutting businesses. Any other business ideas Webb?

Tattoo removal, nail salon, persian night club, too many to list. Or you could just run shine like some of the techs do down here.

 

[LADoc00]

Tattoo removal, nail salon, persian night club, too many to list. Or you could just run shine like some of the techs do down here.

View attachment 343862

okay the fact you specify Persian nightclub has me quite interested.

Like architecture wise or hot women wise or annoying dudes in black on black suits wise??

Im in on the plan but I want details.

 

[Zarniwoop]

Midwest pathologist owned Persian nightclubs? This sounds road trip worthy...Any way to 2-fer with a CME meeting?

 

[futuredoc331]

I got news for all you younger folks. From a financial and job security aspect y’all better hope that every path has to do the work of 2-3 paths. They have cut reimbursement so much that is what it will take to get back to 500k/yr.
Used to do in the neighborhood of 10-12k surgical/yr ( admittedly gi heavy) but it also included all hospital work (180 beds suburban)and I was medical director.

Today it seems lots of paths complain at the 5k level without having any significant other responsibilities. Or won’t do marrows or fna’s or read this or that and did not adequately learn clin path.

There are too many of us and those that are working,frankly, are underworked. Just look at academia. There are places with 12 or 16 residents and there are 35 staff pathologists.
The whole place does 40k surgical/yr. That is ideal for 7 foursomes for bridge and the rest can work/teach.

No argument here. That'll be good for the few people who have jobs. Everyone else will be SOL.

 

[deleted314957]

Maybe this field has become complex and esoteric enough to make it a 6 year program, BUT, the staff/ institution are going to have to “ wink” at the rules to get some real, independent sign out. You CANNOT become a competent and independently functioning pathologist if you are not allowed to bottom line a case yourself, period. If the staff cannot deal with that or cannot trust you, they or you are at the wrong place and they should can you as a resident or they should quit or be pushed out by a program director who actually gives a damned and expects his/her residents to be prepared as we were years ago. We had zero problems in the Navy 40 years ago. Things were not quite as complex and we trained as morphologists. Today, I don’t think a lot of gen x residents can function without lots of immunos and flow. No wonder clinicians think we suck hind teat. Morphology takes a back seat. Juan Rosai is spinning.
And I think the wailing and gnashing of teeth from gen x about the two years would be deafening even though they admit they can’t sign out Jack after 4 years and they don’t learn clin path. And, perhaps folks won’t need so many fellowships.

Please note, I am not demeaning the importance of NGS, flow, immunos etc., but I really think such advances have, of necessity, made 6 years necessary. Maybe then we can look “smart” and re-assume our previous position of respect and prestige. Also, such a requirement might help weed out the deadwood who see path as a “fall back” specialty.

 

[Vicinal]

Kind of crazy that Roche owns
Foundation Medicine
Genentech
Ventana