Impact of digital pathology?

[Dro133]

Saw this press release this morning about new digital pathology software from Roche:

Roche launches uPath enterprise software with improved speed, performance and usability for digital pathology

As an MS3 interested in pathology, this got me thinking about the future of the pathologist's workflow, which I admittedly don't know much about. What do the attendings/residents on this forum think about how "digital pathology" will change the pathologist's workflow in the future, if at all? Does anyone see a day when a pathologist gets a scan of a sample sent directly to their computer, without having to look through a microscope at all?

---

Members don't see this ad.

 

[cmz]

Pathologists are making diagnoses on scanned images alone, right now. This is nothing new.

I think digital pathology will definitely create a new, higher ceiling for pathology when it comes to work-load efficiency. AI is on the horizon, too. To me, these are going to be every day tools that I am going to use when signing my cases out. Our new mid level practitioners will be named Siri and Alexa or whatever name goes along with what Google is developing. I will come to work with all of my slides pre-screened by AI and certain ancillary tests will already have been ordered for me (and cases may already even be pre-dictated). Can you imagine what all of this increased efficiency will do to the job market? 😉

Right now, though, I think that we're greatly limited on the number of digital products out there on the market right now that can scan with a decent resolution. By decent, I mean something at least equal to what looking at a glass slide on your own microscope can provide. The higher the resolution, the larger the data files will be. The larger the data files, the more physical HDD space you're going to need. Running parallel to the storage space issue, you need to be able to move the data across the network at a reasonable speed. This is going to require a decent local and global network infrastructure. Once you exceed your local HDD server storage, you'll need a server farm to offload your digital goods (maybe a HIPAA compliant secure cloud service). This costs $$$. Pennies add up to dollars.

To be quite honest, I would rather pay a service $200k/yr that can digitize all of my cases and manage the bandwith/server/storage needs of my group if that means I can be as efficient as two pathologists. I wouldn't have to hire nearly as many pathologists. If there is a difficult case that requires an expert consult, how easily do you think I could outsource this with digital pathology? Pretty damn easily.

 

[Wantnosmoke24]

This is good for patient care, bad for the job market and pathologists.

 

[jupiterianvibe]

This is good for patient care, bad for the job market and pathologists.

ive been saying for years that pathology will eventually be practiced solely in large centralized locations, probably academic

smaller locations will need more PAs

maybe the CAP should push for cutting residency programs while somehow making it profitable for ****ty small programs to train PAs or something like that

 

[Spikebd]

ive been saying for years that pathology will eventually be practiced solely in large centralized locations, probably academic

smaller locations will need more PAs

maybe the CAP should push for cutting residency programs while somehow making it profitable for ****ty small programs to train PAs or something like that

You can say the same thing for a lot of medicine outside of ER/Trauma/primary care.

“No hip replacements unless it’s performed by a total joints trained orthopod at a regional mega center of Medicare excellence” You can extrapolate that line of thinking to most all surgical specialties, really. The centralization of health care is real. Maybe it’s time for doctors to form unions.

 

[AZpath]

This will probably happen slowly. The cost is substantial. Radiology imaging is natively digital. Not so, with path. We have to produce a slide first.

I ask about low cost slide scanner on this list recently. I did not get a single response. This tells me most paths are not thinking about digital implementation



Max

 

[Wantnosmoke24]

This will probably happen slowly. The cost is substantial. Radiology imaging is natively digital. Not so, with path. We have to produce a slide first.

I ask about low cost slide scanner on this list recently. I did not get a single response. This tells me most paths are not thinking about digital implementation



Max

Agreed. Cost is biggest hurdle.

Problem is that when it does get implemented, it'll make pathologists much more efficient and we won't need as many. What Radiology has going for it is its volume, efficiency is actually a positive thing for them, up to a certain point.

 

[AZpath]

Pathologists are making diagnoses on scanned images alone, right now. This is nothing new.

I think digital pathology will definitely create a new, higher ceiling for pathology when it comes to work-load efficiency. AI is on the horizon, too. To me, these are going to be every day tools that I am going to use when signing my cases out. Our new mid level practitioners will be named Siri and Alexa or whatever name goes along with what Google is developing. I will come to work with all of my slides pre-screened by AI and certain ancillary tests will already have been ordered for me (and cases may already even be pre-dictated). Can you imagine what all of this increased efficiency will do to the job market? 😉

Right now, though, I think that we're greatly limited on the number of digital products out there on the market right now that can scan with a decent resolution. By decent, I mean something at least equal to what looking at a glass slide on your own microscope can provide. The higher the resolution, the larger the data files will be. The larger the data files, the more physical HDD space you're going to need. Running parallel to the storage space issue, you need to be able to move the data across the network at a reasonable speed. This is going to require a decent local and global network infrastructure. Once you exceed your local HDD server storage, you'll need a server farm to offload your digital goods (maybe a HIPAA compliant secure cloud service). This costs $$$. Pennies add up to dollars.

To be quite honest, I would rather pay a service $200k/yr that can digitize all of my cases and manage the bandwith/server/storage needs of my group if that means I can be as efficient as two pathologists. I wouldn't have to hire nearly as many pathologists. If there is a difficult case that requires an expert consult, how easily do you think I could outsource this with digital pathology? Pretty damn easily.

It does not add much productivity yet
No killer app to pay for it

You can get great resolution , files are huge
Then as you say, we need big storage and bandwidth. The best is case is for reducing courier costs and faster distribution

 

[Med Director New England]

I just saw a digital platform presented with digitized thinprep Paps. Fantastic quality and imo will be a game changer.

Will allows CTs and cytopaths to work more efficiently and for large labs the technology will pay for itself in probably a year. Will be FDA approved likely in late 2019 or 2020

I think GYN cytology is already pretty centralized to large labs and will be the first part of AP digitized.

 

[WEBB PINKERTON]

I just saw a digital platform presented with digitized thinprep Paps. Fantastic quality and imo will be a game changer.

Will allows CTs and cytopaths to work more efficiently and for large labs the technology will pay for itself in probably a year. Will be FDA approved likely in late 2019 or 2020

I think GYN cytology is already pretty centralized to large labs and will be the first part of AP digitized.

The game changer is HPV testing. What you saw is decades too late.

There will be few to no community labs left in the future. Almost all of pathology will be centralized. PAMA may make it happen in the very near future.

Do Community Labs Have a Future in the U.S.?

I wouldn't invest in companies working on pathology AI. Slide screening isn't the future.
Better invest in companies like Nanostring and Guardant Health.

 

[Adrian Cocot]

If you still have to make a slide, the tech is superfluous. Digital pathology is overrated, and its potential is dwarfed by biotech.

The money will be going to novel Dx/Tx solutions. Think Foundation One, Exact Sciences, Guardant, etc.

 

[SetecAstronomy]

I'm one of those digital pathology peeps, and I actively work on machine learning for rendering diagnoses.

In my opinion, there would need to be a quantum surge in the real-world utility, plus lower costs in order for digipath to be a real threat to jobs. Where I see this area going, is for QA/QC, retrospective reviews, and speeding up mundane tasks (counting mitoses, measuring depth of invasion, autopopulating some parts of synoptic reporting, etc).

A real world case where this could have been useful:
3 deaths found in investigation of impaired pathologist | WTOP

A technician could've loaded the slides into a scanner, and a computer could analyze the slides and compare its dx to the impaired pathologist's. Any discrepancy would be kicked out to a human pathology for confirmation. This could have reduced the burden on the volunteer pathologists and probably sped up the process.

 

[Dro133]

I'm one of those digital pathology peeps, and I actively work on machine learning for rendering diagnoses.

In my opinion, there would need to be a quantum surge in the real-world utility, plus lower costs in order for digipath to be a real threat to jobs. Where I see this area going, is for QA/QC, retrospective reviews, and speeding up mundane tasks (counting mitoses, measuring depth of invasion, autopopulating some parts of synoptic reporting, etc).

A real world case where this could have been useful:
3 deaths found in investigation of impaired pathologist | WTOP

A technician could've loaded the slides into a scanner, and a computer could analyze the slides and compare its dx to the impaired pathologist's. Any discrepancy would be kicked out to a human pathology for confirmation. This could have reduced the burden on the volunteer pathologists and probably sped up the process.

That's very interesting, thanks for your reply! It sounds like you don't work in this area directly, but do you have any thoughts on the use of machine learning for prognostication and treatment selection as it relates to pathology? Is this an area with a future for someone interested in pathology informatics?

 

[y2k_free_radical]

If you still have to make a slide, the tech is superfluous. Digital pathology is overrated, and its potential is dwarfed by biotech.

The money will be going to novel Dx/Tx solutions. Think Foundation One, Exact Sciences, Guardant, etc.

Those are very interesting as well as NANO, but i have recently been selling and even starting to short some ETFs

 

[SetecAstronomy]

That's very interesting, thanks for your reply! It sounds like you don't work in this area directly, but do you have any thoughts on the use of machine learning for prognostication and treatment selection as it relates to pathology? Is this an area with a future for someone interested in pathology informatics?

I'm not sure what you mean. There are other scientists/informaticians working on computer-aided diagnosis (CAD) for learning EMR's. The diagnoses we provide as pathologists are of immense utility when treating patients, but it really is only 1 part in a complex system of medicine. Maybe machine will replace pathologists (and radiologists!) eventually, but I seriously doubt that will happen in your lifetime. In addition to scientific and technical problems, the financial and legal considerations will slow down implementation. For one, the scanners are expensive and you would probably need more than one for a relatively busy hospital. If the computer makes a mistake, who is responsible? The company that made the scanner? The programmers who coded the algorithms? Etc.

I suppose one day machine learning could "look at slides" and provide a diagnosis and staging/grading, which could be forwarded or inputted into the EMR/EHR where another machine learning algorithm could take that data and incorporate with the labs, rads, ekg's, pft's, notes, etc and provide suggestions or insights into the treating clinicians. Again, I feel pretty certain that's not happening anytime soon.

 

[Med Director New England]

So we are talking about two things now

- Slide scanning to create quality digital slides that can be reviewed remotely without a scope allowing more flexibility with remote consultants, TC/PC arrangements where the technical lab is remote but digitizes the slides for the PC interp back at the original hospital and thus more consolidated TC -- This is already happening with IHC and will be come common for H&E diagnoses within next 5 yrs.

- AI with machines analyzing said digital image and rendering diagnoses. This is far out in the future, wont affect anyone practicing today.

 

[gbwillner]

I want to sign out while sipping on a pina colada on the beach. So I am all for this. I still have to figure out how to designate a portion of the beach as a CLIA-certified laboratory.