- Home
- Member Resources
- Articles
- Twitter Chat Wrap-Up: Digital Pathology and A.I. – How Pathologists Can Leverage Emerging Technologies
On Nov. 18 at 9pm EST, I took over the @Pathologists Twitter handle to host a Twitter chat about opportunities for pathologists to implement digital pathology and artificial intelligence into their practice—both now and in the future. I was joined by expert guests Eric Glassy, MD, FCAP (@apmg_glassy); Marilyn Bui, MD, FCAP (@DrBuiPathology), and Matthew Hanna, MD, FCAP (@MGHannaMD).
Below are some highlights from the hour-long discussion. Also, be sure to scroll to the bottom of this blog to view additional resources related to digital pathology and A.I.
- Open all
- Close all
Hello everyone, welcome to tonight's #capchat I am a pathologist @MSKPathology excited for the discussion
— Matthew G Hanna (@MGHannaMD) November 19, 2019
I am using image analysis, not really AI. #capchat
— Eric Glassy (@apmg_glassy) November 19, 2019
Hola! Paul Gibbs from Scripps Clinic in San Diego. User of digital path for tumor boards, teaching, and occasional image analysis. Can't wait to learn from these leaders! #capchat
— Paul M. Gibbs (@gibbspath) November 19, 2019
Jason Scapa here. Surgical Pathology Fellow at Stanford and Former Resident at UCLA. We used digital path at UCLA to scan all biopsies, interesting cases, tumor board cases, and received consults from China through digital. Stanford signs out gyn and neuro via digital. #capchat
— Jason Scapa, MD (@JScapaPathMD) November 19, 2019
#capchat Hello, Mike Bonham, Pathologist, Chief Medical Officer, Proscia #proscia
— Mike Bonham (@mikebonhamsf) November 19, 2019
>Ingram here. Community pathologist in Charlotte, NC. My group uses image analysis for quantitative breast biomarker IHC, but not #AI as of yet. #capchat
— Frank Ingram, MD (@Chucktowndoc) November 19, 2019
I am also actively and deeply involved with DPA and CAP digital pathology committee. Please feel free to ask me questions about these organizations too #capchat
— Dr. Bui Pathologist (@DrBuiPathology) November 19, 2019
Sara Jiang, @DukePathDept, we use digital path and AI for various clinical care and research applications #capchat
— Sara Jiang, MD (@Sara_Jiang) November 19, 2019
I had a meeting last week with some representatives who wanted to work with me regarding breast biomarker analysis using whole slide imaging and AI (we use an earlier version now). Digital/AI is an inevitable next-step in #pathology as a supplement so we must embrace it #CAPchat
— Alexander Damron, MD (@ADamronMD) November 19, 2019
Eve Crane @WCMCPathology, late to the discussion, but enjoying all the tweets on this important topic. Working to expand the use of digital pathology in #hemepath with @sanamloghavi @feldstej with guidance @mydermpath #capchat
— Genevieve Crane, MD, PhD (@evemariecrane) November 19, 2019
A1. A CLIA license is needed to do dig path primary signout. Licenses are based on location. Signing out from home requires a CLIA license. But “infrequent” signout (<10%) is OK. Be sure to adhere to HIPAA regs. #capchat
— Eric Glassy (@apmg_glassy) November 19, 2019
23s
— Matthew G Hanna (@MGHannaMD) November 19, 2019
A1. Pathologists using WSI systems under CAP validation guidance can use their laboratory's existing CLIA license to report patient's pathology cases using digital pathology #capchat
If I understand correctly, cases can be reviewed off-site digitally (say the beach for example) but verification must be done on site where the existing CLIA license is located, right? #capchat
— Adam L. Booth, MD (@ALBoothMD) November 19, 2019
Actual signout requires CLIA certificate. You can review anywhere. But you can signout cases infrequently without a CLIA license (<10% of cases). #capchat
— Eric Glassy (@apmg_glassy) November 19, 2019
Things will get complicated when cases are uploaded to the cloud and AI performed or the signout happens in the cloud. CLIAC has a workgroup looking at this, recognizing that modern digital workflow does not fit the norm. #capchat
— Eric Glassy (@apmg_glassy) November 19, 2019
A2. The FDA regulates how a manufacturer markets scanning devices. It does not regulate how a pathologist uses that device. It is OK to use a non-FDA approved system for primary signout as long as proper validation and training is conducted. #capchat
— Eric Glassy (@apmg_glassy) November 19, 2019
A2. Regulatory bodies tell vendors/manufacturers how they can market their product(s), FDA does not regulate healthcare/healthcare professionals #capchat
— Matthew G Hanna (@MGHannaMD) November 19, 2019
A2. There is always confusion about this topic. The FDA does not regulate what we pathologists do. We can use any tool (appropriately validated) we want for primary DX.
— Eric Glassy (@apmg_glassy) November 19, 2019
Are there other systems besides @Philips_Path that are FDA approved currently? #capchat https://t.co/1rCJbWHQ9H
— Michael Misialek, MD (@DrMisialek) November 19, 2019
A2. We used a Leica based system at UCLA which I believe was recently FDA approved. Stanford uses the FDA approved Phillips setup. Love to hear thoughts on the advantages/disadvantages to both. @apmg_glassy @DrBuiPathology #capchat
— Jason Scapa, MD (@JScapaPathMD) November 19, 2019
A2. Pathologists should validate dig path systems under CAP/ use professional society guidance. There are laboratories reporting dig path using FDA approved systems and also not using FDA approved systems with validation #capchat
— Matthew G Hanna (@MGHannaMD) November 19, 2019
A2. FDA approval is for primary diagnosis of surgical pathology, this does not cover frozen section or 2nd opinion consultations #capchat
— Matthew G Hanna (@MGHannaMD) November 19, 2019
#capchat . Curious on everyone's thoughts for how AI may impact pathology or not.
— Mike Bonham (@mikebonhamsf) November 19, 2019
The most exciting applications for AI for me-and the ones that seem most feasible for adoption anytime soon-are the ones which streamline workflow-counting mitoses, identifying areas of interest like PNI, LVI, micromets. #capchat
— Sara Jiang, MD (@Sara_Jiang) November 19, 2019
How about intelligent workflow, presorting cases, marking ares of interest, ordering stain upfront, not all pathologist have as great residents as you do :)
— Mike Bonham (@mikebonhamsf) November 19, 2019
sounds good, gathering history, ordering molecular #capchat maybe AI informatics IT
— Julie Feldstein,M.D., FCAP (@feldstej) November 19, 2019
I agree. I see this as the first step. One conversation I had was someone saying AI is right 80% of the time regarding being vs malignant and my response was well we need to be as close to 100% as we can because every % is a patient. The first step I agree is streamline & preview
— Alexander Damron, MD (@ADamronMD) November 19, 2019
Benign** I got autocorrected!
— Alexander Damron, MD (@ADamronMD) November 19, 2019
Believe it would be a good supplemental tool
— Evita Henderson Jackson (@EvitaHe57886045) November 19, 2019
Did you know the adoption of AI is beginning to happen in Europe right now . There are large multiyear clinical studies ongoing and commencing. Helps to be digital.
— Mike Bonham (@mikebonhamsf) November 19, 2019
Implementing digital pathology has been shown to increase efficiency and operational utility in this @ArchivesPath study https://t.co/Fj3XRi6QPR #capchat pic.twitter.com/LnSyj9xONe
— CAPathologists (@Pathologists) November 19, 2019
Any advice for us pathologists working in community/private practice? Digital pathology and A.I. sound great, but it seems daunting to bring up in smaller practices. Especially as a new-in-practice attending. #capchat @DrBuiPathology @apmg_glassy
— Tyler Teichmeier M.D., MLS(ASCP) (@TylerTeichmeier) November 19, 2019
Still early days, most important is belief and desire to bring change. Companies are hungry to prove out technology, we cant do it without pathology buy in. As a new pathologist, this is an opportunity to take advantage of.
— Mike Bonham (@mikebonhamsf) November 19, 2019
Tyler you are perceptive. This is a tough subject the cost and benefits are tricky to calculate. It’ll take time before this becomes commonplace.
— Mick Raich (@MickRaich) November 19, 2019
companies like Protean BioDiagnostics are focused on helping small practices access advanced technology https://t.co/Z9oIZp0Aq9
— Anthony M Magliocco MD (@MaglioccoTony) November 19, 2019
If you are interested in a fairly straightforward whole slide viewer/annotator, #QuPath is both free and open source. Of course, that is only useful once you have whole slide images, which may be more the problem.https://t.co/8aPwsCKiPqhttps://t.co/LZ39lSaHL3
— Still Sciencing (@StillSciencing) November 19, 2019
A3. All medical student histology labs at UCLA are "microscope free" and only use digital slides for teaching the preclinical curriculum. #capchat
— Jason Scapa, MD (@JScapaPathMD) November 19, 2019
Ours are too @DukeMedSchool and have been for several years. Frankly, this dinosaur is a bit sad about that, but that’s another discussion #path2path #capchat
— Sara Jiang, MD (@Sara_Jiang) November 19, 2019
A3. Other WSI resources are available on PathPresenterhttps://t.co/C96JPCO9iy#capchat
— Matthew G Hanna (@MGHannaMD) November 19, 2019
Example image with link Breast biopsyhttps://t.co/QckiirVtkb#capchat pic.twitter.com/2rGMWbvHMw
— Matthew G Hanna (@MGHannaMD) November 19, 2019
A3. There are many. Tumor board, case of the month, image repositories, virtual micro courses, etc. Medical schools have abandoned scopes. CAP has virtual PT products like blood smears, bone marrow, fluids, surgical bxs and HistoQip. #capchat
— Eric Glassy (@apmg_glassy) November 19, 2019
A3. Take a look at this book: over 100 WSI links to peripheral smears. (Shameless plug)#capchat pic.twitter.com/2f071efcxG
— Eric Glassy (@apmg_glassy) November 19, 2019
Here is a sample with WSI link. #capchat pic.twitter.com/4Y2D7n5JXv
— Eric Glassy (@apmg_glassy) November 19, 2019
Yes! As a member of @Pathologists Surg Path Cmte I have to give a plug for the Virtual Biopsy Program (VBP) and PIP-W! https://t.co/fVylvr7jcb #capcast
— Sara Jiang, MD (@Sara_Jiang) November 19, 2019
It is a little sad that Medical Schools have abandoned microscopes. However, don't let them throw them away! Donate them to nearby schools! You never know, that may be the gift that turns someone onto a career in pathology! #medtwitter @ALBoothMD #RFEC #capchat @UNMC_PathMicro https://t.co/DzCZkF2JFC
— Tyler Teichmeier M.D., MLS(ASCP) (@TylerTeichmeier) November 19, 2019
Other than online repositories how can Residents best prepare for the changes to practice if their programs are not using dig path or AI? @MGHannaMD @DrBuiPathology @apmg_glassy #capchat
— Adam L. Booth, MD (@ALBoothMD) November 19, 2019
Adam-- new hires are expected to know molecular and dig path. If your program does not have a scanner, strategize how to get one. Work to get a grant. Donation from a wealthy donor looking to put family name on a beautiful scanning system that will honor a loved one. #capchat
— Eric Glassy (@apmg_glassy) November 19, 2019
Fortunately my program has scanners we use for tumor boards, education, and we’ve recently validated use for frozens on different campuses.
— Adam L. Booth, MD (@ALBoothMD) November 19, 2019
Really helps to go to some of the DP meetings. Pathvisions, Pathinformatics. Follow the leaders on social media, ask questions. It is wide open.
— Mike Bonham (@mikebonhamsf) November 19, 2019
Residents, don't forget that there will be digital slides on the @TheABPath Boards! Need to get used to the interface and how to make diagnoses off of digital slides! #capchat
— Jason Scapa, MD (@JScapaPathMD) November 19, 2019
The @theABPath website has a link to practice virtual microscopy. The DigitalScope software shown is similar to what is on the exam. #capchat #followup https://t.co/BnRCcnAEoj
— American Board of Pathology (@TheABPath) November 19, 2019
#capchat get involved with DPA which gives you a greater chance to connect with pathologists who are using these tools. Consider an outside rotation in these places. Personal scanners are also available. DPA is also working provide more opportunities for residents and fellows https://t.co/vhOlBpOFxW
— Dr. Bui Pathologist (@DrBuiPathology) November 19, 2019
can we link flow, FISH, karyotype, NGS? #capchat
— Julie Feldstein,M.D., FCAP (@feldstej) November 19, 2019
Julie-- any digital image can be linked. An important use case is radiology-pathology integrated reports. Dean Wallace at USC and others have great demos, using custom Epic software. #capchat
— Eric Glassy (@apmg_glassy) November 19, 2019
#capchat . Tell me the one change you would most like to have as a pathologist. Pathologists need to be more vocal about what would make their practice better. In a perfect digital world, my signout is like ....
— Mike Bonham (@mikebonhamsf) November 19, 2019
In a perfect digital world, my sign out is efficient. I would be able to have all of my cases downloaded in th emorning and be able to access them between meetings. It would also be nice to access previous cases online (instead of bothering the slide room to pull them). #capchat
— Tyler Teichmeier M.D., MLS(ASCP) (@TylerTeichmeier) November 19, 2019
functional implementation #capchat
— Julie Feldstein,M.D., FCAP (@feldstej) November 19, 2019
Also interesting thought is digital path integration into the EMR. With Epic AP Beaker, we could launch the Leica Apiero online image scope right from the case. Curious 2 see other ways to integrate the clinical EMR with digital path...could become a PACs viewer 4 path. #capchat
— Jason Scapa, MD (@JScapaPathMD) November 19, 2019
LIS-centric workflow launches WSI from LIS and reporting still in LIS. PACS-driven workflow could be possible in the future.#capchat
— Matthew G Hanna (@MGHannaMD) November 19, 2019
Hybrid human and AI workflows seem to be key for human acceptance of AI. What applications are most suited for a human-AI workflow? #CapChat @Pathologists
— Motic Digital Pathology (@MoticPathology) November 19, 2019
#capchat lower level AI include quantification, classification and locating which are more acceptable by human because AI is reproducible, accurate and tireless; but providing diagnosis, prognosis and prediction is much higher level, AI is not there https://t.co/dUBt4ST0KE
— Dr. Bui Pathologist (@DrBuiPathology) November 19, 2019
A5. Best use case satisfies the intended use. So many applications for digital pathology.
— Matthew G Hanna (@MGHannaMD) November 19, 2019
Primary diagnosis, telepathology/consultation, machine learning, education, research, etc #capchat
A5. For me, best use case right now is telepathology, as @MGHannaMD notes. Other use cases depend on what you want to get done. My favorite is tumor board. Lots of positive comments from clinical staff. One quote: "Best tumor board I ever attended." #capchat
— Eric Glassy (@apmg_glassy) November 19, 2019
A5. #capchat. Depending on the situation. ROSE, telepath for frozen and consult solve problems of shortage of expertise. Image analysis and AI produce produce accuracy and efficiency https://t.co/kWemyOcVck
— Dr. Bui Pathologist (@DrBuiPathology) November 19, 2019
A5. #capchat archiving the slide to reduce the time and labor to retrieve the slide is a great use case for some to digitalize https://t.co/ZQHIvkxDI5 pic.twitter.com/Ho4iCWpCY2
— Dr. Bui Pathologist (@DrBuiPathology) November 19, 2019
Yessssss...can't tell you how often I wish the prior biopsy/case was easily accessible without having to hunt for the slide. UCLA used this very effectively for transplant biopsies and breast cores to compare with excisions. #capchat https://t.co/rY3czB5M8Z
— Jason Scapa, MD (@JScapaPathMD) November 19, 2019
We do this @MSKPathology and it has transformed our workflow #capchat
— Matthew G Hanna (@MGHannaMD) November 19, 2019
As someone who is responsible for a fair number of tumor boards, this is the bane of my existence. #capchat
— Tyler Teichmeier M.D., MLS(ASCP) (@TylerTeichmeier) November 19, 2019
We also had a cabinet fall over that had thousands of slides. Many lost/broken. A digital copy would still be around #capchat
— Steven Hart (@StevenNHart) November 19, 2019
A6. Current CAP virtual PT products: surg path biopsies, cytology, virtual hematology (blood and marrow), flow cytometry, microbiology and reproductive med. All use DigitalScope viewer. #capchat
— Eric Glassy (@apmg_glassy) November 19, 2019
A6. CAP 2020 Surveys include digital pathology https://t.co/6PiCTvDyR5#capchat
— Matthew G Hanna (@MGHannaMD) November 19, 2019
#capchat digital images would be great for GYN proficiency testing. Easily test candidates from their computer.
— Evita Henderson Jackson (@EvitaHe57886045) November 19, 2019
Evita-- virtual pap smears were piloted by CAP a few years ago. Not sure of current status. Cytology glass PT does have some sample links to virtual smears that are not used for grading but no totally virtual Pap PT product yet. We should lobby for one. #capchat
— Eric Glassy (@apmg_glassy) November 19, 2019
#capchat A6. This is new and the CAP DP committee worked with HistoQip program developed this. Advantage: more meaningful feedback because we can annotate the slide! https://t.co/dOhlprJ3TT pic.twitter.com/XCsXWdEfcL
— Dr. Bui Pathologist (@DrBuiPathology) November 19, 2019
How many have printed out an image of ki-67 then manually counted with a sharpie? #capchat https://t.co/eKqELLGaPR
— Michael Misialek, MD (@DrMisialek) November 19, 2019
Excellent method- very helpful to do this way!
— Rebecca Osgood (@OsgoodRebecca) November 19, 2019
I've done this, and I've also used QuPath, which does a brilliant job at Ki-67 quantitation as an unofficial crosscheck of my manual count. #capchat
— Paul M. Gibbs (@gibbspath) November 19, 2019
But when I look at the CAP checklist requirements for validating QuPath or a like image analysis, it's a huge task relative to the times per month my group will use it.#capchat
— Paul M. Gibbs (@gibbspath) November 19, 2019
Does anybody have recommendations on an efficient validation process for Ki-67 by image analysis?
That is how we grade our Neuroendocrine neoplasms at MSKCC.
— Olca Basturk (@OlcaBasturk) November 19, 2019
I’ve done that Michael with Ki-67. I’ve also tried Qupath and ImageJ but can’t figure it out.
— garrey faller (@garreyf) November 19, 2019
#QuPath has the option for both automated quantification and manual tracking with the Points tool (and then it keeps track of the counting!), if you want to compare results. Questions at https://t.co/h0ZDOatyzC
— Still Sciencing (@StillSciencing) November 19, 2019
This is what I currently do for my 500 cell count!
— Sara L. Zadeh, MD (@missDoctorZ) November 19, 2019
@Pathologists what do you think is the greatest hurdle to digital pathology and AI? #capchat
— follow_ur_path (@FollowPath) November 19, 2019
For, one of the biggest hurdles is...ourselves. We need to adapt and be willing to tweak our workflows to accommodate digital. Cost is always an issue but personal scanners are available. AI will help drive adoption. #capchat
— Eric Glassy (@apmg_glassy) November 19, 2019
#capchat 4 big barriers, regulatory, technology, $$ and human, the later is the determining factor https://t.co/cNy6biurgn
— Dr. Bui Pathologist (@DrBuiPathology) November 19, 2019
For those in private practice- making the business case for digital pathology to administrators is difficult. #CapChat pic.twitter.com/DtixKC6xa4
— AmyHDeekenMD (@AmyHDeekenMD) November 19, 2019
Convincing leadership to invest #capchat
— Evita Henderson Jackson (@EvitaHe57886045) November 19, 2019
Currently cost for small groups. The rhetoric by some that #AI will replace us is also a bit of a turnoff. #capchat https://t.co/1E2xGNSQZW
— Frank Ingram, MD (@Chucktowndoc) November 19, 2019
Frank-- Replacing pathologist just is not in the cards. The current crop of AI and machine learning tools are weak AI and will only make us more efficient. #capchat
— Eric Glassy (@apmg_glassy) November 19, 2019
Agreed, data supports pathologist + machine > pathologist alone #capchat https://t.co/fD7z8qNloi
— Matthew G Hanna (@MGHannaMD) November 19, 2019
Agree most current solutions are weak but #AI capabilities are changing fast. Solutions that aim for efficiency, QC, and enhanced visualization will open up opportunities for pathologists #capchat
— Julianna Ianni (@juliannalog) November 19, 2019
And reproducible. Not only can we do more, but it will help standardized interpretations (think ki67 for neuroendocrine tumors) so that patients and their clinicians known that their results are not only accurate but precise. #capchat https://t.co/A2MHFWFZst
— Jason Scapa, MD (@JScapaPathMD) November 19, 2019
How do you purpose we can change that?
— follow_ur_path (@FollowPath) November 19, 2019
Need physicians in leadership positions.
— Jen Buhay (@JenBuhayPhD) November 19, 2019
People, processes, technology. People are critical to ensuring digital pathology/machine learning success#capchat
— Matthew G Hanna (@MGHannaMD) November 19, 2019
Julie-- I love PathPresenter! What a great resource. Created by Raj Singh and Matt Hanna.
— Eric Glassy (@apmg_glassy) November 19, 2019
Also look at KiKo social media tool, developed by Jon Ho. #capchat
A7. CAP WSI validation updates open comment period has closed. CAP Digital Pathology Committee (https://t.co/7TIgdrPAYU) currently updating guidelines for publication https://t.co/xUNPlxWXuC#capchat
— Matthew G Hanna (@MGHannaMD) November 19, 2019
A8. Attend House of Delegates and Spring Leadership conference in May has a digital pathology/AI track. Educational offerings including webinars from the Dig Path Committee. #capchat
— Eric Glassy (@apmg_glassy) November 19, 2019
#capchat definitely going
— Evita Henderson Jackson (@EvitaHe57886045) November 19, 2019
Evita-- and join the DPA!
— Eric Glassy (@apmg_glassy) November 19, 2019
Marilyn Bui is current president.
#capchat
Yes, I am a member and get all my residents to become members to be so part of the next revolution in pathology
— Evita Henderson Jackson (@EvitaHe57886045) November 19, 2019
#capchat Thank you all for your engagement. Discuss dig path with your organizational leadership and challenge the status quo to move pathology into the modern era. Get involved and be an avid end user to influence the future
— Matthew G Hanna (@MGHannaMD) November 19, 2019
Radiology did this decades ago and look where they are with their digital and PACs based systems (how did any one develop MRI/CT on films?!). It made them so much more efficient. We should be next to embrace this technology in our field! #capchat https://t.co/sWDC58pRvw
— Jason Scapa, MD (@JScapaPathMD) November 19, 2019
Digital pathology represents the future of pathology. AI and machine learning will provide tools to improve patient care and make up better diagnosticians. The CAP will be there, helping us implement. Thanks for all the feedback. #capchat
— Eric Glassy (@apmg_glassy) November 19, 2019
Thanks to everyone who participated in last night's #capchat, especially our host @DrMisialek and guest experts @DrBuiPathology @apmg_glassy, & @MGHannaMD. Learn more about A.I. at the Pathologists Leadership Summit, May 2–5, 2020, in Washington, DC: https://t.co/o3RjQ2zfj6.
— CAPathologists (@Pathologists) November 19, 2019
Additional Resources
- Digital Pathology Association Whole Slide Imaging Repository
- CAP Digital Pathology Resource Guide
- CAP Digital Pathology Topic Center
- Implementation of Digital Pathology Offers Clinical and Operational Increase in Efficiency and Cost Savings; Archives of Pathology & Laboratory Medicine.
- The Pathologists Leadership Summit (May 2-5, 2020, Washington, DC) will feature digital pathology and A.I. courses.