Q: Do you have special a special recommendation for labs and countries where they still have to start with building networks?
A: I would suggest to built a network for pathology reports first. Then proceed with WSI.
Q: Is reporting on WSI legally admissible as avalid document? If yes, in which countries of the World?
A: That question is very out of my confort zone! I know of the FDA in the US. Different countries may have different legal bodies or advisory bodies on this. For digital cytology, in NL we only need a quality certification by the labs. They should demonstrate it is at least as good as an existant mode. Of course, rules for screening is a different matter altogether, that is usually in the hand of health authorities (being healthy citizens involved)
Q: As AI becomes more developed, where will the central software tools be? LIMS systems that now can have full cancer data set and proprietary data also present, and also potentially integrating with genetic data? Digital pathology image platforms? Or standalone packages that can extract such data from multiple sources? How can we best future proof ourselves across the spectrum of IT infrastructure in pathology?
A: probably a federated system seems the best solution, where labs/ hospital bring their data into a accessible portal. So, most data remain where they are, but are collected and assembled into display tools where the patients is in an MDT or so.
Q: What is the best slide scanning equipment for simplicity of use & costs
A: It is difficult question to answer.
If you want to use for any type of slides BF, FL, Histology, Cytology, etc, I like 3DHistech Pannoramic 250 Flash III. But for Z-stacking, there is something you need to be aware.
HamamatsuHT2.0 was similar to this. but discontinued unfortunately.
Both are relatively user friendly and stable. Easy to share the images and annotation with other groups. We use both for annotations for AI development.
Cost is 200k?
If you want to use for clinical, Philips is very easy to use but the cost. No training needs to use.
We are going to have an opportunity to use Bionovation’s scanner. It is much more reasonable than other scanner and fast. Please contact me if you are interested in knowing our experiences with this scanner. Probably Feb – March next year.
Q: To Anna Boden: thank you for an exellent talk. Did you digitize sections at x20 magnification? and how long do you keep glass slides in the archive?
A: From start we digitized all WSI x20, today we have a proocol depending on area and specimen type either x20 or x40. E.g. prostatebiopsy; x40, prostate large section x20. At the moment we store all imagedata with no time limitations but will hopefully finalize an aproved new stratgy for storage next year.
Q: How much is the annual cost of implementing digital pathology in Dr Boden's lab?
A: The annual cost is hard to estimate since it is not separate from other lab acitivities nor diagnostic on personal level nor impact on IHC, archive and MDT activities but roughly €500 000 for storage, scanners, IT systems and workstation (for whole department)
Q: The annual cost is hard to estimate since it is not separate from other lab acitivities nor diagnostic on personal level nor impact on IHC, archive and MDT activities but roughly €500 000 for storage, scanners, IT systems and workstation (for whole department)
A: 96.1 (0.9 discrepancies + 3.0 deferrals)
Q: Aren't these problems generic and solved in other systems? E.g. doing the work of LIMS LIMS integration engines, which do exchanges across other physical test requests and movement of results in blood sciences and immunology, though not yet cell path? E.g. the NPEX system of XLabs in the UK?
A: I'm not sure if I understand your question. In NL, clinical immunology lab do not have a country wide network. Similarly for clin chemistry. Radiologists struggle with similar issues. Xlab sounds like XNAT what the radiologist use, but that' s only for research
Q: How many of the skin cases in the 99% concordance were dysplastic naevi or Spitzoid lesions?
A: We have not assorted all the areas in subgroups of diagnosis yet. For skin area there were in total 219 cases (13% of all cases). For the skin area 94 (43%) where categorized as malignant and 124 (57%) as benign.
Q: To Dr Bodén: how did you solve the problem of archiving the digital slides? Thank you
A: The images are archived in local storage environment designed and supported by the hospital IT department in a VNA using propriatare scanner format. The storage is bound to local servers at the moment but different secure cloud solution might be adapted to in future. The issue of costs and storage are under constant revision and we are looking into shorter archiving time as well as access to storage in different levels (tiered) with aim of lesser storage costs.
Q: from Andrey Bychkov (Director of Digital Pathology at Kameda Medical Center, Japan) to Dr. Anna Boden >> I. Practical questions: - do you use tiered storage (hot + cold) or all WSIs are immediately accessible? - what is your current archive strategy, i.e. store all WSIs since 2016 or delete some old slides? if the latter, please elaborate; - did you go digital beyond routine Sx and Bx specimens? e.g. frozen and cytology [to be cont'd below]
A: No at the moment we do not use tiered storage but the IT department is requested to look into this. Also, I want to delete WSI based on time of latest scanning time on WSI level as long as we have glass slide stated as the original, but there is a lot of regulatory issuses to clear out on national as well as international level. How do you handle this questin in Japan? Yes, we are currently also scannig frozen sections and some cytologi as well as internal QA cases, all included in the total storage.
Q: II. Conceptual questions: - it is not clear what is a scope and practical utility of such an extensive validation project (cont'd for almost 5 years)? I guess that concordance rate of 99% was obvious from the very 1st year of the study - what is a motivation for your staff pathologists to continue doing this exhaustive study? - have you noticed any difference in results with regard to different scanner vendors?
A: Exhausitve yes, aim is to involve in and get acceptance to implementation. Slow progression is due to no resourse of extra pathologists during validation, they need to do workactivities parallell. Also, you need every pathologist to perform their own personal validation. Scanners (vendor based and modelbased) do affect image quality mainly in color but also details, also depending on chosen resolution on pixel level in scanner and screen level. Image variation is also depending on decompression of fileformat and viewer functionality and so on. I do not think there is any guidelines for the most optimized level on pixeldata for pathologist perception vs AI algorithms?
Q: For Dr Boden, How often do you go back to digitized archived cases? Is the cost for archiving justified? How long should digitized slides be archived in your opinion?
A: We have looked into logs and there is a decrease after 2-3 years on a slide and case level how many slides are pulled out from digital archive in relation to all cases. Asking pathologists the digital archive is anyhow very appreciated and need to be weighted to access of physcal archive and how do we value the benefit of direct access on a patient level? In my opinion we should store digital slides in a hot and cold storage model and in total for 10 years to cover for data regulations as well as clinical and research need.
Q: Is it important to open the blackbox of AI to understand or explain how it hes been working
A: It has been shown that the trust in AI in Health by users is mainly hampered by the black box nature of AI. Providing some insight in why the decision was made or what it was based upon could improve trust and thus acceptance and use in clinical practice.
Q: For Dr. Boden : How do you manage frozen section examination ? Or do you still do the diagnostic by glass section?
A: Frozen sections should be scanned but we do frozen sections on glass slides if pathologist is in house and timelimitation is too tight. It is tricky to add scanning and import of slide to VNA as fast as you wnat. On the other hand consulting forzen sections is faster on a digitalized slide.
Q: To Dr Yagi - did you develop your own universal WSI viewer, and what were the challenges here? Is the LIS integration commercial or prorietary, and again what were the challenges here? Would you recommend sticking with single vendors, or developoing VNA for geographical digital pathology networks?
A: Yes. Our department has developed.
• Supporting all scanners we use.
• Integration with LIS.
• Normalize pixel resolution between scanner (0.46um/pixel = 20x for all scanners)
• Speed of viewing WSIs and network traffic. We work with hospital IT to improve
the speed. (upgraded network if necessary)
• To make reporting problem process easy for users, added reporting feature in the viewer.
Use just need to click smile or unhappy face icon in the viewer if there is any speed issue
Image quality & missing tissue issues also can be reported from viewer by just one click
I think basic LIS integration is commercial but all hospitals have slightly different requirements
For the integration. This part is proprietrary. Such as how many barcodes to read,
the location of Barcode to read. Also how to display about the WSIs are available in the LIS.
WSI vendors have their image/scanner management system.
One IMS system has to manage more than 10 of their scanners in the hospital.
LIS and each IMS system are communicating to give us the right images in the right places with
The right information. Depending on the vendor, the challenges are different.
It is most tough part.
If you do not have a strong IT support or personnel, better to stick with one vendor.
Hospital cost is also lower if you work with only one vendor (for LIS integration cost).
Q: Dr. Yagi: Efficiency is very important for pathology departments. Q1) Do sign-out times improve with longer experience? Q2 Are some viewer systems better than others for speed?
A: 5. Efficiency is very important for pathology departments
Yes I agree.
Q1) Do sign-out times improve with longer experience?
Yes. By experience, the speed and confidence are improved.
At the same time. We watch computer spec, monitor, human/user interface to improve more.
Q2 Are some viewer systems better than others for speed?
Yes, very different. Often, it is not only by viewer. System structure in the hospital.
(we are not using WSI for primary diagnosis yet except COVID circumstance early this year although
Moving to that direction. All pathologists are using WSIs daily basis with LIS. Right now it is more for references than primary diagnosis. It will make smoother transition to fully digital operation in near future.
Q: For dr Yagi - how do you cope with the concept (aparent) magnification - resolution and image quality? Do you believe that pathologist realize that magnification is not equal to resolution?
A: Yes recently most pathologists know it.
Also our viewer normalize any of pixcel resolution images to 0.46um/pixel =20x
0.23=40x. Not only pixel resolution, they understand optical part like which scanner has true 40x objective lens and also NA.
But display resolution 4k monitor vs 2k monitor confuse them more depending on how to show.
Q: Dr Boden: your example of the biopsy where only two cuts were shown in the scanner, is a bit worrisome. Does that happen frequently?
A: No this is very rare and the example is old, but unfortunately it still happens sometimes on small fragments. We have manual quality control before slide is taken from scanner (especially if automatic scanning mode is initiated), but as you know human visual errors sometimes apprear as well as automatic. Other lessons learned is that new scanners should have a automated quality contol system to compare snap shot from slide with scanned image and rescan automatically. Also in the viewer you must have access to snap shot of slide for manual visual comparison if necessary for confirmation (both of label and tissue).
Q: To van der Waal: to what extent is the differentiation of types of MI actually used in the autopsy protocols? And do you have any possiblity to search for the cases afterwards - for the purpose of research?
A: The 5 clinically distinct types of are not mentioned in the protocols; the protocols provide the information on how to investigate and document the myocardium for occurrence of specific types of myocardial injury (ischemic and others), how to investigate coronary artery pathology, eventual cardiac remodelling changes such as hypertrophy and dilatation which may affect microvascular myocardial perfusion, and the extra cardiac pathology which may affect myocardial perfusion. These are mentioned in the tables of the new (2020) ACVP article on myocardial infarction. Then after, and importantly in combination of the clinical manifestations of the deceased person, it may lead to a discrimination between the different clinical types of MI, and to to see how they contributed to disease progression or death. At this time, as i am aware of, there is no ongoing sytematic research started on autopsy cases.
Q: Preston: Do you know how many abnormal coronary arteries are diagnosed in the living vs. at autopsy?
A: I’m afraid I don’t know, but clinically important ones are often diagnosed in life for symptoms, particularly ALCAPA which usually presents in the neonatal period.
Q: To Preston (or all): Your lectures clearly show that expertise is crucial, esp. in cases of "sudden deaths". Often, PM are performed by residents working alone, or even TAs. What do you think about turning Autopsy pathology into a subspeciality of pathology.
A: In the UK, not all pathologists now undertake autopsies, and I think we need to have people who are interested in autopsies – but the range of pathology is so wide at autopsy that overlap with other subspecialties is also helpful. Senior supervision is important, both for a good autopsy and to ensure the resident can learn in my opinion. I think that good training and commitment to doing it well are more important than how specialties are organised. AvW: I agree with Stephen, although in our institution we try to form a subgroup of pathologists, who are specifically interested and trained in autopsy pathology. What is the opinion of the ESP Working group of Pathology?
Q: Question to anyone or all: How do you examinate the coronary arteries during the autopsy? Do you make cross-sections or cut through the lumen?
A: Cross section through the coronary arteries. The AECVP paper describes a good method of examination.
Q: Dr Giordano. Excelent talk. Do you have experience using IHC for C5b9 compared to C9 in evaluation of acute myocardial infarction? Thanyou.
A: Unfortunately I do not have experience with C9 IHC. Actually a combination of fibronectin (early marker) and C5b-9 (more sensitive and specific) IHC is suggested. Also C4d may be used although exact timing is not known.
Q: Question from a non-academic general pathologist to the panel, In case of a suspected sudden cardiac death without obvious coronary atherosclerosis/thrombus: - Should we use IHC to at least try proving early myocardial infarction? And which one (as they are not very specific) would be the most 'reliable'? C4d?CRP?other? - Should we be advised to allways analyse the conduction system as well? (Is it enough then to mainly focus on retrieving and analysing the AV node?)
A: A combination of fibronectin (early marker) and C5b-9 (more sensitive and specific) IHC is suggested.C4d may be used on paraffin sections, however it is not clear the timing. Very important is the topographic pattern of staining (regional and/or subendocardial) in order to define the ischemic etiology of myocardial injury. About conduction system please refer to: Guidelines for autopsy investigation of sudden cardiac death: 2017 update from the Association for European Cardiovascular Pathology. Virchows Arch (2017) 471:691–705
Q: Has the utilisation of post mortem radiology been seen in Covid-positive cases?
A: In our experience, postmortem radiological findings are not specific
Q: For katarzyna Michaud: can you comment on how timing of execution of post-mortem imaging can impact on the quality of imaging and their interpretation?
A: deally, postmortem imaging should be performed asap. However, some patients are found days after the death and it is still possible to perform even postmortem CT- angiography
Q: Do you recommend IHQ panels (C5b-9, Fibronectin) to detect early myocardial injury? What is your experience?
A: Actually a combination of fibronectin (early marker) and C5b-9 (more sensitive and specific) IHC is suggested. Personally I also perform Cd4 on paraffin sections. What should be clear is that positive stain by itself does not allow the diagnosis of ischemic injury. The topographic pattern (regional and/or subendocardial)is important
Q: For Carla giordano: Excellent talk: can you comment/expand on pitfalls related to postmortem autolytic processes in MI interpretation? Thank you
A: Autolysis manifests with loss of nuclei and eosinophilic changes. The diffuse, not regional pattern of features and the absence of other changes (PMN marginations,.) help to exclude diagnosis of AMI.
Q: When using CTA, how do you know you will not push away plaques (of push them further downwards where they will get stuck anyway)?
A: Of the isolated heart at autopsy: I think there is no risk of pushing away plaques, but potentially it could happen with fresh thrombus depositions. Important is to first inspect the coronary ostia, before introducing the injection needle, and inflate the contrast fluid slowly and gently, or preferably under physiologic pressure (100mnHg) using a fluid column for inflation. This procedure takes circa 10-15 minutes of inflation.Also in postmortem whole body imaging, the pressures should not allow the displacement of ante mortem thrombus
Q: Could you expand on the "brain autopsy was not allowed"? Is consent given for autopsy of separate organs?
A: In our hospital, we have to ask permission explicitly for autopsy of either the body or the brains, or both. In some instances, it is even possible that we receive the permission only for heart and lungs. The decision is to the relatives of the deceased (in a setting of clinical autopsies)
Q: For Stephen preston: Excellent overview of a difficult topic: Do you have experience with post-partum dissection with presence of severe eosinophylic infiltration?
A: We’ve had a couple of cases of dissection, although I don’t think they were post partum, with prominent eosinophils. This is a difficult area – pregnancy is such a strong risk factor that it would suggest that the eosinophils are reactive, but I don’t see prominent eosinophils in aortic dissection. What does the panel think?
Q: Are seremos troponines useful after sudden Death or are masked by CPR?
A: According to the published studies, serum troponins (currently high sensitive troponin) could be useful and they are not influenced by CPR
Q: Excellent presentations and a superb session. For Carla Giordano: How far down the coronary arteries from the ostia do you consider atheroma to be causing significant occlusion?
A: Any acute plaque event (rupture or erosion) independently from the distance from the ostium should be considered potentially related to the cause of death
Q: For Stephen Preston: How would you sign out a post mortem case of significant coronary bridging with no myocardial fibrosis in the absence of another cause of death?
A: Carefully! Coronary bridges are common, and most don’t cause problems. I would look for signs of turbulence, such as depth and length. Does the history fit – bridges appear to be mainly a problem during exertion. The absence of acute infarction does not exclude ischaemia, as Dr Giordano showed, so I think a myocardial bridge would be a diagnosis of exclusion as a cause of death (including toxicology), for example it is common with hypertrophic cardiomyopathy
Q: Allard do you mean that even distal small coronary arteries narrowing is significant in absence of infarction. Lovely talks by everybody
A: To my opinion narrowing of small coronary vessels is important when it affects the critical perfusion gradients in the myocardium, especially the subendocardial layers