The title, the Imaging Informatics Clinical Interspace Between Radiology and Pathology is a mouthful, but just remember radiology and pathology. What we're trying to do today is introduce you to the subject of integrated diagnostics, give you an appreciation of the role of imaging informatics in this integrated diagnostic paradigm, and also make some suggestions about how integrated diagnostics might be driven in your practice by radiology and pathology. Integrated diagnostics represents a vision for the future where data from in vivo and in vitro diagnostics together with clinical data from the electronic health record are aggregated and contextualized to enhance diagnostic insights and direct clinical action. It also implies a more holistic look at diagnostic testing from the pre-test phase, which primarily involves the referring physician and the patient, to the test phase, which does involve us and where we tend to concentrate our activities, but then also into the post-test phase where our test results are put into the electronic health record and used to help take care of the patient as well as, nowadays, the broader populations. In many practices today, we work like this. Laboratory medicine receives orders, does a variety of tests, and puts out some reports that goes back to referring clinician. Radiologists do the same thing. We get electronic orders entry, enter, we schedule them, we scan them, and then we put a report back out, send the reports out. And pathology, similar. And so there's a whole lot of reports that we generate, and then, exaggerating a bit, we kind of throw them over the wall to the referring physicians, our colleagues, and each other, but in a rather ad hoc way of managing what is becoming an increasingly overwhelming data pool for our clinicians to work with. And by the way, our reports also go out to our patients now, so it's even more confusing to them. What we're talking about integrated diagnostics is to where there's a unified front office, if you will, for the ordering of diagnostic tests, and the scheduling of diagnostic tests, and then an integrated diagnostic process that nobody sees, that's in the back office, that performs all of these tests, creates our individual reports, but then at the end, those reports are integrated with each other to give a more in-depth view of the significance of our results. And so that's a little introduction to integrated diagnostics. We're going to focus today on a part of this, which is imaging informatics related to, obviously, image data. And I'll now turn the program over to Dr. Carino, who is our moderator, and will take care of the rest of the show. Thank you. Hi. Good afternoon, everyone. I'm Marilyn Bui. I'm a practicing pathologist at Moffitt Cancer Center. Today we're going to talk about pathology imaging data, because today we're going to compare and contrast radiology and the pathology, and discuss how can we integrate these diagnostic imaging systems. So we're going to discuss the essentials and quality of pathology imaging. So I'm a professor of pathology, and I credit many of my accomplishments in digital pathology to the associations I'm affiliated with, including Digital Pathology Association, College of American Pathologists, and the SIM, Society of Imaging Informatics in Medicine, and also others. And I'm speaking from my personal experiences, many of the projects mentioned in this talk, and I am personally involved, and also some high-level information from what I learned being in leadership roles in these associations. So four things we're going to discuss. Review the essentials of pathology imaging data, highlights the potential impact of digital pathology and AI on precision medicine, discuss the quality improvement opportunities, and showcase, optimize the patient care by integrating radiomic, pathomic together. So in contrast to radiology, it's pretty much an in vivo imaging process. Pathology is very much in vitro. For example, the patient tissue needs to be taken out from the patient, and we go through a process to preserve the tissue, give it a lifelike appearance, and we will physically, mechanically section them into four to five micron section, so to a glass slice like that, and then we stain them. Our traditional stains are HE slides, and we use microscope reveal those slides to study the cells, tissues, and their relationship to their surrounding tissues. That's our first generation of pathology. Our first revolution happened when immunology and the immunohistochemistry was discovered. So we use immunohistochemical stains going through antigen retrieval and the staining, and then we create those brown stains. So now we can look at the tissue in a much deeper level, at the protein level. So this is a very complicated process. Many process are manually done. Some of them are automated using the stainers. With this very complicated process, it's very subjective to pre-analytical source of slide quality variation. Second is the pathology information data is very heterogeneous. For example, this one is a gross image, and it's a cross-section of a tumor, and this is a physically cut, formally fixed, paraffin-embedded tissue of cross-section of a tumor. So we can take a digital image of that. And this one is H and E. Okay, so this is H and E image. So this doesn't really work very well. All right, so this is the H and E image. You can see the blue and purple stain, and the brown one is the IHC image. So those are the microscopic images, and we can do special studies. For example, the third one is immuno-in situ hybridization, and we can also do cytology. So you can see on top here is cytology image, and over here at the right corner is another cytology image. Compared to the histological image, cytology image are thicker, three-dimensional. The cells kind of floating in between the two slides. So that makes digitalizing the slides a little bit challenging. And on the right upper corner is multiflex immunofluorescent. As you can see, we can test multiple markers from that. So it's very cumbersome for human eye to interpret that result. So that gives us the opportunity to use quantitative image analysis or even AI to do it. So we talk about our traditional H and E. Our first revolution is IHC. Our third revolution was molecular pathology, and the third one is now is digital pathology and artificial intelligence. What is digital pathology? So all those glass slides information you're seeing will be digitalized and convert into digital images. So that gives us opportunity to change the analog data in a digital data. So pathology is now speaking the digital language. With that, the pathology can be integrated into the electronic health system, can be integrated with the proteomic, the genomic information. So this gives pathology the opportunity to do image analysis and AI. So with the third revolution, that's genomics. So the pathology that have become more and more complicated. So we really rely on computational pathology and pathology informatics to really mind the data generated by pathology so we can make diagnosis, make predictions, and make prognosis. So this is the role of digital pathology and AI in precision medicine. Detection, find tumors, metastatic tumor in infant node, find the rare events like microorganisms. Second is quantification. For biomarkers like ERP or HER2, they're not just the immunostains. They're scoring and it's very important for how to treat the patient. So AI and the quantitative image analysis is very reproducible and robust. And the third is to do classification, tumor type, tumor grade, and then do prognosis, how the tumor will behave, and the prediction which tumor, which patient will be response to what therapy. So this DP and AI will give us opportunity to improve the quality and the efficiency of pathology to optimize patient care. So I like to introduce the concept of augmented intelligence. So the image on the left is not a very good illustration of AIS. Basically it's implying machine replace human, still read through the microscope. So that's a bad image. What we want is the image on the right. Human is still in the center. Now with this assistant, assistive role, we can integrate radiomics, genomics, and everything else together to provide better patient care. So in order for us to go to the next level, we have to make sure the data, the data sets, and the partnership with pathology need to be quality one. So today we only have time just to give you a sneak preview of the slides and images quality, how important that is. So this one is showing how the thickness of this cutting slides is important in quality. For example, the left HE slice, if it's a three micron, it's not ideal. If it's too thick, it's 10 micron, it's not ideal. Because you either lose nuclei, truncate them, or the nuclei is too thick you can't see them. So ideally, four to five micron is good. And also the thickness affects immunostain. So as a pathologist, we control those qualities. We use three big tools. One is the proficiency testing, one is rechecking, one is on-site evaluation. So this is an example of proficiency testing provided the College of American Pathologists. The laboratory will send the requests. And for example, this test is to ask you to submit four slides with a specific request. The slides will send to the college, and the panel of judge, like me, practicing pathologists, will review the glass slides and the whole slide imaging to evaluate how the slides are fixed and how the slides are scanned. And here's some examples looking for problem. For example, due to histology, the slides are not in focus because of the folds and the floaters. And this out of focus can also be due to scanning. And this is incomplete scanning. For example, there's a piece of tissue on the thumb drive. You can see it's missing on the scanned slides. And this tissue is truncated on the edge, so it's not completely scanning. And the digitalization can create those stitching artifact, and you have a piece of lint, dirt on the slides. It's also not good. So this quality control manually done by us now, but we're envisioning the next step is to use image analysis solution or use those open source AI tools to evaluate those slides. So this is another program. It's every two years, the laboratory will participate in this on-site evaluation by a group of peer pathologists. In order to help pathologists to practice, the college provide guidelines. And I'm only highlighting two guidelines here. One is the whole slide imaging guideline update. The other one is the CAP HER2 quantitative image analysis guidelines. So these are the tools for us to improve our quality. So here, I'm going to highlight one example how to integrate pathomic and radiomic to provide a better patient care possibility. So in this case, it's a preclinical model of sarcoma. We all know that hypoxic reason within the tumor are heterogeneous. And they can be variant and then response to the therapy differently. So in this study, and we're trying to co-register the radiomic and the pathology data. And at the end, we're able to predict how the tumor will response to therapy. So this is the future. And it's really exciting here. So in this short conversation, we talk about pathology imaging data are heterogeneous and prone to numerous pre-analytical and analytical variations. And we talk about the advantages of having whole slide images and what's the foundation for digital pathology and AI and the precision medicine. We also talk about the importance of the quality images that as the first step for quality AI. And we also give an example talking about the integrated radiomic and pathomic can provide optimized patient care. So this is the last slide to acknowledge my community, all the associations, the environment I'm involved. And I also like to thank some of the industries, which gave me the opportunity to be scientific consultant, which I have no conflict of interest for this talk. Thank you. Okay, so I want to talk to you about extending what Marilyn was talking about. But I also want to, one of the biggest difference between pathology imaging and radiology. When radiology went digital, the film went away. But with pathology, as you can see, the glass slides are still here, right? So we are moving to technologies which will allow us to go directly from the tissue to an image. But today, we still need the glass slides. So many of the tests that we do in pathology require glass slides. And that's why that's one of the biggest differences between radiology and pathology. We still have to store the glass slides. We are required to store in some states glass slides for 10 years. So as we think about diagnostic medicine, and I had the opportunity to visit Dell School of Medicine and the Department of Diagnostic Medicine there. As we think about convergence, this difference will make that convergence somewhat of a barrier until we come to a point where we don't need glass slides, right? So we're moving in that direction. So I'd like to talk to you today about what we have been doing at The Ohio State University. So I'm a pathologist. I am actually on service tomorrow. So I have the opportunity to start reviewing my slides for tomorrow already using digital slides. I will talk to you about the role of lab information system, and radiology is very similar in dealing with information as it relates to images. And what are the barriers? What are the challenges as it takes to integrate? What are the informatics challenges? And in the next session, we'll actually highlight some of those challenges in more detail. But pathology workflow, as you could see from the slides that were presented by Dr. Bui, is highly variable. The thickness of a slide can dictate the quality of the image, right? So we have some types of biopsies like muscle biopsies or brain biopsies or prostate biopsies. They have to be prepared in a special way so they can be scanned and produce a diagnostic quality image. And some of these tasks are automated. So if you go to the clinical pathology labs where they do microbiology and clinical chemistry, hematology, many of those are automated. But anatomical pathology is still very manual. Someone has to physically cut the tissue, make sure it's the right thickness, and produce those images. Some of those steps can be algorithm driven, and there are new technologies which are making this process more and more automated. But it requires a lot of skilled technicians, and pathology in general is facing a shortage of these skilled technicians, which is more the reason for us to move into these technologies. And in addition to that, as we think about enterprise imaging, I was walking around downstairs and looking at the exhibits. Many of those solutions that radiology needs, pathology needs the same solutions. In fact, I looked at many of those vendor areas and none of them mentioned pathology, pathology data or pathology images. So we have scanners which make a glass slide digital, and these can be small scanners which are controlled directly by a pathologist, like a robotic microscope, or we have high throughput scanning systems in which you can load up to 1,000 slides and walk away and they become digital images. The other big difference is a glass slide, when you digitize this, it can be up to 1.5 gigabytes, even after compression. And if you scan it in multiple planes, which is called Z-stacking, they can go up to 20 gigabytes each. So in my lab, we scan 3,000 slides a day. So we are generating a lot of data every day, and we still have to store this data for 10 years. We cannot discard it, and we also have to store the glass slides. So the goal is to, as we look at how do we deploy these, the biggest informatics challenge is how do we create an open platform for these tools? How do we make all these tools, which are AI enablers, make it happen in the same time? So I'm going to show you some examples of what we have done at The Ohio State University. And so our initial challenge was to take this glass slide data, millions of glass slides, digitize them, create barcodes, link them to the lab information system, and then link them to the electronic medical record. So all these glass slides were stored in a warehouse somewhere. They had to be brought in, cleaned, barcoded, and linked to the system. So here you can see this is being done manually. All of them now have 2D barcodes. Moving forward, as we are doing it today, we are doing it in real time. So the cases that I'm responsible for today and tomorrow, they've already been digitized, and they are in my queue, just like a radiology work list. So at the end of the day, we have the slides on a monitor. So this is our new histology lab, where the scanners that you can see are being loaded with these glass slides. So in the center of all this is the lab information system, which links interface engines, links outreach modules, barcoding and tracking, and all these images coming out of this image management system, linked to the different scanners, is being rendered and presented to the pathologist in their work list. So this is where we need integration, and this is where all the informatics challenges really come into play. So this is our current workflow, where slides are prepared and stained. They have to be standardized, they have to be of similar thickness. They are now digitized, and cases are assembled as barcodes are added to them. And QA is done, all the QA that Dr. Bui mentioned, and then the pathologist can review them remotely anywhere. They can review them in their office, or on a monitor which is digital pathology certified, and render a diagnosis. So this is how it becomes similar to radiology. So here I'm showing you integration of the lab information system, the whole slide imaging system, and all the apps that you might be using on these systems. So this is an example of my work list, where I'm in the EMR now, and I'm in the module for lab information system, and I can see all the cases where images are available for me. And as I go and click on any of these cases, I can see all the information which is needed for me to review the case and release my report. And I can click on a button and I can see the image. And once I'm in the image, I can use all the tools that Dr. Bui was showing you. So this is a pathologist who is reviewing images on a workstation. So we have, with this workstation, we have a viewer software which enables us to do collaboration in real time. We can do tumor boards, we can do research, we can do teaching and education. And concurrently, several people can be on the same image as we navigate these images. So this is a typical whole slide imaging viewer, which makes navigation really easy. You can measure things easily. You can identify and annotate features. You can lock the different images. As I'm looking at this patient's liver biopsy with metastatic cancer, I can look at all the immunostains, all the biomarkers that are needed, and co-register these images. So think about all the informatics challenges that are needed to be overcome for this to happen. So the scan slides are instantly available in the image management system. I can start looking at these images as soon as they're scanned, because all this workflow is barcode driven. I can consult colleagues. I can send a link to a colleague anywhere remotely, as long as they have privileges, as long as they have access to this for patient. We can sign out with residents. This was really important during COVID, where we were practicing social distancing. And so as we have rolled out digital pathology at my institute, we've already seen improvement in our workflows by direct interface to the LIS, with direct interface to the EMR, doing all the conferences, flagging cases, control slides instantaneously available. So we don't have to distribute glass slides anymore. We don't have to wait for the slides to be sent to a warehouse or to a consultant. And we can also do priors easily now. We don't have to wait for glass slides from the warehouse to come back. And now once you have achieved that, you can start to roll out AI algorithms. So this is an example of an AI algorithm, which screens the prostate biopsies and finds the one which have cancer, and even grades it for pathologists. so the pathologist is controlling the system and the AI is just a co-pilot. Pathologist is driving it, so this is what Dr. Bui was talking about, augmenting the intelligence of the pathologist, using this as a decision support tool. So this is a heat map which shows where the cancer is. And finally, this is really a reality now where once you have overcome the challenges which radiology faced many years ago, in pathology, we now have the ability to review patient images side by side with patient data in the electronic medical record, have access to these cases instantaneously. So for this, the integration is key. So this is our road map currently where we started with static images and now we have whole slide images for education, for quality assurance, and for primary diagnosis. And we are now moving towards a pathology-packed system and even looking at vendor-neutral archives where radiology data, pathology data, cardiology data will all be accessible in the patient context. And finally, we can start using AI tools on it. So this is not something that's going to happen 10 years from now, this is already happening in pathology, although adoption has not taken off as rapidly as it did in radiology several years ago. So with this, I'd like to conclude that this session is really, really important because it really highlights some of the key differences between pathology and radiology, but it also enables you to see what are the challenges as we think about diagnostic medicine. How do we bring all these two fields closer together? So in our institute, the pandemic was actually a catalyst because many pathologists wanted to use digital slides when they really needed to. There were pathologists who were afraid to touch glass slides. They wanted to wear gloves and sanitize all the slides and we went through that phase in April of 2020. So it was crazy, but we've overcome those. And so the emerging informatics clinical interspace is a very, very important area to continue to work and it's going to evolve. So with this, I'm gonna stop and see if you have any questions for me. Thank you. Thank you for the opportunity to share with you some of my thoughts on integrated diagnostics and more specifically on the informatics part of it. So when talking about integrated diagnostics, it's important to realize that it's not only bringing together the diagnostic pieces of the puzzle, but we do have to deal with other relevant items as well. So for example, informatics is a very important component. Workflow is also very important. So I would like you to realize that integrated diagnostics, it's complicated and several items need to be addressed and I will take you to the relevant informatics items. So talking about informatics and integrated diagnostics, we firstly have to think about the IT architecture. Secondly, we also need to think about the levels of integration that we need or might need for a specific case. And last but not least, we are only able to work on integrated diagnostics if we are able to address the right use cases in order to get to clinical implementation of integrated diagnostics. So let's have a look at the IT architecture. And also here, I would like to distinguish between two relevant topics. So first the concept, in the hospital, a lot of data is generated every day. But the thing is that a lot of this data is siloed. It's in different systems and it's simply not possible to bring that data together in order to make, for example, relevant calculations for an individual patient. So that's why we at Rasmus MC are moving towards a health data platform that has one unique underlying data model that allows you for combining the data from the different systems. And of course, it's also relevant to think about research and to think how you connect your data to other hospitals so that all relevant data sources can be combined. The second topic related to the IT architecture is the technical part. Of course, we have to make sure that it's technically possible to store the data into one health data platform, meeting one unique data model, but also meeting open standards. So there is indeed the possibility to realize integrated diagnostics into clinical practice. The second topic is the level of integration. Of course, integrated diagnostic means that you bring together data from imaging, data from other omics, you combine them, and also you would like to combine it with clinical data in order to come up with a very relevant and state-of-the-art differential diagnosis for an individual patient. But the level of integration is important and is related to specific use cases. So first have a look, let's have a look at the levels of integration you would like to distinguish. And that's firstly the viewer level. Secondly, we can think about integration on the data level. And thirdly, we can think about integration on the workflow level. So what does level of integration on the viewer level? That basically means that you're able to see all relevant information in one or maybe two screens. So we have information about a patient, the clinical history, but also about the imaging, about omics data, visually able the doctor to see that in one or maybe two screens. So that's the viewer level. If we go one level, then we have the data level. And as I already mentioned, the data level allows you to combine different or data from different sources so that you're able to combine them and to make relevant calculations to get to precision medicine. So that's the integration on data level. And I'll come back to that in a use case later in this presentation. And then last but not least, of course, is the workflow level. We have the diagnostic workflow, starting with a diagnostic question. And then we have the management and the order process in which also relevant tasks are being undertaken as well. And of course, in the ideal situation, we would like to get to an integrated report that basically covers all the relevant information for the patient. But that's on the workflow level, the diagnostic workflow level in total. And it might be complicated to start with. So starting with a less complicated workflow so that you, for example, as radiologists are able to see the pathology reports and that the pathologist is able to see the radiologist report in the workflow system. And maybe also looking at discordance. Also, we'll come back to that later. But that's the workflow integration. And of course, we do need relevant use cases in order to be able to implement integrated diagnostics into the clinic. So for the viewer level, I will take you to a case on liver FNH. For the data level, I'll take you to a case on glioblastoma. For the workflow level, I'll take you to a lung cancer case. And it is important to use the agile methodology. You know where to go, but you do not exactly know how to get to your final destination. And you should work with very dedicated people that are flexible and are aiming for the right solutions. So let's have a look at the viewer level case, the liver FNH. This case, of course, is usually available in the MDT, where, of course, the MDTs are available, but also relevant clinic information. So for example, in this patient, she had obesity, she had oral anticonception use, and she underwent MR, and you see a liver lesion here, multiple liver lesions. Basically, a biopsy was taken. And here are an example from pathology images. That if you are able to combine this information into one specific screen, that, of course, allows you to do, have efficient discussions on patient level. And here you see more schematically what's possible that you are able to give patient, to give images from different systems. So from the pathology system, from the radiology system, to a more universal environment, the EMR usually, so that MDTs can be facilitated as good as possible. So let's have a look at the use case for data, glioblastoma. As you probably may know, glioblastoma is the most common and aggressive primary brain neoplasm. There is a limited median overall survival, and this paper basically aims for the integration that can capture multifaceted tumor characteristics, facilitating personalized treatment planning. And this paper has a look whether it's worth to combine clinical measures, radiomics, MGMT modulation, and genomics. And the results are, and here you see a more conceptual representation of what I just mentioned. So you have radiomics data, you have clinical data, molecular data, and you use all that data for risk stratification. And here you see the results. For example, have a look at the model one, and compare it to model six. And what you see, you can conclude here, is the more diagnostic information you add, the better the survivor probability is. So here it turns out, if you're able to combine on the data level the integrated diagnostics concept, then you are able to give better predictions and do a better job in individual patients. And then thirdly, the workflow use case. That was a lung cancer, and when you have a look here, you see the cancer that has been imaged at a CT scan, a biopsy was taken, and it was concluded that indeed this was a lung cancer. But, you know, you can think about how would we be able to make a combined workflow that's relevant for this particular case. So we thought, you know, it might be relevant to allow the pathologist to give feedback to the intervention radiologist, whether, you know, the biopsy was undertaken well, whether there is enough material, whether the material was appropriate. And if that's not the case, a feedback loop would be available in order to give the intervention radiologist feedback on the job on the biopsy. Furthermore, you also could imagine that you could have a different setup for diagnostics in lung cancer. So what we did, we added, we set up a radiologist correlation level scoring. So what happened was that an individual tumor was scored on a five point scale, ranging from benign to uncertain benign, to uncertain malignant to malignant. The pathologist did the same, and he used basically a seven point scale using this same system as the radiologist does, but we added non-representative biopsy and insufficient material to the pathology score. And then the system compared the scores of the radiologist with the pathology score. And if there was a discordance, a separate work list was generated so that the radiologist and pathologist could take it from there and have discussions on which conclusion was the right one. So to conclude, informatics plays an important role in implementation of integrated diagnostics. And we have to think about the IT landscape. We have to think about the levels of integration. And in order to come to integrated diagnostics into daily clinical practice, we need relevant use cases that need to be addressed using the agile methodology. So with the product owner, it's crown master, and also somebody, the product manager, who has the strategic overview to make sure that we are getting to the right points for integrated diagnostics. I would like to thank you for your attention. So I'm gonna do the last session and kind of pull things together, maybe not so much what you should do, but show one example of what you could do if working with radiology and digital pathology PACS. So I'm gonna present some of the experience at Hospital for Special Surgery. So we're an orthopedic hospital based in New York City. We do a lot of joint replacements and other musculoskeletal health care revolving around rheumatology and musculoskeletal medicine. Tom Bowers, our chief of pathologists, came from Cleveland and kind of spearheaded digital pathology there and also at HSS. So I wanna thank him for this material. And Inder Kohli is our IT person who worked with this and then myself on the radiology side. So HSS started a digital pathology journal journey a couple of years ago, but in February of 2021 is when we actually deployed it. And from my understanding, we're the first deployment of this vendor in the United States. They have many other deployments in Scandinavia and Europe, but the primary benefits they were looking for was to enable the pathologist to digitally sign out their surgical pathology cases, implement this integrated diagnostics to allow other providers, particularly surgeons, radiologists, pathologists, and other physicians to correlate the imaging within the same system, enable and facilitate the consultations and interpretations of digital images from other institutions. So doing some referral-based reading or telepathology, and then to promote research and collaboration across our institution, which is primarily based in New York, but we do have some other regional sites in the tri-state area and in Florida. So you've heard about the pathologist workflow and about some of the IT infrastructure. So again, from an integration standpoint, we have this physical media of the glass slide that has to get scanned. The scanners, as we heard in large part, do not support the DICOM file format. So I would encourage us and everybody as users to push the vendors that DICOM is a medical standard. There is one that's available for pathology. It makes managing the data better in terms of looking at the metadata, some of the transfer syntax, other multi-frame aspects of that object. But currently they're imported in non-DICOM format. And then we're integrated with other information systems, the EMR, using things like the HL7 messaging to update the status and also provide some of the patient demographics. So this is the way our PACS architecture is at HSS, very radiology focused. You know, on the orthopedic side, we do maybe about half a million radiology exams a year, which is about a moderate sized facility. And we have radiology PACS that's dedicated to that. But now we've integrated the pathology PACS. So with regards to now putting pathology imaging into this infrastructure, we then have our scanner. So just like a CT and MRI scanner, we now have the slide scanners that are pushing the images to the storage. But because of the storage requirements and also the file format, we have now a separate server for that storage for the pathology images. And then we also have a database to help manage that, which provides the pointers to those images. So then from a pathologist perspective, the sign out workflow, according to Dr. Bauer, is the cases are assigned to the pathologist. When they're accessioned, they get ready the microscope slides, stain, cover slip, all of the things you heard from the previous speakers. They're delivered to the scanning room, loaded in the scanner, sometimes could be done overnight so that they're available the next day. Then the default here is to scan all the slides subject to a pathologist preference, what the histology workload is, scanning workload. And then the potential urgency of the diagnosis. But after scanning all the whole slide, images are briefly reviewed for quality control and available for sign out. If the scanning is delayed, then they may choose to distribute the glass slides without being scanned. And so example here, one of the points Tom brings out is that careful orientation of tissue on the glass slide may help optimize the file size. And maybe we'll ask some of the pathologist speakers to talk to that issue about the kind of setup up front, how important that is. So then the pathologist goes to the EMR. This is the vendor that we use in our case here. And then on the work list in the pathology module, there's an icon that denotes which of these have digital slides available. So they know they're gonna be visualizing them within the digital system. So the EMR is considered the one source of truth. So now all of that information is then stored and selected within the clinical information system. So again, ready access to the patient information. And then there's a link to open the whole slide images within the PACS. And then the whole slide images are opened up on a separate viewer. Again, typically a multi-monitor environment as was shown just to kind of facilitate having your patient information on one regular screen, but then having your visualization on a dedicated diagnostic screen. And so this whole slide image is viewed on the PACS monitor. While they're reviewing the slide in PACS, they then dictate it. So for us, we have an enterprise level speech recognition engine that all of our clinicians use when they're dictating their clinical notes so that that enterprise speech recognition is also used by the pathologist and they can dictate into the information system. And then they can also code the case. So one thing I think that is really nice about having the radiology and pathology systems together, particularly for something like musculoskeletal is now the pathologist can readily access the radiology images. It's right currently now, it's a different viewer because there's some certain constraints or considerations you need for digital pathology, but our PACS viewer, so the images are in the same folder here. You can see pathology images are listed right next to the radiography images so they can pull them up on another monitor to look at them and correlate, which for musculoskeletal is often a very important aspect. So the whole slide image and radiology studies can be viewed together or they can go back into the PACS if they want to use a PACS workflow rather than a EMR driven workflow and kind of go back into the PACS and find those cases. The image here is more to highlight that, so you have your pathology image. If you, many systems you can right click and have some tools. So there are, for radiologists, we're familiar with using, invoking these tools here, but the vendor has now included a couple of pathology specific tools. So being able to add the, looking at the mitosis or looking at the cell counting here. And so this is something that's within that user interface. So again, adapting the user interface for the pathologist for their workflow. And then being able to annotate this image, and Tom has identified this as a synvisc granuloma, a little granuloma you can get from a visco supplement that you get injected into an arthritic joint before you get a joint replacement. This was a person with a, went for a joint replacement. So with regards to the monitor, the resolution and quality control, so our vendor currently recommends that we use a test pattern. So this test pattern is known from Task Group 18. It's from the American Association of Physicists in Medicine. It's one that's developed for radiology workstations. So this provides a quality check. There are a number of boxes there that we check as, the way our system works is every week this comes up for the observer to go in and validate their monitor. So I do that every Monday when I go to my workstation in radiology. The same is advocated for pathology. However, it may be that pathology might need something a little different. We can look at other domains such as mammography, where it has its own set of requirements for quality control, which may need to be more rigorous than, say, a general radiology module. And then just to wrap up here, these are just some of the anecdotes which is some information, but this is found to be extremely useful in our environment. Not only do we have radiology and pathology images, we also save arthroscopy images. So on the research side, there's a capability to correlate the radiology imaging, arthroscopy imaging, pathology imaging. It's been great for teaching. And again, for some collaboration across the enterprise, which can be done telephonically and not have to actually be in front of the microscope together. So with that, I'm gonna wrap up the lecture part of the session.

integrated diagnostics

radiology

pathology

imaging informatics

digital pathology

precision medicine

workflow integration

AI automation

patient outcomes