Pathology’s Digital Revolution Is Changing Diagnosis

This transcript has been edited for clarity.Nelly Abulata, MD, PhD, MBA: Welcome to Digital Health Pulse, where medicine meets technology and patients stay at the center. Hello, everyone. I'm Dr Nelly Abulata, a professor and consultant of diagnostic hematology at Cairo University's Medical School in Egypt, a digital health strategist working across global healthcare systems, and your host.Today we're exploring how one of medicine's most foundational specialties is undergoing a profound transformation. For more than a century, surgical pathology has relied on the microscope, glass slides, and expert visual interpretation to diagnose disease. But today that model is evolving, becoming increasingly digital, connected, and data driven.To explore this transformation, I'm delighted to be joined by Dr Saeed Asiry, an attending pathologist at Lenox Hill Hospital in New York in the United States, and an assistant professor of pathology at the Donald and Barbara Zucker School of Medicine at Northwell Health. He also serves as chief of neuropathology and director of cytopathology at Lenox Hill Hospital. He earned his medical degree from Umm Al-Qura University in Saudi Arabia before completing specialist training in Canada and the United States, including fellowships in cytopathology and molecular genetic pathology. His areas of expertise include neuropathology, cytopathology, and molecular genetic pathology.Saeed, welcome to Digital Health Pulse. It's a pleasure to have you with us. Saeed Asiry, MD: Thank you, Nelly, for the nice intro. I’m very happy to be here with you today to discuss this important topic. Abulata: I want to start with your journey. You trained in Saudi Arabia and then continued across Canada and the United States. Can you share a little bit about that, how it shaped the way you think about pathology and how it's evolving today? Asiry: Absolutely. You already touched on this in the intro, but my training, especially at the beginning in Canada, was specifically focused on neuropathology.At that time, I realized that neuropathology involves more than diagnosing tissue; it also involves molecular genetics. That led me to explore molecular diagnostics and cytopathology. In neuropathology, we often use cytology preparations, such as smears and wash preparations, to help make a preliminary diagnosis during an intraoperative consultation, alongside frozen sections. So I became really interested in neuropathology, cytopathology, and molecular genetic pathology. Then, as time passed and as we've seen in the past few years, molecular genetic pathology evolved enormously. And then there is the evolution of AI models, especially in the past few years. So today, to succeed in pathology, you must be aware of the areas that are evolving quickly. It’s not just H&E [hematoxylin and eosin] diagnosis anymore.In a bit, I'll show a few examples to demonstrate that to keep up with the specialty today, you have to be up to date with molecular genetic pathology on one end and advances in AI on the other, while utilizing your digital tools to reach the best possible diagnosis.Abulata: That's a really great way to start. When you reflect on your early training, what defined a strong pathologist in the microscope-based era? Asiry: I think it was, and still is, strong pattern-recognition skills, good pictorial memory, consistent clinical correlation, and knowing when and how to use ancillary studies, especially when they are new. Abulata: And how about difficult and borderline cases? How were they handled before the era of digital pathology existed? Asiry: We used to send the physical glass slides, with all the special stains and immunostains performed on the case, to a consulting pathologist. That could be within our institution or our health system, or to an outside institution. It used to take weeks to months to get an answer, especially for challenging cases. That’s understandable, given the process. But it also affects patient care because patients might wait for an answer for weeks to months, especially if it's an aggressive malignant neoplasm and it's a challenging case. So that’s how we used to do things, and the drawback was delayed diagnosis. Abulata: How has your daily workflow changed with the shift from glass slides to whole-slide digital pathology?Asiry: Well, I still love to look at glass slides in everyday practice. But I find digitally scanned slides super helpful in consulting services, for example. If you have a challenging case, instead of sending all the slides and the whole case to someone, you can just call a pathologist in a different institution or in the same institution who specializes in that area.Digitally scanned slides are also helpful with a soft tissue tumor, such as a sarcoma, that I might not have encountered before. I can call a soft tissue pathologist and it's just faster. They can guide you about what to do, what stains to order, etc. You can have the slides scanned and just give them the case number, and they can look at it in their own time. It's faster. You eliminate the step of sending the slides. Every day, we have multidisciplinary tumor boards. And in those tumor boards, we have clinicians, oncologists, radiologists, and, of course, pathologists. And when you have your digital slides or scanned slides ready, it's much easier. Rather than going and looking for the glass slides, everything is stored with just a few clicks. You can show the scanned slide live or you can take pictures. Another example is if you have a tumor resection and a prior biopsy that you want to compare to see if it is the same tumor. Is it progressing? Is it changing? Previously, we would request the old biopsy slides, review them, and compare them with what we have. Now, if the slides are scanned, I can just go in, check the case number, and review them. Also, if I ask for additional testing, like immunostains, from an outside lab. Now, most labs here, even those in different states, scan the slides and I can see the scanned slides right away and make a diagnosis. I don't need to wait for the glass slides. Those are just a few examples of how digital pathology affects our daily practice.Abulata: And that's just the start. How about when AI enters the conversation? From your perspective, what is AI actually doing in slide interpretation? Asiry: For now — and I stress for now — in my experience at my hospital and in our system, AI is mainly used to improve consistency, mainly for quantification. For example, PD-L1 and Ki-67 are immunostains that require quantification. We use AI to help with that. We’re currently using it for PD-L1 and validating it for Ki-67. It’s super helpful.I can show you two examples.This first slide shows how we’re using AI for PD-L1 quantification. The image on top is a medium-power view of a lung cancer. Almost everything here is tumor, but with different morphologic features. Our oncologist wants to know the PD-L1 status to justify immunotherapy. PD-L1 is a protein expressed on the surface of cancer cells. If it is expressed at sufficiently high levels, immunotherapy treatment is justified.The first picture shows an H&E stain. For non-pathologists, hematoxylin stains cell nuclei, which is why you see those purplish nuclei. It’s a low- to medium-power view. And then the pink stain is the cytoplasm, which is eosin. In the second picture, labeled PD-L1, you’re seeing a brown immunohistochemistry stain. This is the membranous staining of the tumor cells. In order to report on PD-L1, we have what we call the tumor proportion score (TPS) and the combined positive score (CPS). The TPS is the number of tumor cells staining positive for PD-L1, divided by the total number of viable tumor cells.Imagine the challenge if you want to count the number of positive tumor cells in this field and divide that by all the tumor cells. They're in the thousands, and this is just one area of the slide. If you want to do it on the whole slide, it’s very difficult. We used to give a visual estimate. Now, with the help of AI, I chose this area for image analysis, and as you can see, it calculated the TPS. So the PD-L1-positive tumor cells are the ones in red and the negative ones are yellow. We have the counts of positive and negative cells, recorded here, and the number of cells are in the thousands.I could repeat that on the whole slide, on the whole tumor, and get a more accurate estimate. The other scoring system is what we call the CPS score, in which you also include inflammatory cells, marked here in purple. Again, imagine if you're going to take the inflammatory cells into account. It's really difficult. But the image analysis software is doing that for me, and it gives a more accurate estimate of the CPS score. It's a really helpful feature. Having said that, if we look at the second slide here, the green arrow is pointing to the lung cancer, which is an adenocarcinoma. But the yellow arrow is pointing to cells in the spaces between the cancer cells. These are macrophages. So when we stain for PD-L1, inflammatory cells — including macrophages — can also get stained. So when I run the image analysis on this sample, it counts those cells as positive tumor cells. Again, you have to teach your model and you have to be careful. You can use the technology, but you have to know when the technology might go wrong and give you an overestimate.Another example where quantification is particularly important is in tumors like meningioma, a tumor of the tissue surrounding the brain. Ki-67 isn’t part of the grading, but it is a prognostic marker. Some studies suggest that a Ki-67 index above 4% may indicate a higher risk for tumor recurrence. In general practice, we’ll usually make a visual estimate. We’ll look at it quickly and say it's roughly 2%, 3%, 4%, or 5%. But with AI tools, like what you can see here, all the brown dots in the second picture are positive for proliferative cells. Then you see that highlighted in the image analysis. The red dots are positive and the yellow dots are negative for tumor cells.So it will give you a more accurate percentage of Ki-67, and you could do that on the whole slide. It will give you a better assessment of the proliferative activity. These are a few examples I wanted to share with you, Nelly, just to show you how it can be really helpful in terms of quantification.Abulata: So AI can be used as a primary quantifier, a second reader; but, of course, it can never replace a pathologist's judgment. One of the biggest promises of digital pathology is expanding access to expertise. How has telepathology changed the way second opinions are now obtained in your daily practice?Asiry: Oh my gosh, big time. Just a few examples:If you have a difficult case you want to share with an expert pathologist, you can share the case number with other hospitals with expertise in the area. In telepathology, we have consensus conferences for each body system — for example, a thoracic pathology consensus conference, a neuropathology consensus conference. Some of them are daily, some weekly. Again, you can show a challenging case live and get feedback immediately. Instead of sending the case, instead of calling or sending it to an outside institution, you could just show the case and get live feedback. It really improved the practice big time. Abulata: So this definitely improved the turnaround time and the time it takes for the patient to receive a diagnosis as well.Asiry: Absolutely. As I said, it used to take weeks, even up to a month, to get a final diagnosis — and that was just a few years ago. That was because you had to send the slides, they had to be accessioned, and at the other hospital they had to go through the process and then the report had to be faxed.We eliminated all of this with telepathology. It has really improved turnaround time and patient care, so absolutely a plus. Abulata: Some pathology decisions happen under significant time pressure, especially those done during surgery. Where do you see digital pathology and AI adding the most value in frozen section diagnoses and intraoperative decision-making?Asiry: Let me explain quickly to the non-pathologists in the audience what a frozen section is. Let's say a patient is going into the OR for resection of a tumor and they need to determine whether it’s neoplastic or non-neoplastic. They take a piece and send it to us in pathology. We freeze the tissue quickly so it’s rigid enough to cut. We stain it quickly with the basic H&E and, ideally, give a diagnosis within 20 minutes while the patient is under general anesthesia. There are many cases where you encounter an unusual tumor. Or you're in a small community hospital where you don't have the experience of dealing with complicated neoplastic cases. I see this often in our practice: You get a case and it's an area that you're not experienced with, so you have two options. You can scan the slides and call the pathologist you want to consult, and then the pathologist will access the digital slides and give you a preliminary diagnosis.Or you could make a live virtual call while you’re doing the frozen section, show it to an expert in that area, and then they will guide you. I’m specialized in neuropathology, for example. I get calls from pathologists from other hospitals if there is a complicated neuro case and they need expertise or help.It's super helpful during frozen section because it's a big deal, and as I said, they need to make a decision based on that whether to resect or stop. So it really improved the patient care we provide as pathologists. Abulata: Pathology has really evolved a lot. Let's talk more about that evolution. Where do you think pathology has evolved the most? Asiry: Two things have advanced tremendously over the past decade. One is molecular genetic pathology. As we discussed, we’ve seen rapid progression and improvement of AI models in the past few years.We're currently using it for quantification, but I'm sure there will be a time when AI models will be assisting in diagnosis. But going back to molecular genetic pathology, it has evolved so much that tumor classification now relies on molecular and cytogenetic features more than just H&E, more than histology and morphology or even immunohistochemistry.For example, in my practice of neuropathology, the current WHO classification now is literally based on the molecular and genetic features of each tumor. And it’s going to get more advanced with time, and will be more and more molecular and cytogenetic-driven classification.Also, now we have techniques that we didn't have a few years ago, like cell-free DNA. These are fragments of DNA that can be detected in plasma and in cerebrospinal fluid. If you see driver mutations in those fragments of DNA, given that there is a tumor somewhere, they can help pinpoint where the tumor is coming from.So I think that, in the future, molecular genetic pathology with less invasive techniques will play a bigger role in pathology practice. Abulata: And that's only the beginning. Saeed, for pathologists in training today, what skills and mindset will be essential to remain central to diagnosis in this very rapidly changing environment?Asiry: The core is still the same. You have to have the knowledge. Pathology is really expanding. You have to have the diagnostic skills needed for pathology, pattern recognition, and pictorial memory. These remain essential skills. But also, you have to adapt to evolving technologies and knowledge. Molecular genetic pathology is a big part of that. Now, with digital pathology, you have to get used to looking at digital slides more than just the regular glass slides and feel comfortable with reporting cases based on digital slides. I’ve seen pathologists struggle with this.So it's not just the H&E good old days. You have to broaden your perspective and think more about the other technologies and not get left behind by failing to keep up with what's happening in the world of pathology. Abulata: Let me close with one final question: What do you think are the three most important things clinicians should understand about digital pathology today? Asiry: It’s really transforming our practice in a good way. Clinicians are already feeling the benefits. It provides faster, more reliable pathology results and easier access to experts and second opinions, easy access to additional studies.As I said, if you send an immunostain, most labs can now scan the slides before they send you the glass slides. So in short, I think it’s transforming the practice in a good way, providing faster and more reliable pathology results.Abulata: Saeed, thank you for helping us trace the journey from slide to signal, from glass-based microscopy to digital and AI-assisted diagnostics, and from isolated expertise to connected data-driven pathology networks. As we've heard, the future of pathology is not about replacing the pathologist; it's about expanding what pathology can achieve: faster diagnosis, broader access to expertise, and deeper integration with data and technology.My sincere thanks to Dr Asiry for joining us and sharing his insights. And to our audience, thank you for watching. As always, stay curious, stay clinical, and stay connected. Nelly Abulata, MD, PhD, MBA, is a physician, educator, and global health innovation strategist. She is a professor of hematology in the Department of Clinical and Chemical Pathology, Kasr AlAiny Medical School & University Teaching Hospital, Cairo University, Egypt. Abulata has held leadership roles on numerous boards, advisory committees, and working groups where she has made significant contributions to the advancement of healthcare and higher education. Saeed Asiry, MD, is an attending pathologist at Lenox Hill Hospital in New York and an assistant professor of pathology at the Donald and Barbara Zucker School of Medicine at Hofstra/Northwell. He also serves as chief of neuropathology and director of cytopathology at Lenox Hill Hospital. He earned his medical degree from Umm Al-Qura University in Mecca, Saudi Arabia, before completing specialist training in Canada and the United States, including fellowships in cytopathology and molecular genetic pathology. His areas of expertise include neuropathology, cytopathology, and molecular genetic pathology.