245: Why Going Slow Is Killing Digital Pathology Adoption | Syed T. Hoda, M.D.

Is your digital pathology rollout moving so slowly that it’s creating a fragmented workflow instead of transforming the department?

In this episode of the Digital Pathology Podcast, I speak with Dr. Syed Hoda, Director of Digital Pathology at NYU, about why gradual implementation may no longer be the best approach to digital pathology adoption.

Dr. Hoda explains how NYU used an intensive nine-month planning period to prepare for a department-wide transition. The process involved pathology, IT, project managers, vendors, hospital leadership, and approximately 40–50 people participating in regular planning calls.

This wasn’t simply a scanner installation.

The team mapped workflows, configured Epic Beaker, redesigned laboratory spaces, tested integrations, planned training, and addressed the practical concerns of nearly 100 pathologists.

We also discuss why scanner specifications may matter less than integration, vendor support, training, and system performance. For Dr. Hoda, digital pathology had to work as smoothly as glass microscopy. Speed was non-negotiable.

Change management played an equally important role. Through open discussions, town halls, and the ADKAR framework, the team addressed concerns ranging from ergonomics to the loss of collaborative microscope sessions.

The result? Every pathologist adopted the digital workflow, no one left the department because of the transition, and approximately 60–65 pathologists now work remotely using equipment that matches their office setup.

Finally, we examine the next step: artificial intelligence in pathology. Dr. Hoda explains why NYU focused on building a reliable digital foundation before introducing AI. He also raises important questions about validation, transparency, responsibility, regulatory clearance, and the need for greater pathologist involvement in AI development.

Episode Highlights

• 00:00 — Are we repeating the same mistakes with pathology AI?
  Dr. Hoda compares the current excitement around AI with the early promises made about digital pathology 15 years ago.
• 01:04 — Meet Dr. Syed Hoda
  His clinical pathology background and path to becoming NYU’s Director of Digital Pathology.
• 03:16 — Why going slowly can hold departments back
  How partial adoption creates fragmented workflows, inconsistent training, and prolonged implementation.
• 06:25 — Leadership support for rapid adoption
  Why institutional commitment, resources, and an ambitious timeline made the project possible.
• 10:13 — Nine months of detailed planning
  Workflow mapping, laboratory changes, system configuration, vendor selection, testing, and validation.
• 11:48 — The role of professional project management
  Why pathologists shouldn’t be expected to coordinate every part of a complex digital transformation.
• 14:29 — Why the scanner isn’t the most important decision
  Image quality matters, but integration, service, training, and workflow fit may matter more.
• 17:42 — People matter more than machines
  How vendor relationships and departmental engagement supported adoption.
• 19:19 — Setting clear expectations across the department
  NYU communicated that every pathologist would move to digital sign-out within a defined period.
• 20:49 — Change management is a structured process
  How the ADKAR framework guided communication, education, adoption, and reinforcement.
• 25:07 — Addressing practical and personal concerns
  From mouse ergonomics to preserving collaborative case review between pathologists.
• 27:19 — Why NYU didn’t introduce AI first
  Dr. Hoda explains why pathologists needed to become comfortable with the digital platform before adding new AI tools.
• 29:26 — Digital pathology and remote sign-out
  Approximately 60–65 pathologists now work remotely with equipment matching their office setup.
• 30:28 — Why speed is non-negotiable
  Even a small delay or repeated pixelation can quickly undermine confidence in a digital workflow.
• 33:25 — A cautious approach to pathology AI
  Concerns about premature adoption, self-validation, limited regulatory clearance, and lack of pathologist involvement.
• 37:27 — Scientific validation, transparency, and responsibility
  What happens when the AI result and the pathologist’s interpretation don’t agree?
• 40:41 — Where AI could meaningfully augment pathology
  Quantifying microenvironments, feature combinations, ratios, and findings that are difficult to assess visually.

Resources Mentioned

• ADKAR change management framework
• Digital Pathology Association
• Executive War College
• FDA list of AI-powered medical devices
• A radiology mock-trial paper examining responsibility when clinicians use AI: Examining perceptions of liability about AI in radiology (MedRxiv)
• Why AI cannot do good science without humans (Nature Editorial)
• A previous Digital Pathology Podcast discussion about AI-supported colorectal cancer feature analysis (How to use deep learning image analysis for colon cancer with Rish Pai)

Listen to the full conversation for a practical look at digital pathology planning, change management, remote sign-out, scanner integration, and responsible AI adoption.

Transcript

00:00:00
 I think we're at risk of doing the same problems that we did 15 years ago for digital pathology with AI and the pattern I'm seeing is identical. 15 years ago there were all the digital path vendors at a meeting and they were like we're all ready to go. It was not ready to go and nobody was up for it. In the US radiology has 900 AI tools FDA cleared for diagnosis. Pathology has six. My understanding is that it's being pushed as a self- validated AI solution for most companies, but I think it's not

00:00:32
 organized. And the most important thing that I've noticed is I don't think pathologists developed a lot of the AI tools currently. And I think that this is a problem with pathologists not being engaged in the process of creating things. >> Welcome my digital pathology trailblazers. Today my guest is Dr. Syed Hoda. Welcome Syed how are you today? >> I'm good Alex. Thanks for having me. >> And we're gonna be talking about something I don't know if if it's controversial but I would say it's not

00:01:04
 the common way of doing things in digital pathology. However, it was very successful for you say we're going to be talking about going allin on digital pathology deployment at an institution rather than doing it slowly. So before we dive into this say let's tell the trailblazers about yourself. >> Sure. Uh so I'm a pathologist at NYU in New York City. I'm a clinical professor of pathology at the NYU school of medicine and director of the bone and soft tissue pathology service and saroma

00:01:40
 service here. You know I spend a lot of time teaching etc. And I've done this for 15 years and then few years ago I became the digital pathology director at NYU. It's a topic that I've loved for many years but never had the resources or the go-ahad to to reimagine. So I'd been kind of planning this for many years in my mind but but nothing actively happening until a few years back. So that's that's my sort of background >> and we have something in common. So I'm

00:02:09
 recording this podcast in Poland and say you finished your uh med school in Poland, right? >> Totally right. Yeah. Yeah. I uh I know I was born and raised in the US but I traveled when I was young. So I wanted to get out of the US and see the world. So I ended up at a medical school in Hungary for a while and then and then transferred to Poland and then graduated from medical University of Salacia and Kawita and then did my US rotations in Chicago and other places. So that that was sort of my level. So yeah, Polish,

00:02:39
 but I forgotten it all. Alex, if you want to train me again in Polish, then >> yeah, we can we can set up some Polish classes. But I mean, you're not going to have too much use of it in the US. >> That's the problem. If you don't speak a language, you forget it if you don't speak it daily. Where in Poland are you? >> I'm close to the German border, like 50 kilometers from the German German border near Gou Vapolski. >> Got it. Okay. Okay. Fantastic. >> So when we were talking before this

00:03:16
 podcast episode, you reached out and basically told me going slow is the disease in digital pathology adoption. Going slow is totally not the way to go. Even though everybody is like, "Oh, can we go slow? Can we go bit by bit? Can we like do it piece by piece?" This is totally not what you are preaching. Although this is like the conventional wisdom in digital pathology, let's unpack this. >> Yeah, thank you for bringing this up. So, I think I've thought about this a lot. There are a few reasons for this.

00:03:53
 One is that pathology departments and and institutions and groups are somewhat scared to ask for the resources in in grander scale. So they they settle for small amounts of money that get allocated to them to do digital pathology and they end up doing these little minor digital pathology things like I've spoken to so many institutions now in the last year and a common theme is we have one scanner we use it for tumor boards or for teaching or interesting cases but it's not used in any meaningful workflow right like in

00:04:30
 the department and and and I get it I get it the pathology departments if they don't get the money or the resources they can't invest fully in the digital path. Not this takes some like pathologists have to be more bold and loud and say that we need this uh and we need to transform our workflow completely rather than kind of being like okay we'll just take a little bit of money and do every year we'll ask for a little bit more but what it ends up doing is you end up with like a seven

00:04:57
 eight-year journey for digital pathology where some of the people are using it some of them are not there's no organized workflow there's you don't get all the benefits of digital pathology the benefits are actually transforming the department that you're at at once to create a new workflow to create a new way of working for everybody and if you don't not doing that then you're kind of artificially holding yourself back and I would go as just so as far as I've recommend some institutions to even

00:05:28
 aggregate the money over a few years first and not roll out anything save the money and buy it all at once rather than buying one machine in 2025, one machine in 2026, one in 2027, and then kind of slowly doing it. I would think it's better to do it together because you create uniform workflows, uniform training. You're all together in it. You know, there's a unity aspect of it that's missing in a seven-year phase in approach. So >> that's I mean I can talk about this for

00:05:56
 a long time because I I think the early adopters of digital pathology and and they have my sympathies because there was no you know there was no precedent or no idea what to do what's right what's wrong and they had to kind of do it this way like slowly kind of introduce it into the workflow and and and I think that model was good for them because they were early but I think now we're not early anymore. we're kind of at the right time and I think it's it it needs a mind shift a mindset change.

00:06:25
 >> So, how did you come up with this idea? Like how did you decide that okay I need to have the money or I need to have the plan to do it all at once even though like you said the early adopters the common wisdom is oh go bit by bit. How did you become so courageous? like what inspired you to go all at once and to figure out how to get the money and do it all at once? >> This is another great question. The the the the one thing I will say is that I am at a place that was is prides itself

00:07:03
 on on kind of being innovative in that way like they made medical school free here for all the students. Like there's there's some element of them doing things differently, right, than the usual US institution. So when our dean and CEO found out about this, he was a radiologist, right? So radiology and pathology are always sort of uh you know joined at the hip philosophically in in the way we kind of look at medical world. So he was super enthusiastic about it. he really wanted it to happen.

00:07:32
 uh and he he was uh you know we got resources very very fully like thoroughly which is unique in itself right so there's a bunch of things that unique things that happened the CEO was a dean you know radiologist he was enthusiastic about it my own leadership was also enthusiastic about it um and then I had been already enthusiastic for many years so for me this was already in the planning that I would do it this way to do it really seamlessly and more together and in a cohesive way rather than this approach. So, but even then,

00:08:07
 Alex, like I will admit that my thought about it like I had read the papers about five years, seven years, you know, whatever. And I thought I could do it in probably two, two, I didn't think one or less. Uh I thought two. His idea was nine months. So, the CEO suggested nine months. Uh and when I first heard that, I was kind of taken aback. But at the same time within the course of a day I was ready because you know what is two years and nine months and 12 months like I figured we could make it happen and

00:08:38
 with that much support and and sort of um you know enablement like you can do those things if you have that support right on another side if I didn't have that support and somebody was telling me to do it in nine months I think it would be really difficult to to do it that way um but we I the more I found out about it the more I convinced I became that we could do this for sure like we had IT support we had project management involved. We had access to people and and resources and planning, you know,

00:09:06
 even even things like building a room for the scanners, right? Like they there was a plan for all of that. So, it became really like yes, it took some courage on me to talk to pathologists and and to do those things and kind of spearhead it, but it took courage for the administration to give pathology this kind of resources and say, "Hey, can you do it?" Right? and and we that was my side our side was to make it happen, right? Their side was to make it happen on their end. So that's that's

00:09:34
 kind of how it came along. It's kind of like all these things came together at the same time with the same kind of spirit. So >> this is uh so fortunate that your supervisor, your boss was a radiologist because he already knew intrinsically what's the value of digitization. and it was basically translating it to into another department, another discipline. So, uh it's nice to hear that he motivated you to do it faster and now you're kind of an example for other departments. So, let's talk a little bit

00:10:13
 about the logistics. How did you do it? like okay you secured the money you had the support everybody was on board how do you actually do it in 9 months >> yeah the big the big thing that I am really really emphasizing here and what I'm seeing institutions not do sometimes is to take your time in planning first like before rolling out and I think what people end up doing is that they get their resources and then they're kind of in a hurry to like make it happen because they don't want to lose the

00:10:45
 resources or they don't want to you know like But that's not the part to hurry. I think the part to to to make faster is the actual roll out, the planning of it. Like we found out I remember in like late 2023 is kind of when I got most of the word. And then from 20 late 2023 through uh September of 2024, we were always daily on calls planning everything. Like we we looked through our labs, we looked through the function, the workflow, we hired people, we built a room, we like, you know,

00:11:18
 created all these ideas, built Epic Beaker, like we customized the the lab operating system, we tested all these things like we didn't just like roll things out, we would test them endlessly to make sure that things are working and functioning and be selected vendors for things and you know, so that planning stage I think was very intensive and the the most important thing I'd like to say is that I didn't do that planning. I had a project management team that we were given who organized the meetings, who

00:11:48
 scheduled things, who got vendors, contacts, you know, like I can't do it. Like if you put it on the pathologist, they're not going to have the time. They're going to move really slowly through the planning. So that team was really organized. And then so I could focus on the pathologist aspect while the IT team would talk about you know their integration side and testing that and I would be testing you know for validation and to make sure that we have the right people looking at things. So

00:12:14
 there was a real like large team at some point you know it was 40 50 people who would regularly be on calls at the same time and I can't emphasize enough that nine to nine months of planning is why we could roll out in September and have very few issues because things were tested so much in that nine months. Mhm. The planning is not what consumes the resources, right? So like you say even before securing or I don't know saving h or getting the resources you can have a plan how you would want to do it.

00:12:48
 >> You can the only problem is that I I I do think you're totally right like it makes sense to discuss and have a vision and a and an idea of how to do it. That doesn't take any resources to discuss. What takes resources is that when you're speaking to vendors or negotiating with vendors and building out you know integration across software with vendors that stuff is taking resources right because the IT team needs to be involved they put in effort into it they they have to test things so now you're kind

00:13:17
 of doing that but yeah it is part of their job too but they have to be allocated to the project right like if if a pathologist department is like begging for IT people to help it's going to be a problem like they that our IT team like so one of the cool things is our IT team looked at this as a the way I saw it they looked at it like a transformative change of a specialty and for them it was like a project too like they're redoing an entire specialty the way they work so it was really like a

00:13:46
 cool thing to see them be so invested in ATF >> so I had a chance at some point in my career also working with project managers oh my goodness Like these people have superpowers to pull in the right expertise at the right time and actually like they would make me deliver on time where if I was doing it myself and like trying to plan uh everything and according to my schedule it would be like >> I would say suboptimal timing. Uh but project managers they have some superpowers and like you say it is a

00:14:29
 resource but from what I'm hearing from you right now is basically stop sweating about the scanner. It doesn't really matter. You didn't like mention oh we did so many tests about the scanner. Let's talk about that. What do you think about this question that usually starts the digital pathology discussion? Yeah, that's the first question that everybody asks and I looking back on it, I totally agree, Alex. I don't think that that is really an important question at all. I I

00:15:02
 think that of course fundamentally, yeah, there's a cost associated to it. You have to integrate it with your system. But honestly, I after looking at all these scanners, like most of them, most of them are actually good quality enough that any pathologist could use them and be fine with them. The question becomes, can you integrate it into your hospital or your center or your lab and can you afford you know the cost of it and stuff? Um, and does it you know do you like for me it was really important

00:15:32
 to have service agreements with them to make sure that the scanner company is actually helping us with the because we're doing something unprecedented for a large academic medical center like that company like Phillips needed to know that we would need some things from them. We want rapid service. We want training help from their their staff to train our pathologists and make sure that they're okay. We needed an ongoing thing. So, for example, we kept we negotiated to have them on site for a

00:16:01
 year like training people and that made a huge difference for us. Like we always had somebody around. So, if I was signing out a case and I was having a problem or I was confused, I could call or email them and be like, "Hey, Nicole, I need your help. Like, can you stop by today or tomorrow?" and we can she would come by. So that was like a massive thing to have. Uh I think and and um you know I think that that kind of stuff is way more important uh along with navigating the actual workflow like how

00:16:29
 does a pathologist work redefining how a pathologist works not the scanner type. the scanner type. Whether Alex, you ask me this right now, whether I have Phillips or Hamamatsu or Leica or whatever, I would still be having this conversation with you after going through the same kind of things that we had to navigate. So I I I don't think that the scanner is the the biggest uh contrast. >> And it used to be like a discussion, oh, which one is better? Which one is like diagnostic quality? Now with I don't

00:17:04
 know how many cleared devices we have like 10 eight I don't know every month there is a new scanner being cleared by the FDA and this is basically a testimonial to not the scanner as a device is good enough they have to be good enough to do primary diagnosis on them but to me that amount of clear devices means digital pathology is good enough so pick the tools that uh um you like, but what I'm hearing from you is rather than the tools that are all good enough, pick the team that you want to

00:17:42
 work with or figure out how you want to work with the team that is um helping you implement the scanner. So, it's a lot more about the people than the machines. And there are obviously people and teams helping you from the vendor side, but what about your people in the department? How did they react to the announcement or I don't know how you did that that hey next year we're going to all go digital? Well, I I will say one of the the things that I prototyped also was that by the time we were going

00:18:19
 digital in like 2023 2024, most of the people who were doing digital path at at large institutions, they were more of an informatics background, you know, more technologically aware and people who, you know, pathologists who had had informatics training and those things. I am none of those things. I am like the most ordinary pathologist like I I I am you know a saroma guy. I diagnose bone and soft tissue. I have a service. I diagnose just like every other pathologist that works everywhere in the

00:18:49
 world like daily. So I'm not a tech person. I'm not even a technologic. I don't even like technology. I actually use like old watches like I have listen to vinyl records. I'm not one of those people. Um, but I figured that if I could figure out the most simple way to do these things like in a way that I could understand, then I think other pathologists would understand too. And that was kind of the idea here. So when I announced it with my my department leadership, we actually talked about it

00:19:19
 a lot. Like we would talk about it at every faculty meeting. We made an announcement that this is our plan is we're going digital uh and it's going to be all of us and it's going to be in under a year when we roll out. We started with that idea. We knew we were very transparent in the fact that there was no dillydallying about this might be two years, this might be three years like it was that once we roll out in September of 2024, it's going to be a year or less. Uh it's going to be every

00:19:46
 single pathologist signing out that's going to be moving to digital. And that was really the expectation of the department that we were all expected to do that together. There was no idea that some people would not use it and some people wouldn't you know like go slower. None of that. It was really like a communications thing. And this is part of change management like you know change management had been you know I I we can talk about it more later but this is all part of the change management

00:20:12
 discussion of how to talk about it with people how to let them talk about it too. we would take a lot of opinions like I can talk about it later in the change management part two but like you know we had open discussions very open >> we can talk about it now because yeah it's natural to people for people to react in different ways >> so when it comes to change management I think it's used as a buzzword and many are not aware that it is an official term and an official way of executing change in an

00:20:49
 organization and let's talk about that in your case. >> Yeah, it is a very organized thing and it's you know when I when I first heard about change management it had nothing to do with digital pathology. I was doing a leadership course for with NYU for other things. And in that there was a woman who Kelly who was talking about change management and my ears perked up because everything she was talked about was talking about was about how groups of people in a workplace or otherwise

00:21:20
 get a new idea and how they progress into something new. And for me I made a bookmark in my brain then that when we do digital pathology this is exactly the kind of thing that needs change management. But like you said, Alex, like I've seen it mentioned as a buzzword but with no depth associated to the change management side. So when this happened the first thing I did was I actually called her and I said can you come and meet with me and some of the department leadership to discuss change management about this and she did and it

00:21:48
 was fantastic like she went through some of the principles of it like I'll just give you an example there's something called ADCAR. So ADGAR are the principles of change management. The first step is awareness. The second one is desire. The third one is knowledge and then the the fourth one is adoption and then reinforcement. Right? So you go through these stages and like if I tell Alex if I tell you that tomorrow you're going to move to like a different house in Poland, right? Like that's a big

00:22:17
 change and like you would first have to be made aware of that change, right? And then you would hope that at some point you say yes, I I would like to do that change. So that would be the desire component. So awareness, desire, and then you would ask like where is this house? Like where is the information about it? Like where where how many rooms does it have? So that's the knowledge component. And then hopefully you would want to eventually after hitting those first three things move to the adoption where you actually move to

00:22:44
 the house and then you stay there and that would be the reinforcement. Right? So this is the same kind of principle we use for digital path. Like these guys first of all they're all very educated people like really good pathologists but they some of them had heard different things about digital pathology over many years and like you know we had to first give them some awareness about what it is we're talking about what are we going to do why does it matter how how many of us are doing it right and then we let

00:23:12
 them talk about like amongst themselves we had town hall meetings where we it was free you said hey tell us what you are worried about tell us what you Tell us what you're excited about. Tell us what you hate. And they would go off and tell us all these things. And it was very terrifying. >> What do they hate? >> It felt like [laughter] it was really uncomfortable for me. But at the same time, I I also learned something myself is that like if you're leading a process, it's not about how I

00:23:43
 feel, right? I'm leading it for 95 people. So they their opinions matter and their concerns matter. And the only way I can make this better is to hear what their problems are and try to fix them, right? Like try to address them. So, you know, some people were worried about like, am I is my wrist going to hurt if I'm just using the computer too long? I have a list of these. I save some of them. Like, is my wrist is going to hurt off using the mouse too much and not you know, I'm used to the

00:24:09
 microscope. And so, you know, we made a response to that. You know, we have like wrist rests for everybody. We also have different devices like different types of mouses and controllers and stuff. So we addressed issues like that uh over and over again. So that's kind of the idea and we brought in some outside speakers to talk about it and build awareness and understanding and we naturally let it go like kind of let it flow. We talked about it a lot. Um and prior to what you were saying so change

00:24:37
 management was mentioned in some papers but I don't think there was ever a mechanism employed to have a dedicated person like I had a dedicated person who was trained in change management working with me every single day in that nine months of planning before to to we would look at the impacts of change management on every process change. So if sides are going to move to here to here, what who is that going to impact and what would they care about in that in that mechanism? So that's how detailed we got

00:25:07
 into the change management. >> Oh wow. From the concerns that people shared, what was the one that was like most surprising? Something that you did not hear before? >> The most surprising? >> Mhm. >> There were a lot of really strange small things that people said. Like for example, they'll miss the metallic sound of a slide hitting a microscope stage. Really, really really abstract kind of things like that. And then there were people that were like, you know, I miss

00:25:38
 sitting at a microscope with another pathologist like next to them like looking at a case together. There's something very like real about that. And so like for example like that kind of thing like the intrapersonal like communication between people like in my office right now there's a monitor that's facing out away from me and that's for the other person in the room with me. So that really hasn't changed. Like we made sure to create an office space which would still be collaborative

00:26:06
 that if somebody's sitting in front of you just because we're not using this microscope it doesn't mean that that persontoperson thing is diminished. Like I just had a person come in two hours ago show me a case. They sat down and we looked at it together the same way we would run a microscope. So like I but that's a result of addressing these kinds of comments. >> You know what uh was the thing that I heard once because um with AI everybody is like oh you're not going to have to

00:26:39
 do the boring cases. You're not going to have to search for metastasis in lymph nodes. AI is gonna do this for you. And uh it was an example from a conference. Somebody said, "But this is this feels therapeutic for me. I like it. It makes me calm." And actually, I can relate to that because when I read the studies for preclinical uh evaluation, a lot of this h a lot of the slides are just normal and that's what it is. And I sometimes enjoy it. I sometimes enjoy looking at the normal and just like being in this

00:27:19
 kind of slide viewing trance. Uh so that would be my funky thing about even though I'm well actually it doesn't go away with digital that would go away with AI doing part of the job with you. So um I'm not there yet. >> Yeah. Yeah. Yeah. Yeah. we can you know that's kind of a separate topic but the AI conversation is a complicated one uh and I and I think it's still evolving right and I will tell you on our end we did not lead with any AI and that for me the the situation that I've been

00:27:55
 watching is that pathology has a habit of getting ahead of itself so I went to a meeting 15 years ago where the people were saying digital pathology will be ready in five years and everybody will have it and it was frankly wrong uh it was completely completely offt target and the reason why is that they make these big promises and it creates too much pressure and people don't do it. So AI we did not even involve AI in our process. You got to first get people onto the platform. Like can you imagine

00:28:23
 Alex if everybody got an iPhone back in 2009 and then you were already putting Siri on there and being like now you got to use Siri and use nobody would have like people would have been much more apprehensive about it all and like you know you got them into the platform they started using it they got comfortable and then you added other stuff as you went along over time like there's no rush to we're we are not like fasttracking AI is like the the promise I think we have to first get people

00:28:51
 modernized and then we move to the next step you know >> and you're an institution with a lot of pathologists I also work in an institution that has many pathologists we have 160 uh veterary pathologists and do you guys have like a hundred people as well >> yeah we have like 95 it's [clears throat] going up all the time but I think it's something in the high 90s >> so do you guys um did it enable people to work remotely. What is the situation there with digital? Were you able to

00:29:26
 bring in more people? Did you have anybody quit because they had to do digital? What was the situation there in terms of um you know just being comfortable working um differently? >> Nobody nobody has left because of digital pathology. There's not a single person. I was worried about I was really worried about this when this started that I don't want to lose I don't know how it is in other places right now but in I I'm guessing it's the same but in the United States has a huge shortage of

00:29:55
 pathologists and the job market is >> you know they can go anywhere and get a job so it's not hard to move around and so nobody quit because of digital path they all adopted it every single pathologist adopted it at the end and there to remote like we have 65 people working remotely right now are 60 to 65 out of the 95 plus and they love the remote setup. Like NYU provided the remote setup for them. It's the exact same office computer and diagnostic monitor as they have in the office. So

00:30:28
 they have the identical spec. It's super fast. There's no lag just like the office. Like for me it was mega important that this system has zero lag. Like I've seen some systems that had lag and it was not acceptable to me. So, uh, we made it really, really, really fast. Um, and the home setup is just as fast as as the office setup. So, they love the remote. I mean, the remote thing has taken off for us. >> I think the speed is the one like real thing that cannot be compromised. And

00:31:03
 for a long time, it was because the systems just were not fast enough. And this like a not fast enough digital system is basically taken out of the equation by the sentence I can do the same thing faster on glass then what is the advantage of digital if you're slower so that also like you say you went to this meeting 15 years ago 15 years ago the speed wasn't there so even if you wanted it really badly it wasn't on par with the service that you were providing with Glass and all the other additional

00:31:44
 benefits. You know, 15 years ago, remote work wasn't a thing. COVID was still in the future. Um, and all these things, right? So, the the speed component had to be there. And I'm totally with you, the moment it's slower. Like sometimes things break, but like when I have to submit an IT ticket for something, I'm like trying to figure it out five times myself because it's going to take longer for the IT to come get back to me to fix it. And often I find workarounds. Um but yes, speed is a non-negotiable.

00:32:20
 >> Speed is actually one of those things that people don't consider, but uh it can be very very aggravating to have. I've spoken to some places where they say there's a lag there like a one second lag in bringing it up and then there's that pixelation idea like the focus and those things can break a pathologist spirit when it comes to digital pathology like that kind of lag over you know 50 times a day 100 times a day is too much to deal with uh sometimes like it creates frustration so

00:32:49
 uh we again as an ordinary non-informatics pathologist Alex like for me it had to be just as smooth that's putting it on the the the scope and looking at it. So, >> and what do you think is going to come next? Like what's your ideas for what is AI going to provide in pathology? Are you looking into it? Are you excited about uh something in particular? What are your thoughts on the next steps of your particular institution journey and in general for pathologists? But I love your example like you you even describe

00:33:25
 yourself as a nontech enthusiast normal pathologist who was able to do it. Is there anything that you're excited about beyond just this digital experience? >> You know, I I wish I have excitement about AI, but I have a lot of concerns about AI. And I and I >> the more I'm getting into this conversation and I've met with a lot of companies and looked at a lot of things and talked to other hospitals and [snorts] I think we're at risk of doing the same problems that we did 15 years

00:33:59
 ago for digital pathology with AI and the pattern I'm seeing is identical. 15 years ago there were all the digital path vendors at a meeting and they were like we're all ready to go. we it was not ready to go and and and nobody was up for it. And right now I'm seeing the same problem. Everybody's like, "We got AI, everybody can do it. LDT, you know, lab developed tests. Um we're not ready for it." Radiology has 900 in the US. Radiology has 900 AI tools FDA cleared for diagnosis. Pathology has six. And my

00:34:33
 understanding is that it's being pushed as a self- validated AI solution for most companies. I see this changing. I I don't think that this is going to be the way in the United States. This may work in other places. I think this is a very regulated country when it comes to risk and also vetting out technology. And I think FDA involvement needs to increase. I know that the DPA are working on this in the US uh and I I I would hope to I'll be speaking more with them in the coming months about this. But I think

00:35:04
 it's not organized. And the other thing, Alex, the most important thing that I've noticed is I don't think pathologists developed a lot of the AI tools currently. And I and I think that this is a problem with pathologists not being engaged uh in the process of creating things. The first AI tools that we're going to use at NYU are going to be self-developed. We're already we already created a couple. Uh they're looking incredible and they're very basic and they're very practical. They're like

00:35:30
 practical things that are the most common things that pathologists want. you know developing an algorithm for prostate carcinoma gleon grading I think there's some utility for that but is that the first thing that we would want I'm not sure like I I as a pathologist I'm not sure that that's what I see as the first thing you'd want so we're going to redo it in our own way and then external vendors yeah we still look at them and you know we'll keep them in consideration but I want to see more FDA

00:35:57
 cleared stuff for the US because I I think that I'm worried about about a bunch of things related to AI What's the thing that worries you most? >> I was at a meeting for example uh in New Orleans a couple months ago called Executive War College and there were some lawyers there who were on a panel discussion a few different sessions and they were first of all very very apprehensive about using AI in a clinical uh uh work currently. The reason why is that a lot of the times we don't know how things were trained and

00:36:30
 where where the things came from. the transparency isn't there. You know, we don't get a predictive value given to us like negative or positive really easily. Sometimes that data is not there. The different places it was trained on, the data sets it was trained on, and when you implemented on yours, how are we tracking and and making sure that that's the right answer. Pathologists are going to be apprehensive already about that technology suggesting an answer. What happens if the pathologist and the AI

00:36:57
 don't agree, right? Like where is that where is that heading? Who is right? How do you determine that? I think that currently the way it's going and radiology is an example of this. There was a paper published about them doing mock trials about radiologists um opinions of when they used AI and not used AI. Who did the responsibility go to? And most of the time it went to the radiologist, not to the software or to the to the technology. So I think that responsibility is still ours. And and so

00:37:27
 I think if we're going to do it that way, if if the responsibility is going to be ours, then I think we have to vet it out completely uh in a scientific way, not in a trendy way where we try to get the fastest, coolest technology. I think the idea is to get the best things going for AI, not the the fastest or coolest looking. I don't Oh, sorry. Last thing I wanted to say the speed is not there was a great one-page editorial by uh I think it was Eric Topal it might have been somebody else in nature

00:37:56
 talking about the scientific we're talking about efficiency a lot of the times with AI but is that really the most important thing efficiency I'm not sure it is like you know we have a scientific basis of this specialty you as a pathologist know that you know we we're dealing with hundreds of years of science building so have we built AI with that science background or have we built it with speed background and I'm not sure we've figured that out yet. Are we actually building stuff to enhance

00:38:22
 the science or enhance the speed? Um it's two different conversations. >> Yes. I think often it is with speed in mind. It is and when I'm thinking about it, I'm like okay quantification of things to do it faster to do it more accurately um as well. But you are right. These are two different conversations, like two different tracks of applying AI and they are often put into the same bucket and kind of confused. >> Exactly. I mean, we we weigh this anyway, right, Alex? Like in our in our

00:39:04
 lives, we weigh this anyway. Speed versus quality versus depth, right? Like [laughter] there's different attributes like Right. It's our whole life is like that. >> I know. I'm laughing because at one company where I was working like you can have it fast and you can have there was like three um different variable fast and something shoot I have to Google that because that's it was but but there was that you cannot have all three of them oh it was fast accurate I don't remember let me Google it so that I can

00:39:42
 actually say it Yeah, I mean we normally we all compromise to create the best combination, right? Like you want ideally all three of those things, but you can't have them all perfectly. So you try to do the best possible, safest, you know, cleanest way to get all three of them as high as >> cheap. It was cheap. The third one was [laughter] good. So fast, accurate, and cheap. >> You can have it fast and accurate, but it's not going to be cheap. You can have it [snorts] cheap and accurate, but it's

00:40:15
 not going to be fast. Or you can have cheap and fast, it's not going to be accurate. So, >> yeah. Yeah. Yeah. Exactly. I mean, yeah. Well, let's see where it goes. You know, I think we're we're going to watch it, but I think it's going to change a lot. My prediction is that in the next year or so, we're going to see pathologists get more involved in creating things. And if I know the mind of a pathologist, I think they're going to invent the ideas that are not currently easy

00:40:41
 diagnosis to make by by our own eyes. Like some of these AI tools are already straightforward for most pathologists. And I don't think that's a real effective tool. I think like if if AI can go into things that we don't normally do like micro environments and like looking at ratios of things that we don't normally look at, that is incredible. like then you're actually pushing the science forward. So, >> yes, totally. And I like these collaborations, so to say. I had a couple of guests on the podcast doing

00:41:15
 like developing image analysis where the pathologist visually like trains the AI, visually evaluates, but then the calculations are based on the visually recognized features. I think it was a podcast with Rish Pi and he was doing work on colurectal cancer and there are like I don't know how many features did you describe in the report of the colorctal cancer but he basically built AI to distinguish all those features and then made combinations of those features to to make predictions which I thought was fantastic because

00:41:50
 this is the augmentation that you want you have the >> no I think yeah you're right there's a lot of potential there and also for QA things and QC See, and we're I I think there's so much untapped potential in that world. >> Say, thank you so much for joining me today. Thank you for sharing your experience with the trailblazers. I'd love everybody to go fast in digital pathology instead of slow. >> Thanks so much, Alex. Appreciate it. Nice to see you again.

00:42:20
 >> So, I hope this implementation story inspires my trailblazers and I talk to you in the next episode.