149: AI in Pathology: From Pixels to Patients with Dr. Anil Parwani

If you've ever wondered what it takes to implement digital pathology at scale in a hospital setting , or how AI is practically used in diagnostics today , this presentation delivers clear , real-world answers .

So it's really many different journeys and when I speak to attendees , they share their stories . What are their pain points , what are the things that they're looking forward to solving with their journey . So how many of you are buying a new scanner ? A few of them . How many of you have completely moved to digital ? Several , several of you .

Excellent Greetings from Columbus . So I want to talk to you today about these journeys , the role of digital pathology and AI , and list some applications which we can do today . But I want to go beyond the glass slide , right ? So what are the barriers today which can take us beyond the glass slide ?

So what are the barriers today which can take us beyond the glass slide and where are we today in that journey and what are the potentials moving forward ? We have powerful microscopes , whole-site imaging scanning systems , but today we have technologies which go beyond that . So I also want to talk about that .

So it's really an exciting time to be in medicine and pathology . We are generating lots of data , lots of data which is now being converted into pixels , into bits and bytes , and we are making decisions on that data . We are sending the data to the electronic medical record . Patients are consuming this data in their portals .

They're looking at the diagnosis , they're trying to decipher that . What does it mean for my diagnosis ? So it's truly an enabler of clinical decision-making . At the end of this , when you buy your first scanner , when you implement it , the goal is how does it transform your pathology practice ? Right ? So that's what I want to talk to you about today .

So , as pathologists , we play a critical role in cancer diagnosis . Number of cancer cases continue to go up and the demand and complexity of our services keep growing . The workforce is continuing to shrink , there is a critical shortage of pathologists and lab staff and we are learning more about disease than ever before , right ?

So all these combined combinations of shortage of pathologists burnout new knowledge . It requires some disruption , it requires a new way of finding things , a new way of discovery and also a new way of making and enabling diagnosis . So , as pathologists , we perform many tasks , many manual tasks . We count things , we annotate information , we assemble information .

At the end , what is our report ? The report is our endpoint , but it's a continuum in the patient's journey , right ? So if I look at all the challenges in pathology today , which is lack of standardization , subjectivity , many manual processes , there are some labs even today which don't even have barcoding .

We were at the ASDP session yesterday and they did a survey which showed a lot of the labs around the world don't even have basic technologies . I mean , we take things for granted here , but globally there is a shortage of pathologists . Pathologists are overworked , right ? How many pathologists here are for vacation ? They're here to play golf , right ?

So if you ask pathologists in general , they will tell you we are overworked , right ? I see my friend Masood in the audience there from a private practice and he's always complaining we're looking for pathologists , we have more work than we can do , right ? So ? And there is an explosion of medical knowledge . So what does digital pathology potentially provide us ?

Going beyond the glass slide , is standardization , more objectivity , more automation , more accuracy . Maybe you can do it faster If you're first starting out . You're signing out your first few cases . You might say I'm slow on the digital , but once you overcome that , you will see the difference In terms of why today , why now , why in 2024 ?

So the current environment is suited

for digital pathology and AI models . Ten years ago , we used to ask the question which scanner to buy ? Right ? Many people used to ask that question . We now have cost-effective , high-performance computing , which is cheaper , which is more readily available . Many academic centers have supercomputers , so we now have that environment which is needed for growth of AI .

We also have better algorithms , right , so algorithms have continued to evolve and they have become commercial grade . There are some which are FDA approved , and then you have more data than ever before , right ?

So if you look around the room and if you ask each one of you collectively within this room , we have several million images scanned , maybe more than 100 million images scanned globally . Right , so we have experience . We have data . Is that data easily available and shareable today ?

No , but we are getting to that point , right , we have several collaborative networks and organizations that are working on it . We have more . We have FDA-approved algorithms , for example , for prostate cancer . We have clearly the cost of making an H&E slide .

Digital has continued to go down and adoption has continued to increase , right , so we have really evolved from early demonstration in the telepathology in the 1980s to 2024 , where we have several labs which are digital and have implemented AI , and if you walk around the exhibit area you can see all these products . It amazes me .

I've been coming to this conference for 10 years and every year there is continuous innovations and developments in this area , so it's an exciting time . If you think about , where are we , have we reached the ceiling in terms of making a glass slide digital ? So I think there are still innovations that need to happen .

There are specialized areas cytopathology , for example , hematopathology where we need different types of scanning devices . Maybe Z stacking , maybe alternative-stacking , maybe alternative light sources , I don't know . We continue to have to continue to evolve and innovate the whole slide scanning systems , but we can go beyond that , right .

So today you have a suite of scanning products out there . If you go to the exhibit area , you can see many of these in action and they can make slides into digital images . They can do immunofluorescence , they can even do polarization now . So this is a given right . So diagnostic quality images are a commodity now , right , everybody agrees , right .

Who doesn't agree ? Who thinks we need more work in this area ? So all the vendors are in the room and there is room to grow right , but on a given day . So I'm on service today , I'm covering GU and I can look at images which are diagnostic quality and I feel confident in my diagnosis as a pathologist .

We have achieved spatial samplings of 0.25 microns per pixel . We are now able to create multiplexing systems which allow us to interrogate multiple biomarkers . There was a system I saw yesterday in the Wendell area with 60 to 70 biomarkers could be interrogated on one slide . We have gone beyond the glass slide to a digital image and now we are moving forward .

So collectively , this is a really strong study . It's a metadata and meta-analysis of over 2,900 AI and DP studies which looked at the from multiple countries , over 152,000 whole slide images representing not just cancer but many diseases .

And if you look at overall from these studies , 100 were selected , 40 were drilled down into it 96.3% sensitivity , 93.3% specificity . So clearly there are studies which have demonstrated that glass slide and digital pathology are equivalent or digital images are non-inferior or not inferior . These studies must continue . We need to do more of these .

We need to do this in diverse populations , not just very focused on one region of the country , but globally , and I think when we discuss this at the ASDP , that is an important aspect of moving these forward . So any guesses where we are ? This is Orlando . Who thinks it's Orlando ? This is Columbus moving these forward . So any guesses where we are ?

This is Orlando . Who thinks it's Orlando ? This is Columbus , ohio . Yeah , so maybe this will bring some recognition . So we're excited Football season is on in Columbus right now . So I just briefly , with a few slides , want to show you where we are in this journey . We started in 2016 , and we implemented . We did retrospective scanning going back 10 years .

We started prospective scanning , we started primary diagnosis and now we have several pathologists who are in this room who are completely digital . We have David Kellogg here . He runs the operation and the scanning site , and I think several of our other team members are here as well .

But several of the pathologists now have come to a point where , if the digital system goes down , david gets a lot of angry emails . So we've completely turned over right . When we were first adopting it , there were one or two pathologists and they were just the lone rangers , but now many pathologists

are digital . David just shared this data with me yesterday . Many pathologists are digital . David just shared this data with me yesterday Almost 4.2 million slides scanned , almost half a million cases scanned . Year-to-year volume growth , you can see , continues to grow . And what I love about this is the user engagement and that's the key right .

You want to create an environment where users are excited to use the system . This is a critical piece of change management right . Change management is not easy , but it'll come with user engagement . If you have two or three we had a good discussion with my colleagues from UAB yesterday about this and how do you start this journey ?

It's getting a few people super engaged and get this journey started . So the scan slides are instantaneously available . They're linked to our lab information system and we are continuing to add more bells and whistles to this system . Right now we move from a case-level integration . Now we have slide-level integration so we can individually call out the slides .

All those 4.2 million slides are available . We can consult with colleagues easily . This , to me , has been a game changer . Our neuropathologists are in a different building . Our dermatopathologists are four or five miles away . If I have a difficult penile biopsy , I can just press a button and connect with them in real time .

And frozen sections and everything else is much more easier and streamlined because we have a way to share images . We have implemented PatPresenter for our consultation work and that's now being integrated into into our system , so scan slides are instantaneously available . There is no waiting for foldering . It doesn't matter which order they are scanned in .

They show up in your queue ready to be signed out , and we are starting to explore cytology preparations FNAs using small scanners on the bedside . So , in summary , digital slides have continued to improve my workflow . I love the direct interface to the LIS .

I love sharing cases with consultants and colleagues and flagging cases , looking at the prior cases , so it has improved my turnaround time . Because I'm not . Our histology lab is three or five , three miles away from the main campus , so it's allowed us to share those images more easily . All right , quiz now which city is this ? No , close to Columbus , though .

It's Cleveland . Very close to it . So if you take 71 , you have Cleveland , columbus and Cincinnati . So 71 corridor , right . So what do we do next ? Right , as the systems mature , as you start your own journey of buying your own scanners , implementing it as these systems mature , how do we go beyond that right ?

How do we start thinking about other things which we cannot do with glass slides as adoption increases . So , if you look at a typical product adoption curve , you have innovators , you have early adopters , but many of them hit this chasm right . So every one of you have their own chasms .

You might be starting your journey , you might be exploring how to implement digital pathology , but the key is finding out what is your pain point , what is your chasm ? How do I implement digital pathology ? But the key is finding out what is your pain point , what is your chasm ? How do you go beyond this chasm ?

And we are facing the same thing , right , as we build , as we think about implementing AI , we have to deal with integration , we have to deal with interoperability . All these are key things . So how do you go beyond the glass slide chasm ? Right ? So we have established in this room we can now create diagnostic quality images . How do we go from there ? Right ?

So what can we do with digital images that we cannot do with glass slides ?

So it really boils down to managing the information , sharing the images , connecting with each other , connecting with experts , but also exploiting the pixel pipeline to build algorithms or buy algorithms to identify , quantitate , synthesize and create knowledge pathways , create important clinical decision-making skills , right ? This is the next part of my talk .

I'm going to focus on what are some institutes doing and where are we today with this chasm ? Where are we today with computational pathology and AI for clinical decision-making ? So these are some of the things which I feel are already here today , in 2024 , right , we have very sophisticated image analysis algorithms , ai algorithms for detection and diagnosis .

Right , to analyze digital images , decipher the pixels , identify features like morphology , cell shapes , size of nuclei , architecture and staining patterns .

So , pattern detection , feature detection these are things I learned as a resident , and I learned that by looking at over and over again and building an algorithm in my brain which allowed me to look at a prostate gland and say , okay , this is cancer , because A , b , c , d , e , same thing , right .

So we are at a point in 2024 where many institutes have implemented AI in pathology . Right , and they started with very basic things counting cells , looking at the size of nuclei , differentiating different nuclei , differentiating positive versus negative signals . So all these focus on biomarkers which could be diagnostic or predictive .

Right , so these could include detecting , classifying , segmenting , quantifying and localization . Here are the four applications that we are starting to use in our Institute . And again , we have our own chasm . Our chasm is that we cannot readily integrate many of these algorithms .

So we still have to go to a third party , launch the algorithm and use the algorithm and bring this information back to the LIS . In an ideal world , in my wish list , we want this to be completely integrated and that's where we are heading towards . But quantitative digital image analysis for biomarkers this is very , very routinely done .

Now there is a separate CPT code for this . You can actually get paid a little bit more , maybe , I don't know twenty , thirty dollars more for a digital image analysis system . If you use in your lab , it's more objective , more accurate , more faster . But is it easy to do ? Is it cheap ? So the answer is it is easy to do .

It's easier to do if you already have digital slides in your system . It may not be the first application you launch in your lab , right ? If you look at this gastric neuroendocrine tumor and you can see a 2-millimeter square area was analyzed in 28 seconds .

So if I asked each one of you in the room to count the blue dots and I give you three seconds , you all will have a different answer . Right ? Everybody agree with that hypothesis .

And if I showed you this , it's gonna be even harder , right , but we can get this data objectively and it's reproducible every time you do it , if the answer will be the same in Cleveland or in Columbus , as long as you're using the same algorithm and you've validated it . It's a locked system and so on . These are a given .

Other type of algorithms we are using is detecting rare events . This is a lymph node detection algorithm , right , so we can actually launch it directly from the viewer today . But we still have to do a lot of manual work to get this algorithm queued up . But pathologists still use it . Pathologists still want to use it .

So imagine a world where this is even easier to do . It will become much more easier to use . So identifying metastatic foci , and the pathologist in the room might say why do you need an algorithm ? I can just eyeball it . This is cancer . What if you have a few rare cells and you missed it ? But the computer didn't miss it ?

These are type of things I call them rare event detection . Overall , right , it could be finding microorganisms and so on . Let's go beyond that , right . So today we can find metastatic cancer in lymph nodes . But we want to go beyond that right . Pathologists

today do immunostains to figure out which cancer this is Tumor of unknown , primary , unknown origin we do several immunostains . I had a case last week . I had to do 20 immunostains and consult a metopathologist and soft tissue pathologist and even then you know guess how it was signed out High-grade malignancy , see comment .

So this is an ongoing issue in diagnostic pathology and today we have algorithms which can predict for you , just like when you send it for next-gen sequencing . It can predict with greater than 95% certainty this is renal cell carcinoma . So this is an example of such a prediction model where the computer has predicted this is colorectal cancer .

We have GI pathologist , dr Chen , in the audience and she will say why do you even need AI for this ? I can look at this and say this is colorectal cancer . It has necrosis and dirty necrosis and all the features of colorectal cancer . But the point is there are algorithms out there and they are available .

They will continue to evolve and get better and when you are ready in your journey , you will buy that algorithm and use it . But before you do that , we still have to solve the interoperability issues . We still have to integrate them and some sites have , some institutes have done better with these , others have not .

So if you look at diagnostics for cancer overall , they can help you in a pre-sign-out process . They can help you during sign-out or they can help you post-sign-out . In pre-sign-out setting it could be a good screening tool . I showed you the data from hundreds of studies about sensitivity , right ?

So for prostate cancer specifically , similar studies have done and shown high sensitivity and high specificity and it's one of the most commonly exploited cancer for building algorithms . Like every company I talk to , they're building their own prostate cancer algorithm , pre-ordering IHC , right ?

Imagine a world where you're coming to work and the system automatically screens the cases and chooses the ones that you should do immunos on , and doesn't do it automatically but makes recommendation to you During sign out . It can find other features , like intraductal cancer of the prostate . It can help you with supervision , primary diagnosis or without supervision .

It can also create automated reporting templates for you and then do a second review , like what about post sign out , right ? So imagine a world where you are overworked pathologist , you've just booked a ticket to go to Las Vegas and it's six o'clock and you're trying to rush through your cases and you make a mistake .

But you have the safety net , you have a gatekeeper , you have the AI assistant . They check your work and say wait a minute before you board the flight . Are you sure you want to call this cancer ? And you'd look at it again and again and maybe you change your diagnosis .

So some of the studies out there have looked at this specifically and there are cases where small foci or cancer were missed . Did it make a difference ? Maybe not . Maybe there was cancer in other cores , but what if it was the only core and you missed it and you wish you had the system , or not ?

So how many of you if this was free right , all the vendors will give this to you for free how many of you will use it ? And what if it wasn't free ? You'll still use it . So again , we are also , as we build and use these algorithms , we're learning new information about these cancers .

So , again , these tools are becoming , I would say , not in routine use , but they're getting very close to routine use . So you can see this work list as you are looking at your work list for the day and the cases have been flagged for you .

So I would say we're very close to getting a routine use of these algorithms and , again , you can turn it on and off . If you're in a residency program , you can have the resident use it or not . You can use it as a way to assess their competency , right ? You can make them review the case and then turn on the AI or make it so .

Ai will only be turned on after five minutes of review . How many residents in the room You're ready to use AI , right ? So this is an example of breast cancer , right , invasive lobular carcinoma . And again , the red areas are where the cancer is . You can go look at it more closely .

You can also have algorithms which can distinguish the subtypes of breast cancer or subtypes of prostate cancer . What about finding mitosis ? Who likes to count mitosis ? Residents , you like to count , you know ? I asked , I did this test .

I asked three residents to count mitosis on the same slide and I got three different answers and I would say pathologists would also give the same different answers . Right ? But you don't have to , right , today , if you have AI , you've implemented it . They will find the mitosis for you and let you check it or not .

So again , this is an example of lymphoma classification , or finding abnormal white blood cells in a smear . These systems are already in place , like many of these systems are being used in the clinical lab . What about finding acid-fast bacilli ? Who likes to do that ? These are algorithms which are available . Some of the centers have implemented it .

What about finding H pylori ? So these systems are being built . They're going to become commercially available . There will come a time where you will have your own personalized dashboard of apps that you can download from the App Store , and you'll have to pay a subscription to it .

Some of them will be free , some of them will be per click , but these are things which we didn't have when we were talking about whole slide imaging . So let's go beyond that . What else is coming ? So we were talking about whole slide imaging , so let's go beyond that . What else is coming

? Right ? So I talked about the whole slide imaging . Is that the end of the journey ? Is that the end of a glass slide ? What about starting from the tissue ? Right ? So we're getting closer to radiology , and radiology is getting closer to pathology , so it's a spectrum . Right ?

So you have synthetic data , we have virtual staining , we have 3D pathology , integrated large language models . We have all these new tools which will become available in your work list soon , from pixels to diagnosis , to prognosis and predictions . Now we talked about the segmentation .

But imagine a world where I want to go inside this segmented area and pick one or two cells and interrogate those cells . What if I want to look at this image in three dimensions , right ? What about just finding similar things , right ?

Today you can take a picture on your phone and you're going to the supermarket and you take a picture and find similar products . It gives you the prices . What if you could do this in pathology ? When I was a resident , how did I learn ? I learned by opening textbooks . I was in the resident room and it was 7.55 .

The attending was going to come there at 8 o'clock . I didn't know what this was . I wanted to scribble something , so I flipped through pages and I put something down . At least I tried . But today this could be done electronically .

It could be done using artificial intelligence , where , in this case , the query image was a brain ependymoma and the algorithm found the best matches which were similar to this and created . At least it gave you an indication of what you're dealing with . And you're going to see examples of these different things , like virtual staining .

So if you walk around the exhibit area , there are vendors out there which have worked on it , right ? So this , to me , will be going beyond the glass slide . So what about prostate biopsies which are unstained ? On the top panel A , you have unstained prostate biopsies . In panel B you have H&Es which were generated in the lab .

And in panel C you have prostate biopsies which were generated by virtual staining , and then you can actually see the areas where the cancer was . So what about staining , right ? See the areas where the cancer was . So what about staining , right ? So , as a prostate pathologist , we do prostate triple stain to look for basal cells .

So here you see , on the right-hand side you have real HNE and on the left-hand side you have virtual HNE . But what is amazing to me is the ability to now predict where the stain might be and even predict intraductal carcinoma , which is a difficult diagnosis , in my opinion , you know , and there is a lot of controversy around it . So 3D pathology is coming .

We have commercially available scanning systems now which can scan in 3D , right ? So here you have a cancer prostate core and a benign prostate core . So what about helping you while you're signing out , right ? So these are slides from Dr Singh , who's built this in Path Presenter the ability to create a chat with the image .

So imagine a world where you're looking at this image . You start chatting with the chatbot what is this ? And help . These large language models can help proofread your report , alert pathologists if you had missing information . You know this happens when you're in a hurry . You're rushing through a case .

You might miss information , you might put something like a T4 where it was actually a T2 . What if you had this way to proofread your reports and help with billing right ? So imagine a world where you have this image where you can press a button and look at the features for the residents and the trainees .

It can highlight the features in this image , but it can also suggest a diagnosis for you and suggest some IHCs for you and also try to help you find similar cases , create a differential diagnosis , take these images and put them into your tumor board pile and present it at the tumor board .

So these type of tools are coming in many areas in pathology and again , we're not there yet today . But what I'm showing you today are prototypes of things that are coming in the pipeline . So we've gone beyond the glass slide . We have now demonstrated that we can make a good diagnosis on digital images .

We have demonstrated that there are algorithms today which can be used clinically , and I've shown you examples of workflows that are coming soon . So I just want to also separate artificial intelligence and real intelligence , right ? So when you think about artificial intelligence , it's very task-oriented . Right ?

So when you think about artificial intelligence , it's very task-oriented . Find me this feature how many nuclei are positive ? Biomarker quantification , lymph node met ? I showed you all these examples , but they're very task-oriented . Each algorithm is composed of small steps , but real intelligence is goal-oriented Many algorithms per task .

So , as a pathologist , if you look at your journey from residency , from medical school , to where you are as an experienced pathologist , it's experience-based , it's context-based , it's knowledge-based . So we're not there today where we can say AI is equal to real intelligence , right ? So that's a journey , just like digital pathology is a journey .

So this is an example of a case of prostate cancer where , on one , one of the cores , I have clearly established prostate cancer , right here and then this is which looks like prostate cancer , and I actually signed it out as prostate cancer , and that's where real intelligence comes in .

So I signed out this case , I released it into the patient's medical record , but something bothered me when I reached home . You know , that night . I just had a thought that this , maybe this is rectal cancer . Maybe we should think about it . So I dug through the notes .

I found a note from one of the primary care visits where this patient has rectal bleeding but refused endoscopy because it was too expensive . So I went in , I ordered some stains and it turned out to be colon cancer . So I had to admit I was wrong .

I amended my report , I called the urologist and the patient actually got treated for rectal cancer and prostate cancer . So I think that's where we need to be and we might get there in 20 , 30 years , but we have significant advances in the field even today . So , putting it all together , everybody went to Magic Kingdom .

I was there last night Amazing fireworks , the best fireworks . I mean , I went to Disney with my kids many years ago , but I went now Amazing , so I recommend it highly . I don't have any stocks in Disney or anything . So , in conclusions , how will digital pathology and

AI help pathologists ? Right ? So we talked about the decline in number of pathologists , increasing workloads , fewer trainees going into pathology , and clearly what I've shown you today can help some of those issues . Help us with sharing cases , help us with connecting subspecialists together , connecting spaces together .

We have increasing workloads , so are we ready for a digital disruption ? And then augmenting your diagnosis , checking your work , do some of your work and share the work with others right , so you have to tame your own digital pathology chasm , right . Each one of you is here in this conference because you want to learn more about it .

If you've already bought the scanner , you want to learn about AI . If you've already implemented AI , you want to learn about 3D pathology . But it's a journey , right , so all of you have to solve it on your own . I'm not going to you know . I'm just showing you where we were 20 years ago , where are we today and where are we going 20 years from now .

So maybe you will have a workstation like this in your office where you will customize it . You'll buy your own apps . You will create your own chasm building , right , so you will use AI to assist you to augment your diagnosis . And maybe autonomous drive , right .

Yesterday , my friend drove me from the fireworks back and the car was driving itself and he left the steering wheel and I was super scared and it was pretty crowded . It was Sunday night , it was rush hour . I said , no , I'm not ready for this . So take over the steering , please . I have a talk to give at 7.30 .

So he didn't listen to me and the car drove itself for 10 minutes and it did fine , right . But you have to build that trust . You have to build that trust with AI , right . So where are we going ? Right ? The journey from glass to digital to prediction will continue for all of us . We are today . We can improve our analysis .

What will come is next , we'll be improving your diagnosis . Maybe we will have more integration .

We will solve the interoperability issues , regulatory issues , reimbursement issues , but in the future , this will really be clinical decision-making exercise and integrating multiple types of data , and we're probably this will be the era of AI-based precision medicine in its true sense , and that's probably in pathology .

Ai could probably diagnose easy cases independently , just like the Tesla was driving autonomously yesterday , but I don't know how many pathologists feel comfortable about that riding the Tesla of pathology . So with this I'm going to conclude . It's been an amazing journey and I think we're going to . I actually encourage you to talk amongst yourselves .

Go to the exhibit hall , look at all these different products out there , and there is something for everyone . You know you might be just buying your first baby scanner . You might be just getting that one big check from your administration and you're ready to spend it . This is the place to do it . This is the Disney world of digital pathology .

So with this , I'm going to end with a picture of football again , which I'm excited . Next weekend I'm going to be at the game , and so I'm going to stop and if you have any questions , I'll be glad to answer it . I'm going to be around for the rest of the conference , so I hope to interact with many of you .

You and I want to thank Apreedia for inviting me here as a speaker , and looks like a full house . I'm sure it's because of the breakfast . That was really good .

3D pathology and actually everybody who is in this room and who comes to this conference , they can be at any like single point of what you described . So let's start with the beginning , and you were talking about the chasm . So everybody has their own problem , their own pain point to solve . But let's start with the scanners . What scanners do you ?

So I think one scanner may not address all the needs of a big academic center but a smaller lab can actually do with one or two scanner types .

But there are so many complexities in pathology and I'm glad many of the vendors out there , including Apreedia , leica , are actually solving those problems individually or incorporating them into your dashboard , into their system .

You know , just like when you go to Best Buy to buy the next TV , you have all these bells and whistles , but they specifically build those based on feedback . You know , today I cannot do this . Can you build this ? So I think digital pathology scanner market is also evolving in that direction .

To me , it's like multiple trains have left the station and everyone has their own train and they have their own next stop . The next stop might be I want to do frozen . The next stop might be I want to use image analysis . I want to do large language models . I want to go directly from the tissue to an image , you know .

so , all these trains , individual trains , are users and they have expectations and they have end and they have an endpoint . Not an endpoint but a stop on the way . And they want to get off the train at that . Stop , do something and then get on the train again .

And I think that is an important thing , which is important for your listeners also to know .

How do you do that ? That's the next step , step next stop on this journey .

A good leader is also a good change manager you know if you're not , you're not going to be a good leader if you cannot manage change . The change can occur at people level . The change could be at the technology level , it could be financial , it could be regulatory , but collectively you have to orchestrate all those changes to see and execute your vision .

You have to hire the right people , the right team and keep them engaged , keep them excited in this journey so what is your chasm right now ?

Do we need to create technologies which are not compatible and try to bring them together ? So that is the chasm we're trying to do . You know we can create diagnostic quality images . We know we can use them .

I know we can make diagnosis on them , but I think the next gap is can I create a dashboard where I only have three AI tools but they could be launched from one viewer , and so on ?