 So, for our last talk in hall B, we have Dr. Akilesh to talk about python-based simulation and treatment planning in craniofacial surgery, the advances in the challenges. Thank you. At the outset, it's a statutory disclaimer that I would like to give here, that I am associated with the Government Institute, Oral Institute of Medical Sciences, and abused opinions expressed here on my own, and did not reflect my employees, organisations or anything. I have a long stint with many Government Institutes earlier. My training was in a Government Institute, then I worked again in Jipmar, in Pondicherry, then right now I am with Oral Institute of Medical Sciences at Gurugpur. And I also have no financial interest or any conflicts of interest to declare with any of the codes, frameworks, people who were, you know, any authors that I have cited over here. So I come from Oral Institute of Medical Sciences, Gurugpur. Basically I am a dentist who specialized into Oral and Maxillofacial Surgery, I will come in details of it later. So there was this interesting discussion that I came across, is python the new basic? You know, in 80s and 90s, the initial programming language used to be basic. So the question that came in was, is python the new basic? So for that, I had an answer that it's not the new basic, it's actually better than the new basic. So why this presentation? You know, is it just because this is a, you know, Pycon India, Python based conference, so is that the whole basis of it? No, that's not my reason. The reason for this presentation is that wherever there is technology meets software, there is python that is actually coming in. So the other area that came in was, I am a surgeon by profession. So the question comes is where technology, surgical technology, you have robotic surgeries, navigational surgeries and so on. So I am trying to think about where such technology comes to meet software, what would be the role of python or python based projects? So who am I? I said I am Akhilesh, Assistant Professor in Department of Industry, Ames Gurugpur. So I can call myself as a tinker, tailor, soldier and spy, I am a tinker because I fix broken jaws, I fix jaws in general, I am a tailor because I see wounds in the face, I am a soldier because I am constantly challenging myself and I call myself as a spy because I am always adapting and adopting the new tools and strategies that need to be done. So what lies ahead is, you know, the five core areas that I would, sorry, four ones, four that I would like to actually concentrate on is first one is python and medical imaging, what is the current status and the possibilities and opportunities in cranium axial facial surgery, which is my area of specialty where I am using some of the python based tools for some of the treatment planning aspects and then a little bit of brainstorming about other important 3D frameworks that is available and some python based projects or frameworks which are useful in other surgical areas. So these are the areas in which we will be covering the forthcoming talk. The current status of python and medical imaging, the good part of it is that it is simpler. There was an interesting talk that's been, I think, Sadhanar, somebody was presenting I am not sure on that, coding in python is not actually easy. I am sorry if I got the, you know, the speaker wrong, but I saw the talk in, you know, the list and simpler, maybe easy, I am not sure, programmers from your site and since you are programmers, you would know better, but there is a convenience for abstraction of data. You know, it's easier to abstract data when you use python and there are many ready for use successful python wrappers available, which could be used up, you know, which can wrap itself around C++ codes or C codes or even Java, you know, there is lots of wrappers available, which could be adapted and there is no deficiency of enthusiastic python status. So this is the area that is a good part of, you know, python and medical imaging. The bad part of it comes in, you know, every part, wherever there is a discussion on using python and medical imaging, the challenge comes into the processing speed. When you are dealing with a single big object, you know, python manages well, but you know, when that, when it is a fragmented part, the processing speed increases, you know, it is not as efficient as the other languages that have been the workhorse of image analysis, say C++ or C, that's the, you know, the primary limitation that I came across when I was going through, you know, the softwares that are already successful. So this is the whole bad of it. So what's the ugly part of it? What we find is that there are so many orphaned experimental and experiential projects related to python. And there are no specific market drivers when it comes into, you know, medical imaging part for python, per se. And ultimately, you know, what's happening right now is that python is very good for creating prototypes, you know, you can speed up that process. So you speed up the process, create a python based prototype and then you, you know, you just put that code, whole code into some other language. So this is what is happening right now. So the verdict is that successful projects have done something like a dub smash, you know, so what they do is the heavy lifting, you know, the weightlifting is done by other codes, code bases like C or C++ and python is more or less like a modular aspect, which is more into the, you know, the programmers part in enhancing the usability of that particular software project. So the verdict I would say is that python is here to stay. So with this, we'll just go ahead. So when I just made a search, you know, into medical data in PubMed, one of the search databases for medical publications. When I searched for python and software, it was just done two days ago. You know, you can see the exponential increase in the number of publications. Increase in number of publications actually means there are lots of people who are interested and who are working on that area. There are lots of work coming out in that area. In fact, the talk that I gave some say eight to 10 months ago in Chennai Pai meetup group and right now, you know, when you just say in 2019, there were just three papers at that time and right now there are 119 papers that have already published in this particular heading. So I mean, I didn't do a detailed search. If you're going to go into the subcategories and subbeddings, definitely you'll come up with more. My idea is to say that python has lots of user base, there's lots of opportunity in the healthcare industry, especially including imaging. So most of these papers are concepts or concept papers. They don't translate as end user projects, usable projects. So the request here for me, you know, from an end user, you know, as a surgeon, as a clinician, to a programmer is that we need to bridge this gap. You know, there is a gap between the research part of the concept papers and its implementation. So my request is that programmers should actually come up in developing projects which are usable for the end user for the clinician. Potential areas of translational research that's available be computer vision, image processing and preprocessing. Again, most of these titles or most of these topics would be such that they are not totally native to Python. Most of it would be wrapping around other language frameworks which have to be adapted or you have to wrap Python around it and make it usable. So the workhorse of medical imaging happens to be this VTK and ITK. You know, your visualization tool kits and image segmentation tool kits, which is a registration, image segmentation registration tool kits, ITK. So these are the workhorses for image analysis and that is what is the workhorse for even medical image analysis. So you do have already successful Python frameworks available, wrappers available, which should be utilized to expand the utility or usability of Python-based projects in areas like augmented reality, virtual reality and such. And the other challenge that comes is compliance with standards. You know, the data has to be transferable. So whatever data that we get has to be transferable to others without any problem. And among different softwares, it should be transferable. And most of this workflow is not a single stage. You know, there are different steps and things available before it could be totally usable for a single purpose. And machine learning, deep learning, and then, you know, semi-automatic and automatic image manipulation analysis, it's already in the pipeline. Again, many of these projects are trial projects which have been left undeveloped. Most often, that's what is happening. So where does Python fit in my workflow? As a cranium-axle-official surgeon, what we see is, in Blender, when you say a cranium-axle, you say, you see, this is the part where there is a big gap, you know, after surgery in the skull. So how do you actually bridge that gap? Is that, you know, there is a virtual... I just fill up that gap with a virtual material here and this material could be 3D printed using biocompatible materials or it could be duplicated into biocompatible material once this structure is gone. So the fit is going to be better. This is a simple usage analysis or a use case scenario. There is another well-developed or an almost-developed project which is actually a customized Blender project. It is called Orthogon Blender, developed by a Brazilian 3D designer, Cicero Mures. So this has a total workflow. Like, say, it includes every step or almost every step that is available in a proprietary software as well. Say, there is a photogrammetry done where the patient's surface, you know, facial scan is taken. That is included or that is actually adapted with the skull data that we already have from CT scans. And based on that, we need to actually find out how much of bones to move. This happens to be my area of expertise where we need to restructure the, you know, the jaw have to be moved to a new position so that, you know, the structure of the face could be changed. So this is one area where we see if how much of, you know, movement of what part of the bone might cause what kind of a changes. So we are virtually creating or doing the surgery in this particular aspect there. Different cuts that we actually do in the operation theater. See it in the virtual arena. Find out what might be the potential result that is expected and what we really want. And based on that, we could actually plan the surgery. So with this, you know, I can make any number of cuts, any number of times without actually causing any harm to the patient. So this is a very potential area. And what you're seeing there was a heat diagram which showed how much of stress is there, how much of changes at what part. So all of this is already built into blender. It's about adapting that to a use case scenario as a module. So here we are just seeing how much of, you know, if slight movement of the maxilla, the upper part of the jaw would create how much of changes in the upper jaw or the profile of the face, the soft tissue structures of the face. So the other area that we come into place is that where we would benefit from such a three-dimensional plan is that when there is a loss of bone, you know, after cancer surgery, there is a loss of part or maybe, you know, whole of the mandible, the lower jaw. So we need reconstructed and it is common practice to pick up a bone from the leg and place it there. The bone from the leg is a straight bone. Mandible is a curved, you know, three-dimensional complex structure. So how do I adapt that bone to this structure? So it's a question of actually optimization, you know, it's a question of finding the optimal cuts that I can do and, you know, keep the bone so that it will fit into the near normal profile of the patient to try to bring him back to near normality. So this is not something that we can experiment inside the operation theater. So usually experienced surgeons are able to, you know, by their experience, they've already got the skills to do that. But still it is time taking in the operation theater. So perhaps with this pre-surgical planning, it will be possible for us to, you know, make two or three templates and find out which is the best fit and make that, you know, from that template, you will be able to find out how much of bone to cut, where the bone has to be cut, which is most compliant or which is most congruent to the mandible and also create, you know, what do you see there in the second and the fourth one or actually the depth, the cutting guides, you know, which could be 3D printed. So I would know how my saw or my surgical saw or my surgical blade, the cutting saw has to go in which path to cut it into precise areas accurately so that, you know, all of this is actually automated or semi-automated. And, you know, the operation time is decreased and the outcome is better. So this is another use case scenario which has been published. So 3D slicer and CMS slicer, again, these are Python-based tools. I mean, most of the modules here are Python-based. Blender, everybody knows it's, you know, it's Python-based. Of course, the core is written in, you know, C and C++ if I'm not wrong. And, but the modular part is basically Python. This is why I said it's like a dub smash. The heavy weight lifting is done by C and C++ codes, whereas the modular part is done by Python. And it is working out good. You know, Blender was able to bring out a whole totally animated, you know, film just based on, you know, just with Blender as these things. You know, the day where they had to use some other property software was gone and they just showed it is capable of doing it, you know, a production work. So the 3D slicer and CMS slicer are somewhere similar. CMS slicer is a branch of 3D slicer, specifically for head and neck surgeries. Your 3D slicer is actively being developed program, which is also, again, written in low-level language, whereas it has modules which are based on Python. So here, say for example, you have an accident where you have, you know, parts of these skull bones fractured. So this is actually an implementation where 3D slicer has a VR module, a virtual reality module. So it is easy for us to actually, you know, manipulate the bones, segment the bones into different parts, the broken pieces, and I can manipulate the, you know, each part of the bones to put it into that specific part and find out where it would actually fit in. It's like, you know, solving jigsaw puzzle online. And I don't have to sit and do it on the table. You know, it is possible for me to actually try to do it here and find out what is the defect and based on it, it might be possible for me to devise, you know, a better surgical plan in this particular case. There's another one area where, like, you know, neurosurgical procedures. Again, it is a VR-based module where it can be used for surgical simulation as well as for training purposes where a very tricky surgery, you know, a screw is actually inserted into the backbone. You know, inside the backbone, you have a very sensitive structure called the spinal cord. You can't touch it. Anything touches there, there's going to be some nerve damage and everything. So what is a path in which this particular metal item has to go and reach so that it doesn't damage any of the structure? So this can be simulated prior to the surgery here, you know, on the module. The second thing is it is possible to use this as a training module for surgeons. You know, surgical trainees do not have to do this on a patient. So without any harm, they can do this part. They take the patient data and this could be simulated by using the VR technology, using haptic devices. So they can actually manipulate that screw, pedicle screw, the screw that is there that could be inserted into the bone and to find out what is the optimal path and if they really violated any of the biological areas. So this is a possibility again with 3D slicer with its VR module. So again, the same thing could be adapted even to CMF. There is already an attempt on that where for the craniomaxiophysus surgery, also such a plan could be done, where the cuts on the bone could be made by using this haptic devices. Again, it could be a very good area for surgeons to practice. So the question that comes here when it comes to surgical planning, this prediction, predictive precision versus practical consideration, the sense that the reality, how much of the surgical simulation is going to be as real as it is in the patient. Of course, there is not going to be any bleeding when it cuts something here. There's no blood, there is no distraction for me, whereas in an operation data, we are going to have other factors. So when you're talking about training, we shouldn't miss out on that. So there is a potential area for other 3D environment designers to actually try to give as high fidelity as possible when it comes into surgical education or training. Real-time response versus lack. So we talked about this earlier. The processing time, when you need real time, we cannot actually code totally with Python, because there is a possibility for a lack. So this is a major limitation when it comes to that, but of course that could be offset by using a low level language and in combination with Python. So that's what has been shown earlier. The importance of precision and the level of precision that is required. Say you have a Raspberry Pi based Robo, which has to actually navigate this particular hall. So you have a laser-based distance estimation system, obstacle identification system. So it's almost like real-time. There is a data that is collected and the Robo reacts to that almost on a near real-time basis. So this similar kind of technology is what may be required even for surgical simulation. So there is again a very good area of working and expanding the uses from here to there. And the important aspect when it comes to programmers as well as to the clinicians is that we need to understand the limitations and aim for maximum productivity at a minimum of cost. So this is again important because most of the projects that have implemented are open source. So the problem of cost doesn't come there, but how about the support there? How about it's actually the usability of it on a end user basis? How can it be actually used in a production environment? It's okay to research on it, but how much is it actually reliable on a real-time scenario? These things have to be vetted, have to be researched, have to be tested, tried before they are actually implemented. So it takes lots of efforts from both the sides. So it should be patient-centric and user-centric and it should be modular. So these are the requirements. The other areas are, there's a big list of areas where Python can actually come into place like stereo vision, photogrammetry, and near lifetime 3D scanning with precision. Again, this is already partly there when it comes into navigation systems. Stereo-endoscopic visions, this is again a part where usually in endoscopic visions we have only a single vision. It's a monocular vision. It's like closing one eye and having a look. But if you have a two-dimensional, I mean a three-dimensional view, you need both eyes. So a stereo vision is what will actually give you the depth perception. So deep learning and automation is also in an area that is actually catching up very high in healthcare. And it should go beyond just the data I feel. Just data analysis is what is actually taking up more of this deep learning, machine learning, and stuff, and artificial intelligence. I think it should come into areas of image analysis, patient-centric care, and everything. And image-guided radiotherapy, again, most of it is image-based where you direct the radiotherapy at a particular spot. So how do you plan the direction of the radiotherapy? That could be done by using the software that I had told earlier. Surgical navigation, robotic surgery, again, there is a big lag, I told you, because robotic surgery and navigation would require almost very fast responses and real-time processing, which is a limitation for Python. So the question comes as to, as a programmer, where do I start? It's not where do I start, but it's a question that comes to every programmer. So I would say the best place to start would be adapting VTK and ITK into Python, and integrating that, because these are the workhouses for image analysis, as well as for medical image analysis. And there are lots of actually training modules available, even in Slicer, even in Blender, especially Slicer has very nice things. I have a list of resources that I'm sharing again, and I think this will be circulated as well. I'll make sure that that is done. So the other thing is, the haptic devices have become so commonplace. Like, you know, it's like, even a consumer can use it these days. You know, it's become so commonplace. So how can we utilize those haptic devices for at least, you know, part of the education purposes, or for innovation purposes, or for actually, you know, patient education purpose, or surgeons' education, clinicians' education. So these are the things which we need to adapt and hook onto at an early stage. It's at a very early stage right now. You know, the market is at a very early stage, as far as healthcare industry is concerned, with respect to these haptic devices, feedback devices, and everything. And again, the question comes as to, how can you make a realistic, three-dimensional environment? So again, you know, there are lots of players, there are lots of places to innovate, and lots of areas where programmers can make a difference. Design engineers can take a place. You know, any augmented reality and virtual reality workspace, you have the work of a design engineer, you have the work of a programmer, you have the work of hardware people. It's a big platform for lots of people to work on. And yeah, these were some of these publications which were actually relevant and which have kind of published workable modules for surgical planning. And these are the resources for those of you who are interested. It has a big list, you know, the, when it comes to the second one was the one for Orthogon Blender, then CMF Slazer had it, Slazer had it. Slazer has lots of nightly training modules, which is talking about, you know, specific image analysis for different medical purposes. And it is, and what we saw about the surgical navigation platform that's called SNAPI, that also has, you know, they're trying to use it for, use Python based projects for surgical navigation system, which is a cutting edge right now. You know, surgical navigation system, robotic surgery is not widespread in India, even all over the world it's a leading edge. I would say bleeding edge rather, actually. You know, so this is an area where Python, Pythonistas should catch up. And of course, I'm sure most of you would be aware of most of the names, but you know, Pillow is actually a branch of this PAL Python image library. SNAPI is a project that is one more called a SNAP PY, you know, that's a different one, but the SNAPI is a whole project for navigational projects and made by PyDicom is for actually manipulating your Dicom data using Python and you have tomographic reconstruction frameworks. Like I said, there are lots of frameworks available, there are lots of projects available. Unfortunately, many of it have been orphaned, some of it have never gone beyond, like, you know, they were developed, say, the last update you can see in JIT Hub is like, perhaps five years ago, four years ago, you know, no idea what happened to the developers or what happened, mostly it's like a PhD thesis or a, you know, master's thesis and then it just, I should say, I wouldn't say dies away, but rather, you know, I would say it is just quiescent, you know? So the perfect place I would say to begin is where you are right now. You know, all it needs is some motivation, internal motivation to get this done. So that brings me to the end of this talk. Hello. Yes. Thanks a lot for coming. This is absolutely mind blowing. Thank you. We are usually just playing with data sets and GraphQL API as your saving lies with code. So how did you actually came across all of these technologies and like, how did you integrate with the programming world? See, the part where I integrated is only the small part where I talked about the blender that is used for orthognathic surgery. And of course, there are certain other modules which are being used for dentistry as well. Again, which has been implemented in blender. I came across this quite accidentally. You know, I was just trying to find out, you know, I have my interest in open source softwares. So whenever it's just hunting around for frameworks, projects, readymade softwares, or something that can be adapted, I just stumbled upon it. So when I dug deep into it, you know, we just keep coming and you know, Python, I think as Pythoners you will agree that it kind of gets onto you, you know, once you know, okay, fine, you know, it makes your job so simpler sometimes. Is it so easy? Maybe you should try, maybe you should learn. You know, it started out like that, you know, but I'm not a hardcore programmer, but it is something that I'm kind of experimenting. Yes. Can you hear me? Yeah. Yes, sir. So hi, thanks for the talk. This was really great. It was actually high opening. I just have, okay, the question I have is not really regarding programming, it's more about adoption. So I have seen people in industries like, suppose the aerospace industry, when you tell them, let's ditch MATLAB, let's use Python, they start staring at you. You must be facing that at a hundredfold. If you tell people, let's start using Python and some open source library to solve our problems in dentistry. That's definitely the issue there, but when you see this, it's not totally true, I would say, because one of, there was one project that was related to dental CADCAM software. Like, you know, where you can digitize the whole dental workflow. You know, you take a three dimensional impression or something and the whole part of constructing a broken tooth, you know? That is actually, it has a very good engineering concept. CADCAM technology is very well integrated into it. So that technology was developed by, you know, a dentist. And that person has actually ended up, there is an open source framework still available there, but ultimately that person has shifted himself into another company and has, I mean, improved his development there. So this happened. So it's not always true that this, it's about actually, you know, trying to show the user case scenarios. You know, if you can compete with proprietary softwares, it's going to be a challenge definitely, you know, an open source to compete with the proprietary software. It's going to be a challenge, nevertheless. But the question is about actually, you know, the whole VTK, ITK, you know, that whole framework is open source. And the development came in that being the, you know, the standard, the workhorse for whole image analysis. You know, I think that's a very good example. You know, that's one area of innovation where if not in the end project, you know, there could be some areas like this, like even photogrammetry is one area where we could integrate your C++ codes with your Python and use it, your open CV is again one area. Yeah, thank you so much. Yes, sir, thank you. Hello, doctor. Yes. Thank you so much for being here. I think almost everyone in this room would say the same thing. It's an absolute eye-opener for all of us. Just to learn about how we are using stuff. I had a very simple question. I think even through the whole talk, the bunch of tools that you mentioned and you even give a bunch of links at the end, right? I couldn't help but notice that most of these are developed outside of India, right? Because there's, I guess, more of a culture of having technology work in the domain of medicine outside of India. Like, what do you think is the scope right now in the current scenario of like institutions, like say, AMs to work with tech companies or like people who are interested in tech to get things like these developed out? Because like my friend mentioned, we don't do this on a day-to-day basis. We'd love to, but we absolutely don't know how to contribute other than listening to this and wanting to contribute, so. That's one of the primary areas why, you know, my earlier talk with Chennai Pai was with the same idea and the organizers came in touch with me with the same idea that we need to have a diversified speaker profile as well as, you know, something that should reach on a larger level. From just from Chennai Pai meet, they wanted me to actually come here and deliver this talk. This was the whole purpose of it that most of this is happening outside. And in fact, the work given by Cicero Moray is what I said. It's almost like a single developer working on it with few others contributing to it now and then. The majority of the work is done by that person and there are other areas as well, like in forensic areas, which that person has done it using the same technology. So, like, you know, finding out what could have been the structure of the face of a mummy or of something that was excavated, say, 2,000 years ago or something like that. You know, these are amazing areas where it could be worked and I believe that with the human resource available here and with the kind of talent available here, it is just a matter of, you know, putting your idea forth and being heard and seen, you know. Like, I just stumbled upon that project. And otherwise, even I would have not known. Even when you actually search for ORTOG on Blender, it actually goes to a, you know, correction in Google, it leads to somewhere else. See, this is the actual state that happens. So, the idea here is that, you know, you got to be seen, you got to be heard. Sometimes you have to knock the door many times. It doesn't happen once. You just don't pass on. Sometimes you have to knock the door until it opens. Sometimes the only way to get in is to break in with a breakthrough. And that's the whole point of me coming here and, you know, giving this talk. Hi, sir. So, love the talk actually, pretty good. Sir, just one thing, like as per your experience, like you have been like in the craniofacial surgery part for like so many years. And you have had so many insights for like each and everything, for each and every nerve, you know, where exactly to cut and how to proceed. So, like being programmers, like we people like are good with code and dealing with the memory part and all the stuff. We are nowhere near your knowledge level. So, how can we put in that kind of insights? Like, let's suppose if you're cutting out the way you showed, like the bridge cap that you showed. So, how will we put in our effort? See, one part like I told you about was the whole workflow. It's not just about one step or something. But when you're talking about how as programmers you could do, it's about closely working with the team. There is no other way. It's going to be a two-way area. It's not going to be one way. It's not going to be where I give you a list of specifications and you write a code and you give it back. It has to go through, you know, possibly if it's in a very nascent scenario or nascent level. How do surgeons learn how to operate? You know, at some point, we were also, you know, people who have never held a knife to cut somebody's skin or something. Whereas right now we could do that. How were you when you started writing code? You know, there was a stage, okay, if I forget about it, you and I were at a stage where we needed somebody to catch your hand and write A, B or E, whatever it is. But right now we are at a stage where, you know, you can say, if you tell me what to code, I can code it, right? So it's about actually working in tandem with each other. So this will, that's the reason, you know, I said about for research collaborations, we are always open, especially institutes like AIMS are always open for research collaborations. So if there is a use case scenario where it could be found and if you would like to contribute, you would be very well open for it. You know, it has to go through a process, that's a different case, but there is potential for it. And regarding these nerves anatomy, everything you said, there is already data available. It's all about actually taking the data and using it for yourself. And as data specialists, I'm sure this is not a big job for them. I mean, there is variation at every point. That's a different story. But you do have norms, you know, like this is where we expect something and this is the variation of it. And we just take that into account, we do the plan. And we are not asking for something that's going to be totally automatic. You want something where a manual input is possible. So that correction can be done by the surgeon. You know, when the surgeon sees that they'll be able to identify the nerve or the vessel going in there, they know that a cut cannot be given there. So when you have two or three options available, you can manipulate that cut to a place where the damage is minimum. So that's about it. That's the whole idea of similative surgery that you do not try it on the patient directly. You know, you try to do it earlier so that you anticipate problems during surgery and avoid it and minimize, you know, the damage done. Thank you and all. And yes, okay. Hi, my name is Raveena More. I'm from Pune. I did my master's thesis in ultrasound images. wherein we were trying to find out the blood velocity at different parts of the body. And I must say that it is very exciting to work on medical topics because you see that your use, you know, your thing is going to help someone in the real world, you know, unlike data science, where you know, you're just going to help someone make money. I don't totally agree with it, but yes. That's what is being done. I would agree. But then data changes lives. That's true because we work on data as well. Yeah, yeah, but one challenge that I faced was that out of my 12 months of M-tech thesis, I had to spend eight months to just get the data. And that did not leave me a lot of time to actually work on the data, try out more things. And that's why I did not get enough data. So if there is like, you know, this problem that hospitals are, you know, they are a bit... Hesitant. Hesitant to give data also, even if they agree after going through the whole procedure, the doctors or the resident doctors are, you know, not wanting to collect the data samples or to help you with it, you have to do a lot of rounds. So these are, I think, some of the big roadblocks in, you know, doing such kind of research in India. So what do you think is the, you know, way out for this? Because I approached like some of the best hospitals in Pune, you know, though I won't take names, but it was difficult to get the data from them. Data sharing is a problem and it comes into healthcare because of course, you know, when you talk ethically, it comes into, you know, the ethics of it, you know, the privacy issues and everything. But, you know, these workflows are not well defined, but of course there are, like you said, ultimately we're able to get into some hospital and do some work. And when you talked about eight months required in gathering the data, that is one of the reasons when you see about the research protocols that happen even from medical point of view, it goes through a pre-start vetting, you know, it is vetted at different levels before it actually goes into. And my suggestions is that, instead of actually planning for something at a higher level in the first place. Control large amounts of data. Hello? Yeah, there was just a crosstalk. It's okay. So the only challenge that you would have faced, I agree to it and it's unfortunate I would say that sometimes this happens. Even sometimes we face struggles when you have to go through some kind of a specific research project to complete it. But I would suggest something has to be started on a smaller level. And then, you know, once you've got a, you know, you get used to the workflow of it, you get used to the requirements and what actually happens and the, you know, behind the scenes of that particular scenario. And then it is easier to actually get the data. And with respect to actually working with hospitals, I would say there are roadblocks, but the only way is to, like I said again, you know, keep knocking until the door opens or find out the people, there are people who are interested, I would say, you know, see, I wouldn't say I'm the only one who's interested in, you know, collaborating or working on these areas. There are lots of people, but we do not know where to go and we do not know where to go and ask. And sometimes, you know, our workflow or our work schedule is such that you do not have the time to actually go and search for people. So we just say, fine, after some time your interest starts dying out because you get busy with other things. So this is a flame that we need to keep kindling from both the sites, you know? That's the only way I would suggest. And research projects always have a challenge. And many times, some research projects do not end up with a, you know, workable result. You just end up with a thing stating that this should not be tried or this is not a feasible one. And I would say that that is a very valuable data because at least somebody else would not waste the resources doing the same thing again. So yeah, thanks for the talk and I would actually thank the organizers for calling me here, you know, in this gathering to share my thoughts and views. Thank you and goodbye again.