 Okay. I think it's eight o'clock. Well, welcome to Moran Eye Center, Grand Rounds. I'm Kathleen Diggory, and I'm so pleased today to be able to introduce Dr. Dan Malia, who is a professor and head of the Visual Neuroscience Group at the Singapore Eye Research Institute. He's also a professor of the Duke NUS Graduate Medical School in Singapore. He's a professor of ophthalmology at Angers University in France, and an honorary professor at the Copenhagen University in Denmark. He did his medical school training in Sao Péter in Paris, France, where he also completed a fellowship in neuro-optimology, specializing in ophthalmology, and he has his PhD in neuroscience as well. This lecture is so exciting, because I think it really tells us where neuro-optimology could be going in the future using artificial intelligence to help us understand and detect optic disc abnormalities. You know that we've been doing some artificial intelligence with retinal disorders, but this optic nerve and optic disc study is really excited. He created this large international group called the BANSE, the Brain and Optic Nerve Study with Artificial Intelligence. And if you haven't seen his paper that came out in the New England Journal of Medicine, I highly recommend it. Dan, we are so grateful for you to come today and visit with us and do Grand Rounds, and tell us all about this exciting new work that's taking place. Thank you so much, Kathleen. It's really a high pleasure and privilege to be here with you. As you said, I'm just very sorry I couldn't make it in person. That would have been a wonderful opportunity, but I highly encourage you, if you happen to come to Asia, come and visit me. This is Singapore and the background. Singapore is a fantastic city and country, very vibrant, a lot of energy, lots of research, and tomorrow's technologies. It's an absolute phenomenal environment. These are my financial disclosures. And this is Singapore National Eye Center. It's a very big center, and we are doing a lot of research on the right side. It's the CERI, Singapore Eye Research Institution. And historically, we've been doing a lot of artificial intelligence for diabetic retinopathy. But before going into that, I would like to thank my closest collaborators and partners in crime, Valerie Buse, Nancy Newman from the United States. And locally, this is Raymond Najar, Liu Yong, and Tianyin Wang, who are my collaborators. This is my visual neuroscience group at CERI, a very dynamic and I would say very diverse group. We're having engineers, neuroscientists, biologists, lots of coordinators. And this is the group interested in artificial intelligence in ophthalmology. Quite a big group. And ophthalmologists are probably the minority of this group. It's a very interesting setup. So let me start with a case. This is a patient we saw last week or 10 days ago. And now Kathleen very kindly introduced Bonsai. Bonsai is the brain and optic nerve study with artificial intelligence. So basically, we're testing pretty much every image, every patient who comes in our clinic has pictures taken of his fundus and we're testing, if possible, in real time, the algorithm, trying to understand what does the algorithm say compared to what do we say. And it's interesting because for this patient who had a recent onset of headache, the way the algorithm displays the results is the following. It's saying, right, we're doing another thing, which I think it's probably interesting to know, they were also trying to take pictures with a portable camera to try to do this in real time to make things easier. So this is how the results are displaying. The machine provides a probability for this being normal, papillodema, or others. And in this case, it was quite interesting because the machine did not say papillodema. It says it's certainly not normal, very, very low probability in each eye. Papillodema also very, very low, more so in the left eye than in the right eye. But the machine said this is others. This is not papillodema. This is not normal. And interestingly enough, the resident forgot to take the blood pressure and this was just malignant hypertension. So this tells us a bit of what the machine can do. And I think it can be easy to misdiagnose this fundus with papillodema by non-expert ophthalmologists. That's why our question was, is there a place for artificial intelligence in your ophthalmology for optic disc abnormality detection, especially when we're dealing with optic discs that are not overtly small like in this case, which nevertheless could be a percent two or three. But however, there could be a life-threatening condition that is underlying. So let me tell you a story that made a lot of noise in the UK. You probably heard about the story. This is a kid who went to an optometry shop in the UK. He had an absolutely normal examination and he was discharged after having glasses. But this became big news a couple of months later in the BBC because the child died from undiagnosed hydrocephalus. And when looking back at the pictures, there was papillodema that was misdiagnosed by the optometrist. And the optometrist ended up being convicted. That was revised later. But you can imagine how much emotion is around that situation. And I don't have to tell you that this was followed immediately by a very high peak of prescriptions of MRI, absolutely unnecessary MRIs. So this is the context in which I think that fondness photographs could have a place. For instance, the emergency room and this is Valerie Beuse and Nancy Newman and the DR team who did extensive studies about this. And I'm just citing one of their papers in neurology. They're showing that there's a variety of optic disc abnormalities in the emergency room. So patients coming in with neurological symptoms or visual loss or high blood pressure, 12% of them, which is quite considerable have a normal disc and 3% have disc swelling. So our question was, could AI in Singapore, could AI identify abnormal optic discs for non-specialists? Well, the problem was, I was told, don't even think about it. We have 500,000 images of diabetic retinopathy in Singapore. Of course, the number of papillodema images were considerably lower. But we thought it might be a solution to this. And this is collect worldwide. Go and do a big study with all the friends, all the neurothemologies, which I think are a very special population. And we created the bonsai consortium, brain and optic nerve study with AI. We have now more than 30,000 images in 30 centers, 20 countries with an incredible variety of images coming from all the places from India, from Asia, from Russia, from the United States, even from Africa. And of course, ethnicity is a big issue in order to be sure that the machine is performing equally well with all sorts of populations. So these are a few of the collaborators. I apologize, not all of them are here, but you certainly can recognize Valerie and Nancy, probably Neil Miller, John Chen, Patrick Uyman, and many, many others from the United States. And just to remind you, before going to the results, how to conduct an AI study, we're training a system. So the system is exposed to images of a variety of images with a very strong ground truth diagnosis. And the machine is literally learning from these images. And then once we feel comfortable that this has achieved a reasonable performance, the system is tested in an external dataset, an image that have never been seen by the first system of the training. So training followed by testing. And we created a CNN, a convolutional neural network, first for segmentation of the disks and then for their classification. I don't want to go into details, but we have expert, wonderful engineers doing all the AI work. And the output of our system was to look at images if the disks were normal, papillodema or other optic disk abnormalities. I don't want to go too much into the inclusion criteria, but what is interesting is that we had not only atomic diagnosis, but also general diagnosis, like for instance, venous anus and boses, etc. And this is a retrospective collection of more than 15,000 images, multiple cameras. We tried to mitigate the risk of bias, of course. And they were testing in five totally independent countries. What was very nice, I think, in our study is that we had a very strong reference standard. We didn't look at images saying, oh, this is papillodema. That's not the way we did it. We looked at every single patient at brain imaging, at CSF opening pressures, at everything that we could do to be sure that our reference diagnosis is strong, that that was papillodema and not pseudo papillodema, for instance, which I think made our study quite strong. And these are some technical details. I don't think they are very interesting. What is probably more interesting though is that we could have almost 10,000 normal, 2,500 papillodemas. And I think that's quite considerable. I don't think any human has seen 2,500 images of papillodema. Our computer did and remembers every single detail. And other update, these abnormalities, quite considerable numbers to come to 15,000. And this is the way our API shows the results. So we're just uploading the image. And on the right side of the slide, you can see the probability that the machine comes up with. So in this case, it's quite obvious that papillodema, the probability is 95.6. And we always encourage the clinician to make his own choice. This is just an assistant. This system does not take into account clinical science like headache or tinnitus or other things that could be quite interesting. But in this case, for instance, obviously the system says 95%. This is papillodema, very confident. The system makes the diagnosis on one eye only, not at the patient level, but at the eye level. And if this is quite obvious in this case, I think it's less obvious in this case where the picture is a bit not very clear. There is probably a lower free sense of verity papillodema. Yet look at the system, it's still 95%. Why? Because the system has seen so many images. I think a human would be less confident than that. And I'll come back about it later. And this is another image, an example of optic disc atrophy. And the system feels very comfortable, 100%. I'm always a bit worried when the system says 100%. But in this case, I have no worries to agree with it. 100%, this was optic disc atrophy. Now the results, they are expressed as the performance as area under the curve. I'm sure you are familiar. 05 would be a straight line, which is due to chance. And the higher, the better, of course. And these are the heat maps. We're trying to understand what is the system looking at when it feels so comfortable about calling this papillodema, for instance. And these are the external data sets. We did five data sets in Bangkok and the United States, in Copenhagen, Tehran, and Germany. And look at the curves. For identifying papillodema, the system, the area under curve is 096, which is really very good. And what's interesting is that we have several curves for each of the five centers, as you can see here on the right side, and they are quite superposable, I think. Now, what's interesting, I understand that this is an audience of neurologists predominantly. So my neurology partners, they said, you know, papillodema is great, but we're probably even more interested to know that this are really normal. And the area under curve for that is 098. So the negative predictive value is very high, meaning that if the system says this is normal, the chances are really very high for an optic disc to be normal. And again, I'm talking about neurologists who are not very familiar to do ophthalmoscopy or not, don't feel very comfortable about doing it, yet they highly need that. And as Kathleen has kind of alluded, we have published this two years ago in the New England Journal of Medicine. I think not so much because the paper had that intrinsic high value. It's because this is for the first time that a system could connect the eye with the brain, making inferences about papillodema and intracranial problems. Now, one of the questions we asked ourselves is the machine making errors? And while it's interesting, the answer is yes, obviously, but very often we had to agree with the machine. Let's say, for instance, a provider would give us an image and the machine would say something different than what the provider was giving us. And in front of this discrepancy, sometimes we had to go back to the provider saying, you know, are you sure? Because we pretty much agree more with the machine. And this is an email that I received in 2019. I think it's quite cool. And the provider is saying, oh, sorry for my mistake. This was a coding error. Your machine is better than me. Your machine is better than me. Is this our future? Should we be worried about this? Are the machines going to be more performant than we are? We can discuss about that at the end. A second step, you probably recognize Nancy Newman and Valerie Beuse. They came to Singapore in an incredible challenge to fight against the computer. And yet, although this was, I have to say, I'm sorry if this is recorded, but this was quite tense. Yet they kept their smile and they were very fair players. And they had this competition against the computer looking at 800 images. And the time of Valerie and Nancy ranked between 74 to 61 minutes, the machine 25 seconds, obviously we're talking about 800 images. And the accuracy was quite similar, I have to say. But this is interesting. That means that our system, at least on this data set, was at least as good as two expert neuroautomologists who had absolutely no clinical information. Subsequently, we published a different story trying to classify the papillodema severity into two classes. We simplified the free-send. And that worked quite well. We published that in neurology. And what we're doing now, we're trying to compare the performance of the system with non-automologists. And obviously, this horizontal line is the performance of the system. So it's around 15% errors. And look, everybody else, neurologists, ED internists, everybody does higher numbers of errors compared to the machine. Now, we're also doing a perspective study that we reported at the North American Neuroathlon Society in February. I don't think this has much interest. It's just to say that even studied prospectively in consecutive patients, our system performs fairly well. And its performance drops only marginally. And let me finish with one case. Again, a real live case that we tested. We didn't see the patient, but we tested those images received from some providers. And again, in this situation, the system was that this is not normal, papillodema. Not normal, but it is probably others. It's not normal. It's not papillodema. And then we're developing a new system. We discussed about this with Kathleen. If this is an ischemic optic neuropathy, because it was a acute onset in both eyes, the next interesting step would be to try to compare the probability of being AA-ION and GCA, or NA-ION. And guess what? The probability of being AA-ION, I'm sorry, the display, was very high. And this is exactly what happened. This was AA-ION related to giant cell arthritis that was proven later. So the future, I don't know about the future. I think handheld cameras might have a role. This is the way we're looking at every single patient now, because having a proof of what we're seeing is important. And also applying bonsai on an image in real time, it's an important thing for us. Could one day, those cameras have an embedded AA-ION that could provide an instant diagnosis? That would be absolutely wonderful. Could other machines benefit from AA-ION for new or off applications? We don't know yet. But I think the future holds for AA-ION, new or off, probably improve diagnosis. We don't know if we can predict one day outcomes. And why not? We can always dream some treatment recommendations. Thank you very much. Thank you so much, Dan. Thank you so much for that really interesting talk. So we have time for questions. I asked him to leave some time, because this is such a hot topic in neuroophthalmology. I'd like to start out by asking you about one of the enigmas for us in neuroophthalmology and also pediatric ophthalmology, is trying to discern Barry Druzen from papillodema. What did your AA-ION system, what could it do with Barry Druzen? That's an excellent question. I have another talk, which is 20 minutes about that only. So it's a very hot topic, absolutely. So we did look specifically at that. So we had something like 3,000 papillodema images and something like 1,000 optic dysdrusin. And we, for this purpose, we choose to make an alliance with the members of the Druzen study with Stefan Harmon, with Fiona Costello, with Claire, with Valerie, with many others. And guess what? The performance is super high. It depends what we're looking at. So looking at severe papillodema, comparing severe papillodema, let's say 3,500 against visible optic dysdrusin, the performance is close to 100%. And I mean, any new resident can make that difference. But then we're looking at different subclasses, the performance drops. I mean, unsurprisingly, looking at lower 3,000 grade versus let's say a buried optic dysdrusin, the performance drops. Yet it is above area under the curve. I think we're above 90%. We're writing up a manuscript about that. And you are absolutely right. I think the right question is in the hands of pediatric ophthalmologists who really see those buried druzen. If it's a calcified druzen, I don't think it's a big deal. Although sometimes, I mean, druzen and papillodema can coexist. Did I answer your question? Yes, that's, that was just great. We're also very excited about your study about AION and arteritic AION. Because as you know, that is such a critical difference. And AION often also is confused with optic neuritis and having those two. So have you got a study going on comparing optic neuritis and AION? Right, that's very interesting. So the answer is no. The answer is no for that. For two reasons, which are geographic reasons to start with is for some bizarre reason, AION is extremely rare in Asia and optic neuritis just as much. So we probably will have a very skewed population here looking at NMOs and MOG optic, which are more common than the common optic neuritis. But we also feel that the critical, I mean, how often do you really hesitate between AION and optic neuritis? I think this is quite a small population age 40, a bit atypical. So if you want to do a very clean study, we would have to train only on these populations and the numbers will be extremely small. We thought about that very, very carefully. And I think we might do it, but it's a bit early for now just because we don't have access to the right populations. However, as what you said, AION versus NAION, we are looking at that very intensely. And the results are promising. I don't have definite results yet. But again, we have a number of limitations because we're looking only at the acute onset, although patients can come two, three days later and we wouldn't know how to, I mean, we're excluding those for now. Thank you. Dr. Brad Katz has asked, at what point will the machine be able to use clinical information to help narrow down the diagnosis or able to ask questions about the patient's presentation, which is an interesting thing to take the image and then apply it to the situation. Right. Hey, Brad, nice to hear from you. Yeah, so that's a very interesting critical question. So I still think we're doing the first baby steps and we did not take into account clinical data because we are not sure what sort of clinical data are we going to get. We probably will have to do this with only highly selected centers. We have centers from literally patients from 30 centers. Looking at retrospective data, we are very worried that we might have dirty or inaccurate data. So how do we know if the patient really had headache or a pain in eye movements? Missing data is a big problem in AI. So that's why we prefer to limit ourselves to images, but that is absolutely the future. I fully agree. Dr. Olson, you got your hand up. Yeah, incredible presentation and obviously AI is going to be powerfully important in regards to the future and what we're all going to deal with, but a lot of data sets, sadly, as you mentioned, are dirty and where you have a garbage in, garbage out problem, then AI essentially doesn't have a very good ability to differentiate. So I think that ability to really make sure that you've got very clean sets. I look at, for instance, some of the data mining in association with billing records. And in many instances, and the one that I know that is probably one of the dirtiest would be glaucoma, that probably a third to a half of the people who have glaucoma are being diagnosed as normals right now. And that of those that are diagnosed with glaucoma, many probably don't have glaucoma. So where you've got those kind of data sets that you're doing and they're not clean, then AI obviously is going to have a much, much lower predictive value. So we need to appreciate the fact that this is a, not just grab huge data sets that everything will be resolved and we'll be able to answer any tough question. There's a lot of work going in to make sure that the data sets that are there are accurate and clean. And then you can get these kind of results and they're going to be extremely helpful and particularly helpful to see that important diagnoses are not missed. I 100% agree with you. On the other hand, our approach has its downsides as well, being too clean makes us going away from the real world. And that's the next step. How much does a wonderful study, super clean study, will apply then to a real world? And that's what we're trying to do now in a prospective study. So far, it looks pretty good, but that is the ultimate challenge to translate those methods into real world to make real use of them. Sure. No, good point. Other questions from anybody else? In the New York Times today, it said, AI, when is it going to, when is it going to start having a return on investment? So I would say to you, Dan, when do you think we'll be using the schema in our emergency rooms and also, especially with papillodema in neurology offices or primary care offices? That's a very good question. So I think to people like you and me, it's quite obvious that the ability to discriminate between optic disk and papillodema should be very useful, but there is no cost-effectiveness study that has been done. We have to show that, you know, using this system, if it works, will save us money in terms of less MRIs, unnecessary MRIs, or targeting better our patients. So there is still, I'm afraid, a long way to show that this is cost-effective. That's great. Any other questions from anybody? You can unmute yourself and or feel free to put it in the chat. Hi, Kathleen. This is Jeff Petty. Dan, thank you so much for this talk. It's extraordinary. It's really exciting. I have about two questions. First one, I'll ask, you know, we get our pre-test probability, you know, on our clinical examination, and that of course then goes to guide whether or not we order MRIs, et cetera. And in this particular case where, you know, we're given an actual number now, it seems as if the, you know, think about type 1 errors, type 2 errors, not wanting to miss the brain tumor, it seems like a lot of the clinical decision-making of simply to order an MRI or not to, you know, do a lumbar puncture or not, that could be fairly algorithmic at this point. You mentioned about the emergency departments. Do you envision at some point this being utilized as, you know, your screen, you just automatically go through a physician-less process until we get that information at the end? Absolutely. I think that would be the goal thing. So as you know, you have to deal with sensitivity and specificity. So for now, and those things have to be tailored for the users. So I would say, as you mentioned, for the emergency department, you don't want to miss. So probably you're going to go for a very high sensitivity and just raise the bar for the higher sensitivity and the risk is that it will have probably less specificity and there will be some unnecessary findings, you know, falsely diagnosed papillodema. So those things have to be tailored according to the department where you want to use them. But ultimately, I fully agree. I mean, I have so many collaborators and emails from the world saying, you know, we do not have access to an ophthalmologist at two o'clock in the morning and a patient with heavy headache. We only have access to one machine. Can you just license the algorithm to us and we'll use it? The answer is no, because there have to be approvals, all sorts of regulatory issues. But I think that there is a big demand worldwide. Probably not so much in the US, you know, in high academic centers, but in more peripheral centers, I would say so. Well, I would say even in our academic center, you know, saving our neuro-ophthalmologists from having to leave their clinics to enter into our clinics to tell us if it's papillodema or not, I think they would welcome that even at our center. I agree, fully agree. Just one other question then. So, you know, use a supervised model, you know, where the inputs were clearly identified, very rigorous, very impressive. Question I have is, were there any particular categories that started to flush out where the algorithm, you know, was not as accurate? I mean, any kind of subsets of categories where it showed a little bit less confidence, or was it able to pretty much just be accurate throughout the broad cross-section? Excellent question. So, as I said, we really want to have superbly clean data. And this is not the real world. So let's say, as you say, diagnosis with less confidence, or let's say optic disc swelling, papillodema, with some atrophy, you know, older or chronic papillodema, the performance drops immediately because we taught the machine on purpose only with fresh papillodema. So other categories were mixed diagnosis. We had it, for instance, a few patients would combine drusen and papillodema. The machine was all over the place. Didn't know how to handle that. That sort of things. Great. Dr. Warner has her hands up. Judith Warner. Good morning, and thank you, Dan. That's a great topic. Good morning, Judith. I love it. And I feel like we're sort of working peripherally to try and get some of our giant cell and AION patients to you, although I don't know. But the comment is really more of a comment, less of a question. The comment that I had is, because I don't know if my ophthalmology colleagues who are on the line here know that we had a handheld non-madriatic camera for use in the emergency room. And Sean Cullen has actually been rolling that out to a variety of other places, like the People's Clinic up in Park City and down at Navajo. But how do you see the non-madriatic camera playing a role in an urgent diagnosis in the emergency room? I know that Valerie and her group did a study of that at Emory and found it to be very eye-opening as to how often diagnoses were missed. Right. Absolutely. So you're spot on. So in that study, this photo ED, she found 13% of optic disc abnormalities. So there has been a lot of controversy about portable cameras because the periphery is not very well seen. But I think we're lucky because we are looking precisely at the optic disc at the posterior pole. So what we're doing is we're doing a segmentation. So we're just extracting literally the optic disc. We don't care too much. That's why our providers are sending us saying, you know, the one 30, 45 degrees, we're fine with anything as long as we can extract the optic disc image. So what we're doing now, we're doing a pilot study. We trained, Bonsai was trained on medriatic images on desktop, wonderful Rolls-Royce conditions. But now we're trying to apply that algorithm to its poor brother, which would be non-medriatic images with a handheld camera. And so far, it looks quite okay. But if you guys have that camera, and if it's still a possibility to collaborate, I think it would be a very, very cool study to do, mainly to test our primary bonsai algorithm on non-medriatic images with a handheld camera. I think that's where the future is going to have to be, because not everybody has the fancy cameras that we have the luxury of having an ophthalmology. Dan, unfortunately, I think our camera got dropped and is currently back in in its home country getting fixed, but hopefully we'll have it back soon. Dan, we want to thank you so much for joining us today. This was really extraordinary. I'm so excited to have had you here. You can stay for the rest of our grand rounds, if you wish, but I know you have family obligations as well. So next, it's my great pleasure to introduce Dr. Eric Kasky, who is our current neuroophthalmology fellow. Dr. Kasky has been really outstanding, amazing multitasker, helping us all in neuroophthalmology. He's going to be heading out to the Wheaton Eye Clinic in Illinois, going back to the Midwest, where his roots are. But today, he's going to talk to us about his photophobia curriculum project, and we're excited to have you here. Dan, thank you. Eric, thank you so much for giving us this presentation. You have to agree. Thank you for the introduction. And Dan, that was fantastic. Thanks so much for coming. So my research this year has been about photophobia and this has been presented at Neuroophthalmology Society meeting this year. And I'll kind of dig into that, and as well as the second project that we worked on, this curriculum that you may have heard of recently. I know financial disclosures, so the group here at Utah has been focusing somewhat on pediatric photophobia. Dr. Buchanan and the group here recently published this retrospective review of 36 pediatric patients that presented with the primary complaint of photophobia, and found that a large majority or 70% left the clinic without a specific diagnosis for their symptom. And I think most of us are familiar that pediatric photophobia is a difficult thing to work up and make a diagnosis on being both uncommon and potentially a sign of something pretty serious. So that's something that we wanted to look further into this year. So we were hoping to better understand why this was happening. We're all pretty competent eye providers here at RAN. So we wanted to delve a little more into this, see if we're making progress over the years, if patients are still leaving undiagnosed, and use a larger patient population as well to see what the presumed causes of these patients' photophobia are. And we wanted to hopefully identify some knowledge gaps that could help us to educate ourselves and future ophthalmologists to better handle this condition. So we ran a retrospective chart review of all pediatric visits with photophobia as a diagnosis from 2016 to 2021 at the University of Utah facilities. We had 304 visits identified and 125 of those were unique patients. So those were the ones that were analyzed. For the demographics, it was a pretty even split amongst the age groups. And we wanted to delve into some of the associations as well. It was very even with male and female. Migraine was a common associated condition as well as try to get injury in dry eye. And some of the things that we commonly think of in ophthalmology with autoimmune disease and blepharospasm were present, but not as common as the others. So right into our results here, the similar to other studies that have looked into this, the most common cause is idiopathic, unknown, and about a third of the patients. Better than the previous review, but still the highest percentage. Migraine was very common, especially in the older age groups. And otherwise, there was a widespread and many different things that could cause photophobia, including dry eye, head injury, blepharitis, and other abnormalities, aniridia, dry eye, and many others, including some more serious things. There are two pediatric alcomas, two inherited retinal diseases, and three intracranial masses. So that would be about 4.6% of the patient population that we studied. And treatment is directed towards symptom management and underlying cause. Tinted lenses were recommended in a little less than half of patients. And observation was the only thing in about 21%, as frequently the eye exam is normal. So that was common, especially in the young patients. Focure surface disease and headache management were commonly addressed. And there were some surgical interventions, and that's where most of the other treatments come in. I want to focus in a little bit on the children less than five years old, as those most difficult patients to examine and diagnose. And this is where a lot of the unknowns came in, greater than half of the patients less than five didn't have an identified cause for their apparent photophobia. Migraine, of course, was much less common. And then there was quite a spread of other things that were associated. So our conclusions were that about a third of the patients in the study still don't have a cause of their photophobia. The second third is headache related migraines, especially, as the main cause. And then the last third is some other identifiable cause. And that includes the high morbidity disorders, like congenital glaucoma, inherited retinal dystrophy, and intracranial masses. Our study is limited by its retrospective design. And being a referral center, we may collect the more difficult to diagnose patients or the more high morbidity causes. Also, this was ICD-10 code based for selection of the patients. And as you may know, there's not a specific ICD-10 code for photophobia. It falls under visual disturbance, which so we may miss some patients and collect some others that are not quite exactly what we're looking for. We don't have a good follow-up for our treatment. So we don't know if what we're doing is really working or helping the patients with their symptoms and their quality of life. And that's where we're going to go next. So we're, especially with the unknown diagnoses, reexamine those patients and figure out the natural course for this photophobia, evaluate our treatment if someone's undergone, and if a diagnosis was identified at a later date. So that will be upcoming. But the follow-up for this is this curriculum that was designed to help address this knowledge gap for photophobia, especially in the pediatric population. So along with the team here, we've put together this for the providers here. So as we just heard, the photophobia frequently being undiagnosed, even though we're all pretty competent at our jobs. So there must be something that we're missing to that we can address. So we made a multimedia curriculum for residents and providers for making better care, more prompt care for these patients. Our objective is to make a validated curriculum that can be widely available for use around the world as the education mission from Iran. So hopefully, after completing the curriculum, the participants can explain anatomy and pathophysiology of photophobia to patients as well as their parents, be able to describe and approach to making a diagnosis and knowing the most common causes, and then to discuss photophobia treatment options, of which there are quite a few. So we really wanted to make a special emphasis on a child with otherwise normal alphabetical exam, as those are our most difficult ones to diagnose. And we wanted to make it useful and efficient for our busy lifestyle. So it's a self-paced combination of some reading, a video lecture, and some interview style podcasts. And we're using a pre and post test quiz to help measure that learning has taken place. So we have made 20 pre-test questions based on the didactic material, which includes two articles by the group here and as well as a video lecture that is available on Moran Core currently. And then the didactic podcasts include Dr. Katz and Dr. Degree, myself talking about the approach to photophobia, the appropriate exam and workup and treatment options. And then the participants would take the same 20 quiz questions to give us an idea that we're making some progress, but also get some feedback from the learner to know that we're what we're doing well and what we're not doing so well. So again, we have this curriculum now available for participants to take part in to address this knowledge gap of pediatric photophobia. And once it's validated, this curriculum can potentially kick off a new facet of Moran Core and its educational platform with a symptom-based instruction. So look out for similar symptom-based curriculums in the near future. So to become involved, everyone that is a faculty, resident or fellow is invited to participate. You will be held anonymous and blinded from the investigator. So show up on your feedback reviews at the end of the rotation. So stop by the fourth floor clinical research office during clinic hours. You can ask the group there for the photophobia curriculum, pre-test and its didactic content. And once you complete the curriculum, you'll complete the post-test and you can collect your one hand and dollar Amazon gift card to use at your own perusal. But also you'll advance the education of world, providers worldwide, as well as your own. So hopefully that is enough incentive. We've had some early feedback that it was educational and useful. It takes between two hours and three hours to know how quickly read and how much you fast forward the podcast. But we would certainly welcome more feedback to know how we can improve. I want to thank the whole neurophthalmology team here, especially Dr. Cass and Dr. Degree for their direct participation in the podcast. Dr. Warner and Dr. C, Dr. Crum for feedback on both projects. arena for helping me with the presentations. Chance Arman is a medical student that helped with some of the data collection in our clinical research team for helping to run this. Great. Thank you so much, Eric. That's really wonderful. And I hope everybody will sign up. It's not hard to sign up. Just go to the fourth floor and get the pre-test and sign up. Are there any questions for Dr. Caskey? Dan. Excellent talk. Thank you so much, Eric. Can I have a naive question, please? Do you have any longitudinal data? Could this be a maturation problem in children in the old IPR GC melanapsum story? Yeah, absolutely. So I try not to delve too much into the nuts and bolts of why fortophobia happens because that's a big part of the curriculum. So hopefully they won't give me spoilers. But we don't have any longitudinal component of the study quite yet. We had the previous study, which was similar to mine, limited in its scope. But that is something that we hope to figure out some more as we have this follow-up study with reexamining these patients years later and seeing if is this just something that resolves on its own? Is there something that is diagnosed later? Do retreatments make any difference? So there's still a lot to figure out in the real world of how this fortophobia works. Brad Katz has his hand up. Unmute yourself, Brad. Thanks. Thanks, Eric. And thanks, Dan, for being here this morning. So we just got an approval from the IRB to go back and look at those 30 pediatric patients who left the clinic without a diagnosis, without a treatment plan. And we're going to be calling them up this summer. We've got a couple of medical students are going to be helping us. And we'll bring them back to see if we can figure out what was the cause of their fortophobia. And we'll use that information to help direct the curriculum. And the long-term goal, of course, is to improve our ability to diagnose and treat this condition in kids and also adults eventually. But I want to encourage everybody. This study that Eric is doing, it's a curriculum for ophthalmologists, optometrists, residents, fellows, faculty, everybody can be in the study. You get a $100 Amazon gift card at the end. And if you play the podcast at 2X speed and play my video at 2X speed, you can get through it in under two hours. We sound like chipmunks. But otherwise, you get the information, direct information. I'm going to be very interested to find out how many of those 30 kids have a migraine background, because I think the migraine story is so underdiagnosed in our clinics. And children don't always have headache. And so the photophobia could be part of their migraine, just as babies have colic, which is migraine, and kids have car sickness as a child, which is migraine. And so the photophobia piece of this may end up being a lot of migraine. We'll have to see. I'm very anxious to see the results of this. Any other questions from all of you? Dan? I'm sorry, but it's such a chance to be with you. I don't I don't want to take advantage of it. So about tinted glasses. So are you doing some trials about that? Are you tailoring this to the filtering spectrum? Or how do you deal with those? So Dr. Katz has got a couple studies ongoing. Do you want to take that question there, Brad? Yeah, thanks, Dan. That's a little bit tricky, you know, because I have a conflict of interest there. And so I've really not been even though, you know, we think that FL 41 is very effective for a lot of these conditions, including migraine, blood for spasm, even traumatic brain injury. We're not studying that directly. I am collaborating with a company in Canada called Avalox that's trying to develop the FL 41 technology a little bit further to make it more directed towards intrinsically photosensitive retinal ganglion cells. And they have done a small pilot study in a cohort of patients with episodic migraine. And I hope to publish those data soon. Thank you. Yeah, thanks, Dan. And, you know, FL 41 was developed in England, actually, by a group that studied this back, oh, gee, in the early 80s, late 70s and early 80s, and published it a long time ago. And we, I had come across the article many years ago, and then introduced that to our optical center. So we've been using it. We have studied FL 41, for example, in blood for spasm. We also did a trial in in migraine, using something similar. And so I think there's work to be done in this area. It's a really tricky area to study. As you know, migraine number one is tricky. What's your, you know, what is the outcome? What's going to be the primary outcome? But it's an interesting area that I think has a lot of potential to study. And on the other hand, I feel that in Europe and even more in Asia, the therapeutic options are really low. We don't offer much to those patients. So I would love to hear more. I mean, we can take this offline. Yeah, well, it's a, as you know, it's been a passion for, for me and for us. So it's an interesting topic. Any other questions for Eric? Otherwise, we're going to be giving everybody a few minutes back to their day. I just want to thank everybody for a great presentation on issues that often are somewhat ignored. And in particular, the photophobia, these people just quietly suffer. And I'm proud of the group here, they're really trying to figure out what's going on. The thought that everybody that's in that category has some element of dry eye. I mean, there's this whole thing of the super sensitivity corneal syndrome associated. I mean, we're digging into areas that I think have just been ignored by organized ophthalmology for a long, long time. Yeah, that's for sure. I think the cornea has more clues for us. So I encourage all the cornea people to keep working on this too from that end, because it's the highest density of trigeminal nerve endings in the entire body, which connects right to the trigeminal nuclei in the brainstem, which has a lot to do with migrates. So it's stay tuned. I think there's going to be a lot. I hope there's going to be a lot of work on this. I wish we could do something with AI in this area because it's so cloudy. But until we can have a clearer vision, it's going to be very difficult to study. Yeah, I mean, it's been there in front of us for a long time where we've seen patients who have maybe dry eye, but essentially nothing. And they're absolutely miserable. And we see others with significant dry eye changes you think. And they're saying, no, that doesn't bother me. I'm actually pretty comfortable. My vision's fine. So the thought that we had somehow assumed in all of that there wasn't another answer was clearly naive. So anyway, we're digging into it. And that's good. Yeah, Dr. Warner, unmute yourself, Judith. Judith, you're not often muted. I have to be allowed to unmute. So sorry, I was trying. I was trying. So I think it's important to have a shout out to Kathleen for all the work that she's done with her colleagues around the country and around the world, but also right here with our fellows, that really terrific study on the corneal nerves in patients with photophobia and migraine. It was just really trying to bring migraine down to the microscopic level. And I think one of the most important findings out of this and has been shown before is that the overlap of symptoms of dry eye and signs of dry eye in migraine and in other conditions that are actually very, very poor. So just asking people about their dry eye symptoms does not give you the full answer. And I think that the overlap of symptoms is probably a clue. It's probably what's going to lead us in the right direction. Probably, Kathleen, you have a much more erudite approach to that comment. Well, I think that we just have to keep studying it more and recognize that this trigeminal system is not just in the brain alone and the trigeminal system innervates the eye and the orbit and everything else. And it's going to be susceptible people that are going to have some problem with this because not everybody, as Dr. Olson said, that has dry eyes gets photophobia. It's really interesting. That would be almost an interesting study in itself to do. It would, wouldn't it? Yeah. Well, thank you, Dan. Thank you for joining us. We really loved having you here for Grand Rounds. And thank you all. Eric, you did a superb job as always. Everything you do is very good. And thank you to all of my colleagues who joined us today to hear some updates in neuro-optimology. Thanks so much.