Lecture: Eyes on the Future: Transforming Diabetic Retinopathy Screening Globally through Innovative Technologies

During this live webinar, we will discuss various models of diabetic retinopathy (DR) screenings in countries, while highlighting and understanding the local health systems, opportunities, and barriers. As ideally every patient in the world with diabetes should be screened regularly for DR, we will examine how high-risk patients can be triaged for screening. Lastly, we will explore predictive models and how prevention programs can be initiated at the patient level. (Level: All)

Lecturer: Dr. Sobha Sivaprasad, Ophthalmologist, Moorfields Eye Hospital, United Kingdom of Great Britain and Northern Island


Good afternoon, good morning to everyone. I’m professor Sobha from Moorfields Eye Hospital London. Today, my topic is on eyes on the future. Transforming diabetic retinopathy screening globally through innovative technologies. I will try to cover four sections. The first is the importance of diabetic retinopathy screening. Section 2 is on diabetic retinopathy screening models. And section three is key diabetic retinopathy screening programs and their impact, success stories, and challenges. Lastly, I will go through emerging technologies that transform diabetic retinopathy screening for the future. So the first question we all would want to know is why do we need to screen for diabetic retinopathy? So diabetic retinopathy is an asymptomatic condition that is the patients do not have any symptoms due to this in the beginning of development of the disease. Therefore, it’s crucial that we identify those that are progressing to vision threatening diabetic retinopathy. And that is why we need to screen these patients for retinopathy because the patients will not be able to identify early retinopathy. By the time they identify retinopathy, it will be too late. They may have developed advanced stages of diabetic retinopathy. So I have a question for you. What region in the world has the highest increase in the rate of diabetes? Please answer. Europe, south Asia, China, Middle East, South America? Okay. So 8 percent got it right. It’s China. So a lot of attendees did mark south Asia and it’s sort of true because after China, it’s India and Pakistan. The countries with the increased prevalence of diabetes. So we spoke about diabetes just now. And now I want to stress on diabetic retinopathy. So there are 103 million people living with diabetic retinopathy in the world today. And this is going to rise to 160 million in about two decade’s time. Our next, why is diabetic retinopathy important? There are a few points I need to raise with you. First of all diabetic retinopathy is the most common cause of visual impairment in working age group. So it is irreversible if it’s discovered late. But it is preventable if identified early. And that is why diabetic retinopathy screening is very important. So you identify the disease and treat them early. You look at the global prevalence, one in five people with diabetes have diabetic retinopathy. 6 percent of people have vision threatening diabetic retinopathy. And 4 percent have clinically significant macular edema. The public health burden is extremely high. If you now look at an individual’s quality of life as a result of any of the sight-threatening complication, they do not parallel the visual acuity. People’s quality of vision are often worse off than recorded visual acuity. This will affect their work. They end up with disability during their working life. So it’s, we need to improve their awareness of this condition, therefore they look after themselves. So self-monitor. Go to the hospital and get their diabetic retinopathy checked and treated if needed. Now, this is at personal level. But at global level, this is also a huge public health problem or a global health problem affecting several countries. And you can see the figure on the right on the economic burden. It runs in the billions. It’s important that we all help reduce the social and economic burden of diabetic retinopathy. I have a next question for you. What is the most common sight threatening complication of diabetic retinopathy? Is it macular edema, proliferative diabetic retinopathy. Severe non-proliferative diabetic retinopathy. Vitreous hemorrhage or traction retinal detachment. 45 percent got it right. It’s macular edema. Macular edema is now the most common of the sight-threatening complications. So I’ll go through each step. Who to screen. When to screen. How to screen. So I have yet another question for you. Who should be screened regularly for diabetic retinopathy? Is it all people with type two diabetes from the point of diagnosis. Or is it people with type two diabetes aged 40 years or older. Or all people with type one diabetes from diagnosis. Or only people with other diabetes complications? Very good. All people with type two diabetes from the point of diagnosis should be sent to a screening program. Let’s move on. Who to screen. You already know who to screen. If you go down to the screening schedule. For type one diabetes, you can start screening five years after diagnosis. For type two diabetes, because of the late diagnosis of diabetes in most adults, they need to have their diabetic retinopathy screening at the time of diagnosis of diabetes. Diabetic retinopathy has several stages. Patients start with no diabetes changes in the retina and then they go on the mild non-proliferative diabetic retinopathy to moderate diabetic retinopathy to severe non-proliferative diabetic retinopathy. And proliferative diabetic retinopathy. So they are marked in red, severe and diabetic retinopathy because these are the sight-threatening complications that have to be screened and sent to hospital for treatment. What you see below is macular edema. A patient can have no macular edema or noncentral macular edema or central involved diabetic macular edema. So we have various diabetic retinopathy classification systems. And it started off with the early treatment diabetic retinopathy severity scale. It was done as part of a clinical trial and you can see how complicated that levels of severity is from level ten to level 81 to 85. That was because at that time, they saw so many features or lesions on the retina but had to take them systemically to find out the lesions that cause the increasing severity. With time, we now use the international clinical diabetic retinopathy severity level, on the left-hand side you can see no apparent retinopathy. Mild, moderate. And severe NPDR to proliferative diabetic retinopathy. And below you can see no exudates and no apparent thickening within 1DD from fovea or the presence of exudates in that area. This is the most commonly done classification throughout the world. In the UK, for screening purposes, we use the national screening committee or the NSC screening. You can see the same — graded as R0, R1, R2, R3 and M0 and M1. These are done in this way so if a grader identifies anything from R2 or R3, they can be referred. Prediabetic retinopathy may not be vision threatening but we need an ophthalmologist to see that patient and that is why R2 and R3 are referable retinopathy. Below you see M0 and M1. M1 is the referable retinopathy. We have diabetic retinopathy screening in Scotland. Here it’s broken down further to make it easy for referable. These are the types of classifications used. If you choose one, you need to stick with that so people know exactly what you mean when you refer a case. Coming to various models of diabetic retinopathy screening, a patient can be screened anywhere. But there are three major types. There’s the hospital-based screening models. These are very important in countries and areas where there are no systemic diabetic retinopathy screening services available. So these patients are mostly seen by a diabetes physician, general practitioner or in a clinic or tertiary care. And it can be an ophthalmologist seeing a patient in secondary care. So these patients needs to be screened because they would not have had their diabetic retinopathy screening anywhere else. The most wanted or robust way of screening is community-based screening model so all patients with diabetes have their diabetic retinopathy screening. But for this you need investment, you need a systemic screening program. These can be done either as a stand alone system or at pharmacy, optical shops. Or you can find your own area where this can be done in the rural areas, especially when there are no ophthalmologists around, et cetera. And the newest model is teleophthalmology screening model. This is where you have a combination of both community care as well as hospital based. So patients get screened in the community and images are sent to a central area for grading and this is usually placed in the hospital or it can be placed in any area and the images are graded there. Takes a few days for the patients to know about it. And then if there is referral, the letter will state the patient has to be referred for an opinion or treatment. Now, that’s been developed for over 20 years now. It’s running really well in several regions around the world. But very recently, we have this artificial intelligence programs that can be integrated into the tele ophthalmology screening models and this can make screening less time consuming and I will continue on that later. I want to cover some guidelines for you. We have the American diabetic association guidelines, the NHS UK guidelines and Australian and New Zealand guidelines and several others. Rather than going through these in detail, I think if you see a patient in a clinic and you have no cameras, no eye facility, the least you can do is to take the — record their vision. If their vision is good, they can be advised to go to a diabetic retinopathy screening program to have their eyes examined for retinopathy. As I said earlier, good vision doesn’t mean they do not have diabetic retinopathy. So they still have to have their diabetic retinopathy screening done. If they have poor vision, 4, 12 or worse they should be referred to an ophthalmologist. (6/12) of course they need to rule out diabetic retinopathy screening as well. There are cases where they do only screening without vision testing. In that case, any patient with NPDR or DME or PDR should be sent urgently to an ophthalmologist for monitoring and treatment. Now, our guidelines usually talk only about color photographs being taken to identify retinopathy. But with time, now, we have also approached doing OCT in the screening programs so that you reduce the number of patients required in secondary care. Despite all the development we have with diabetic retinopathy screening, there are significant number of challenges. I will first go through the challenges that are seen more in low, middle-income countries. I explained already there is lack of knowledge of diabetes related complications in the eye. Secondly, there is lack of knowledge or awareness about the need for eye examination in people with diabetes because they wait for their vision to drop, which is too late. Next, there is a lack of knowledge about the availability of eye clinics near a person’s living area. There is also cost and skill limitations in implementing retinal photography service. Of course, there is a global shortage of healthcare worse force for screening and treatment of diabetic retinopathy. We need cost to set it up, so infrastructure and resources are further problem. And then a lack of legislation and policy support. If your government supports this screening, it’s easier to implement it. And financial stability. The patients need to be systemically screened regularly. And for that you need financial sustainability. If we tried to solve all these problems of world financial and workforce and absence of awareness of the disease, in upper middle and high-income countries where some of these are not an issue, we still face problems. There are poor provider services. There is negative self-perception about a person’s eye condition. There is a lack of communications between physician and patients. Then we ideally should be dilating the pupils to take clear photographs and mydriasis or the dilation of the pupil, called mydriasis is a complication for patients and they do not turn up. There are socioeconomic inequities. And there is always a reluctance to change behavior among all of us. One of the key challenges in diabetic retinopathy screening is many countries do have laws that do not permit optometrists to dilate their pupils without on ophthalmologist supervision. In these countries we cannot do diabetic retinopathy screening outside of an ophthalmology center where there is an ophthalmologist. That’s a huge barrier. A solution has arisen with non-mydriasis scanners. We can take images with these cameras without mydriasis. It’s noninvasive. Patients find it more acceptable. It’s portable and easy to use and cost effective. However, there are limitations. In low and middle income countries there are a lot of patients with media opacities and we can’t do this in eyes with small pupils. And non-mydriasis cameras can have a limited field of view. If you’re in an area in the world where there is no diabetic retinopathy screening program, I would rather have a non-mydriasis fundus camera image of my eyes rather than nothing. So non-mydriasis cameras are hugely important to ensure global coverage of diabetic retinopathy screening. Now, I will go through some diabetic retinopathy screening programs and their impact. I will mention their successes and challenges within the programs. I will first take UK national program for diabetic retinopathy screening. It’s a country-wide population-based screening. A very successful program in many ways. This was based on the St. Vincent deck layer ration in 1989. A reduction in diabetes related blindness should be reduced by one-third in Europe. It’s a huge undertaking but UK has achieved it. You can see here the systemic diabetic retinopathy screening program started in 2003. All patients diagnosed with diabetes should be sent to a diabetic retinopathy screening program. Once that was implemented throughout England and Wales, you can see the population coverage increased over time and at present it has 81.5 percent coverage in 2020. And because of this, we found that we could urgently refer patients with vision threatening diabetic retinopathy to a treatment center and therefore diabetic retinopathy is now no longer the leading cause of blindness in the working age group in England and Wales. You can see how dramatically a ten-year investment in the national screening program led to a significant impact in terms of reducing the risk of blindness in these patients. In England, we do two-field, 45-degree field of view and use mydriasis for all our patients before we capture images. The images are graded by trained technicians and arbitrated by ophthalmologists. It’s a well-established program at present. And there are — as you know, the number of people with diabetes are constantly increasing all over the world. These type of systemic diabetic retinopathy screening programs are difficult to sustain because the number of people with diabetes are increasing and they need annual screening done to them. So one way to decrease the workload or decrease the burden of diabetic retinopathy screening is to extend the screening in people with no diabetic retinopathy. And the second plan that has now been implemented is the use of optical coherence tomography or OCT in the second-line digital surveillance clinic so only the screen positive diabetic MAC cue low edema need to be referred to the hospital. Despite all the plans we have and all the successes we have, we still have challenges. We do need to introduce automated analysis for grading. So if we can say that an AI system is robust enough to differentiate a patient with no retinopathy or the presence of retinopathy, you already decrease the workload by approximately 70 percent. So it’s very important that we try to implement automatic analysis for grading. Secondly, at present we still use mydriasis for all our patients. And a better or more effective way out would be to use new camera technologies for screening. For example, staged mydriasis. That is, you try to take your image using new cameras, non-mydriatic cameras and use mydriasis only if the images are not gradable. So that’s called staged mydriasis. And the field of view is approximately the same as that in standard cameras or desktop cameras. What we want to do is to have an ungradable image rate of less than 10 percent. And apply a very cost-effective service. I will cover Scottish national DR screening program. Here the country is smaller. There are only 200,000 patients with diabetes but the coverage is extremely good. What they do here is that they do staged mydriasis. And they also do only a single field 45-degree field of vision imaging. And they have incorporated OCT into the diabetic retinopathy screening program since 3 or 4 years ago. So the advantage here is, of course, the workforce is decreased. You don’t need to increase the workforce as the number of people with diabetes increases because you have automated grading for a considerable number of patients. The third is a SEED study for Singapore. Here again, the images are captured at various sites. Clinics, hospitals, op toll trysts and mobile clinics and the images are transferred to the reading center where graders grade the images. And then they are transmitted to the clinicians with the result. Now with the integration of AI into this system, it’s — the results are usually transmitted within an hour to the clinicians and their patients. So clearly remarkable improvement. Now, Los Angeles, of course the US is a high-income country but there is difference in healthcare throughout the US. So this is an example of a Los Angeles tele retinal DR screening program. You can see that approximately 20,000 patients were graded and they get 11 percent of patients referred to ophthalmology. And the rest are — so let me get this right. 20,000 patients are screened of which 20 percent have DR and another 10 percent have other conditions. So here they do three standard field protocols and they dilate only if the image quality is not acceptable. There are various ways of DR screening models but end of the day, the aim of all these are to identify patients with vision-threating diabetic retinopathy and macular edema early rather than wait for the visual acuity to drop. Here you can see the Los Angeles one needed a lot of work. But it did lead to 89.2 percent reduction in waiting times. And what are the issues with that program? Implementation, of course, when you implement the DR screening program in any country, you’ll be faced with several hurdles. One of them being education and stakeholder support. So you need buy-in from the local health authority. Of course, you have to address skepticism. Why do we need this. You need to increase access of care for the patients. And you also, if you’re a general practitioner, you need to collaborate with an eye care professional to send the patients off for treatment if they have referable retinopathy. So these are the barriers. There are also barriers in terms of decreasing workforce and implementation of tele medicine. And the need for standardization. But the major impacts were the fact that we have an adjustment period and it’s also impacted by new EMR system. If your electronic medical record system changes, we have a significant impact on the system, on the screening program as a whole. There may be concurrent changes in health policy. Today you may have a policy that states that all DR patients need only be screened in a certain area or something like that. So that will derail whatever program you were trying to build. And for this particular study, they had only historical controls. So there were incomplete assessments of the treatment axis as well. This is a huge problem for any screening program in that a screening program may work very efficiently in identifying patients with vision threatening diabetic retinopathy. But there’s a further step in that these patients that need to be informed, educated and told to go to on ophthalmology center for treatment. That communication often breaks down and the patients don’t show up despite being identified as potentially have been diabetic retinopathy. There is a need for further assessment of these centers and linking up communications between them. Now, coming to a low middle income country, this is Bangladesh. Here there is no standardized DR screening program. But for the last 7 years they have tried two stage DR screening in a few hospitals in Bangladesh. You can see they have used disk and macular focused mydriatic images. And they were done by nurses and paramedics and sent for reading. It is feasible for a low, middle-income country. Smart India is one study that I led in India. It’s a multicenter cross sectional screening study. We had UK funded a large grant for us to establish DR screening program in India and assess their processes, estimate the prevalence, and come up with guidelines on how to run these services. So we involved ten Indian districts and one union territory. We screened 42,000 patients and used non-mydriatic camera. We took one disc and one macula centered image and a tele ophthalmology program and four tertiary centers. The results, 78 percent of patients had a gradable image. That’s a very important point. We went house to house surveying patients. Mainly to prove that there is no primary eye care system in India. And therefore could any form of community care help identify diabetic retinopathy and any other retinal conditions. Approximately 3 million people aged 40 or older had vision threatening diabetic retinopathy. And a higher prevalence in those with non-diabetes. We found that of the 42,000 participants that we screened, 7,000 of them had already had diabetes of which 4,350 had known diabetes. The rest were undiagnosed. So a considerable number of patients don’t even know they have diabetes. Another part of that project is called the — India project. It was the project that we started in India. In Kerala. Here is a particular example where the government was engaging. They wanted a diabetic retinopathy screening program embedded in their new electronic medical records that they had in family healthcare centers. You can see how a government buy in helps us to very quickly establish a DR screening program within primary care. So the government of Kerala had an NCD registry like many other states in India. It also had 16 family health centers in Kerala. Of these, there were several staff in there that were trained to use the remedial camera which is a non-mydriatic camera. And this, a positive step here was the remedial manufacturers themselves came in, worked with the government and trained the staff and therefore were able to go through the feasibility quite well with them. So we enrolled 4500 patients and the images were transferred by PACS to the tertiary center. Where there were retinal specialists and they graded the images together with the graders that they developed. They were optometrists. And then positive patients, if they were positive or referral patients, they were then informed through the family health center that referred them. And then the patients were referred for treatment. Either laser or anti-VEGF treatment. This is called a complex intervention because you can see despite government buy in, there are several steps we have to take to ensure that the pathway works. So what is the impact of doing such an exercise? We realized that one in three people with diabetes in this district had diabetic retinopathy. And 8 percent had to be referred to a secondary or tertiary care. 38 percent were between the age group of 41 and 60 years. Clearly stating the public health problem this condition has. And 99 percent of the patients were not aware of diabetic retinopathy and their consequences: All this costed only 400 Indian rupees for screening and treatment. So clearly, it’s a very cost effective way of identifying people at risk of blindness. At present, the impact is that the government found it cost effective and have implemented it throughout Kerala. It’s a huge impact but everyone needs to initiate a feasibility exercise like we did to ensure that it can then be implemented in the larger area. For example, in district region, et cetera. So these are the public issues that you can read about, the complex interventions on the Kerala protocol. And then the Kerala policy is in addition to diabetic retinopathy screening, they also screened for all other end-organ damage due to diabetes. Like they had a foot examination, a nephropathy check, a cardiovascular check. And therefore, it became significantly valuable for patients to identify all complications of diabetes within primary care. And we have, of course, had to do diabetic retinopathy screening guidelines as well as a result of our efforts. One big question we have as I mentioned in my first slide, people with diabetes are only increasing by the day. Whether we bring in workforce or bring in AI, it is going to be a problem to screen these patients every year. So one way out is to identify the high risk groups. And people with diabetes to fast track them for diabetic retinopathy screening. And to control their risk factors so they delay the development of diabetic retinopathy. This is more a preventive approach. Instead of screening everyone at every visit, we do this prediction modeling and send patients for diabetic retinopathy screening in a triage. Everybody with type two diabetes needs screening but because we don’t have the workforce, is there any way of making it more cost efficient. This predictive model will tell you which patients are at very high risk of having vision-threatening diabetic retinopathy in the next three years or even at present. And then that group needs to be referred out. The second groups, of course, will tell us what the risk is like and whether they can control their blood pressure, blood sugar, and cholesterol levels to decrease the rate of progression of diabetic retinopathy to vision-threatening diabetic retinopathy. Predictive modeling are really good and cost effective. You have that for cardiovascular disease. You have that for nephropathy. So this is one such model that you could use for diabetic retinopathy. But I stress, all people with diabetic retinopathy, with diabetes type two should be screened at diagnosis and regularly after that. It’s only the timing that will be reduced as a result of this predictive modeling. So the summary of key challenges in diabetic retinopathy screening programs are mydriasis. The challenges with policies that do not allow mydriasis. And most importantly is the quality of the images that are captured. Of course, you need cloud systems if you do tele medicine. So we need the internet. And we need to try to do automated instant report and referral system so they can educate the patient while the patient is still with you. And of course, you need resources for image grading and training and education. I will now go onto the last point and that is emerging technologies that can transform diabetic retinopathy screening. So artificial intelligence as I mentioned earlier can identify people with diabetic retinopathy. If we have an AI tool that tells us this patient has no diabetic retinopathy, very accurately, then that patient can stay in the screening program without a physician or a grader having to grade the images. Because it told you instantly that you have no DR. If artificial intelligence identifies a patient with diabetic retinopathy, although there are AI tools that do different referable retinopathy from non-referable retinopathy, it’s still wise that we — as accurate as possible. All these patients need to be seen for treatment or prevention. What’s the benefit of AI? It’s faster analysis compared to manual methods. It’s scalable to larger populations and with higher accuracy in detecting early signs of diabetic retinopathy. However, AI is not without challenges either. Because there is regulatory uncertainty. Financial challenges. There is uneven financing and reimbursement schemes around the world. Data privacy is an important issue. And of course, we don’t like change. Doctors and patients may be resistant to this new technology. We do — some centers don’t have the infrastructure to run an AI-integrated screening program. Or they may not have enough IT infrastructure. And may not have the adequate governance to run an AI-integrated project. There are significant challenges that one needs to cross to integrate AI into DR screening programs: But that does not mean it cannot be used. So this is one example that was done in Singapore. No this is one that was done in Thailand. You can see that in Thailand, they had mydriasis and several cameras involved and integrated this into the integrated health record system. This is a Scottish system where the AI system is already integrated and you can see how well the — you can see the accuracy of the screening programs. AI integrated screening program. This is the Thailand one: They used non-mydriatic cameras. Here they came up with a few points that are really important for low and middle-income countries. That is quality of the images matter. You can apply or deploy a deep-mind learning system into healthcare but if the images are not good enough, we will end up with a lot of un-gradable and inaccurate imaging. So this is what I already mentioned. When you implement, you need a streamlined referral system to an ophthalmology clinic. You need an electronic recording system. And un-gradable images and slow internet are significant barriers. And with time, if AI is integrated into cameras, it will be a huge achievement because the patients can be given their results then and there. This is the ORNATE India project where we looked at whether we can integrate deep learning grading system in non-mydriatic cameras. The one point I need to stress is that deep learning DR grading system was developed from images in another country taken on mydriatic pupils. Now, when you apply a deep learning DR grading system from mydriatic retinal camera to a non-mydriatic screening, we will face challenges. Because the AI system is not used to seeing poor quality images or un-gradable images as much as they would if they were using a stand-alone mydriatic camera. So ORNATE-India showed us there is a way of integrating AI algorithm into a camera and it’s extremely accurate to identify referable DR from fundus images captured by minimally trained field workers in a home environment. So it can be done. There is a need for automation of referable DR systems so it’s much easier to implement this. DR models can learn to classify gradeability of hand held 2-field non-mydriatic retinal images. And for that, I suggest the deep learning model is built on the non-mydriatic cameras, on the images that were captured by the non-mydriatic camera. So remedial has got one automated online DR, artificial intelligence system in their smartphone fundus camera. The images are taken as macular and decentered. The AI image is from the DR algorithm that is offline. This is integrated on the smartphone fundus camera. You take a picture and click the button and you’ll be told whether this patient has referable retina or not. This has been tried in several countries that you can see in the right upper part of this. All over the world. And the reference standard is of course using certified graders and using the diabetic retinopathy screening program. One advantage of this camera is that they have automated offline model to screen pre-disease areas. Not only diabetic retinopathy but also glaucoma. And AMD. So you can imagine the impact this can do to people with diabetes. It can soon spread to just ocular screening for anyone. This is yet another study that looked at head-to-head comparison of 7 automated artificial intelligence diabetic retinopathy screening systems in the U.S., in Seattle and Atlanta. This is very interesting. It clearly shows that the performance at Atlanta is better than that in Seattle. One point that we need to know is why. The reason is because the performance is as good as the quality of the images taken. This is why it’s extremely important that the quality of retinal images are good and then AI is applied. So with this, I have some take-home messages. All persons with diabetes need to be aware of diabetic retinopathy. Patient education is key to success. All care providers should be aware of diabetic retinopathy and make sure the patient is advised about screening. And all sight threatening complications should be refer today a retina specialist for treatment. There are various screening models. I could go on for the rest of the day. But I think the forward looking solution is an AI integration into an existing non-mydriatic retinal camera as a first line for screening for diabetic retinopathy as we really need a clear solution for insufficient workforce and increasing incidence of diabetes. Thank you very much. >> Do diabetic retinopathy severity levels relate to HBA1C? Yes, it’s the strongest risk factor for diabetic retinopathy and its progression. Next question, is it possible to regain vision for a patient with diabetic retinopathy? Yes, if it is identified early. It depends on the stage of the retinopathy. If it’s severe, it is unlikely for us to get on top of things. That is why early diagnosis is important. Do you know of any AI solutions in diabetic retinopathy? I just showed an example of remedial camera having integrated AI. There are several AI systems. There was a slide where I compared several AI screening programs and there are a lot around the world. What’s the prognosis and possibility of VA improvement after management of DR complications? As I said earlier, it will improve if you identify the patients early and treat them early but sometimes if it comes too late, it will be difficult to improve visual acuity. But we mostly stabilize vision. Telemedicine versus AI. It’s a complementary approach. We can do the images and send to human graders or integrate AI and tele medicine and make sure part of the system is graded by AI. Shall a general ophthalmologist give anti-VEGF for a patient with clinically significant diabetic macular edema. As long as this ophthalmologist is trained how to read OCTs, how to implement retreatment criteria. And how to add laser, how to switch patients and how to stop treatment. So clearly there are a lot of things that the general ophthalmologist should train and know about. Of course, the general ophthalmologist can do the service especially where there are no retinal surgeons around the world. Sorry for a patient who can’t afford referral. That is correct but there are government patients or hospitals where there are no charges and these patients can be referred to. There is always a need for quality assurance and clinical governance which is currently undertaken by the PHE in England. Who is responsible for the QA and governance and screening program or the AI company? That is a difficult country. This is why this is being delayed in the UK, because it’s not enough if AI is of sufficient quality and very accurate in identifying sight-threatening diabetic retinopathy. It’s also the governance around it there. Is a data policy, governance. There are several steps to be taken. That is what England is trying to do. How much is the cost of the smartphone attached tool. I wouldn’t know. Can optometrists advise for DR screening in India? Anybody can advise for DR screening. It’s extremely important that all of us take that one point away today. That any of us can tell a person with diabetes whether you are at home or in a store or a museum or anywhere, if a person tells you I’m diabetic, I think it is our role to tell them they need to have their eyes screened for diabetic retinopathy. That’s the only way we can get coverage. How to do screening for ischemic retinopathy. That is something we haven’t started doing yet. It would take a significant amount of time but if we can do screening according to the current classification, we will be in effect, screening people with peripheral ischemic retinopathy. Do you think a screening program with AI should be a priority for lower income countries? It’s easy for me to say yes but as a question was raised earlier, we also need to ensure the governance is correct, that the images go to the right person. You’re grading the right person’s eye. Reporting is correct. Data integrity is maintained. Data protection is ensured. There are lots of other issues but if there is nothing to offer, then yes, AI cameras will provide a good solution where there are no ophthalmologists around to do the screening. I will give you an example. In the UK, you have a very good diabetic retinopathy screening program. What we’re trying to do is to see whether integrating AI will improve the efficiency of such a system. It is a different question if you are asking, can we use an AI integrated camera in a place where there are no ophthalmologists anywhere close by. There are people with diabetes with no primary care. And there is no advice given to these patients. Of course, then, at least we should integrate AI into their cameras and get them identified early and treated early. Sent somewhere for treatment. So you can understand it also depends on the scenario where the AI is being integrated. Any course for becoming a grader of retinal images. There are several courses around the world and some are free. Do including OCT increase the DR screening efficiency? Yes, it does. Definitely. It reduces all the unnecessary referral for non-center involving diabetic retinopathy. It’s good to have a patient education when you’re having screening to tell them they have non-center macular edema and need to be reviewed for screening in 4 months but they don’t need to go to an ophthalmologist to get their eyes treated. Can Remidio camera be installed in any camera? No. It has to be in their camera. As far as I’m aware. There are lots of repeats of questions. Just for reference, I’m wondering how long is the wait for newly diagnosed diabetes patients to be screened in the UK? They need to be seen and screened in three months and most patients do have their screening done. It says as soon as possible though. Doctors working in remote areas where OCT or B scan is not available. How do you estimate the visual prognosis for cataract surgery in patients of suspected DR. Are you asking about cataract surgery in a patient with suspected DR where there is no OCT. Okay. You have to take the risk. Do the cataract surgery and then treat the patient for the diabetic retinopathy. Because you can then see the fundus better. Kindly specify any possible training modules for using smartphone AI camera. I think you can contact Remidio and they may be able to advise you on that. AI with smartphone for screening, is it affordable in Africa. I’m sure it’s affordable and I mentioned the governance around it but it’s definitely affordable and available. What is the level of diabetes retinopathy in Pakistan? It’s between 22 percent and 30 percent. Almost everywhere. Okay. I think I have answered most of the questions. Or nearly all of the questions. What’s the average time it takes far type two diabetes patient to develop DR? If the diabetes is under control. Less than 6 millimole for a period of five years. Difficult to understand. Because despite diabetes being under control, because you developed diabetes, there is micro vascular — in the retina and diabetic retinopathy can still develop. But it’s the strongest risk factors of hypoglycemia. So thank you everyone.

Last Updated: May 29, 2024

4 thoughts on “Lecture: Eyes on the Future: Transforming Diabetic Retinopathy Screening Globally through Innovative Technologies”

  1. Good presentation but we still have much challenges in low income countries
    No equipments for screening
    Distance to tertiary Hospitals where services can be obtained couple with high cost


Leave a Comment