AI-powered Glaucoma Management: Predicting the Optimal Surgical Treatment

AI-powered Glaucoma Management: Predicting the Optimal Surgical Treatment | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article AI-powered Glaucoma Management: Predicting the Optimal Surgical Treatment Guy Mole, Uvais Qidwai, Umair Qidwai, Nathan Kerr, Keith Barton, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6244012/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Introduction: Glaucoma is the leading cause of irreversible blindness worldwide, and due to changing demographics leading to increased prevalence, pressure on ophthalmic services is growing rapidly in many countries. Recently there has been a rapid increase in new surgical techniques to prevent sight loss from glaucoma with the introduction of Minimally Invasive Glaucoma Surgery (MIGS). This is a relatively new set of techniques with increasing evidence regarding efficacy however it is not yet clear which glaucoma surgery is the optimum procedure to perform in different clinical scenarios. Methods: We developed an Adaptive Neuro Fuzzy Inference System (ANFIS) AI model to help surgeons decide which surgical technique would likely have the best outcome for an individual patient depending on core clinical parameters such as vision and intraocular pressure. The model was also able to accurately predict clinical outcomes such as vision, intraocular pressure and number of medications at 1 year. Results: The ANFIS model had a very high degree of accuracy both in predicting clinical parameters such as vision and intraocular pressure 1 year after surgery and in determining the optimum surgery in different clinical scenarios. Discussion: With the increasing array of available MIGS procedures as well as traditional glaucoma surgery, AI could provide a powerful tool to help surgeons decide, in collaboration with their patients, on the optimum procedure. As the training data comes from an international registry, and so represents real world results across different surgeons and surgical centers, this makes it a powerful tool to help surgeons to practice evidence-based medicine whilst harnessing these new techniques to treat patients with glaucoma. Health sciences/Health care/Health services Health sciences/Diseases/Eye diseases/Optic nerve diseases Health sciences/Medical research/Outcomes research Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction The escalating global burden of glaucoma, the leading cause of irreversible blindness(1), necessitates innovation in both its medical management and surgical intervention. Recently Minimally Invasive Glaucoma Surgery (MIGS) has emerged as a promising group of techniques due to their ability to reduce intraocular pressure with a lower risk profile(2) compared to traditional surgical approaches(3) although this does vary depending on the mechanism of action(2). However, despite the emerging evidence base supporting the efficacy of MIGS(4–6) the selection of appropriate candidates and choice of the optimum surgical option remains unclear, in part because few trials directly compare different MIGS against each other(7) or against traditional glaucoma filtering surgery. Artificial Intelligence (AI) has been revolutionizing many industries and holds transformative potential in this domain, offering the potential to harness complex clinical real-world data to enhance decision-making processes. This paper explores the development of an AI model designed specifically for glaucoma surgery, combining the integration of machine learning algorithms with large-scale clinical datasets aiming to improve patient outcomes. We describe the accuracy of the model and how it could be deployed as a predictive tool that can personalize treatment plans, predict surgical outcomes, and ultimately improve the quality of life for patients with glaucoma. By leveraging detailed patient data—including demographic variables, clinical parameters, and surgical histories—this study underscores the significant contribution AI can make to precision medicine in glaucoma surgery but also has implications for other aspects of care in ophthalmology(8, 9). The strategy employed in this study uses the less data-hogging AI technique of a Neuro-Fuzzy model trained on real world data from the International Glaucoma Surgery Registry (IGSR) at IGSR.org. This anonymized registry is used by leading surgeons around the world as a valuable and constantly updating source of real-world data. Ultimately the aim of this work is for the model to act as a clinical decision support tool for surgeons, in collaboration with patients, to help them in choosing the optimum glaucoma surgery in specific clinical situations. As a result, we explore how the positive findings could be used in clinical practice and how the model could be further improved and developed in the future for new applications. Methodology The complete data selection process used for this model is set out in the PRISMA diagram shown in Fig. 1 . Dataset: The data used to train the model is real-world from multiple surgical centers in different countries and which was recorded in the anonymous International Glaucoma Surgery Registry. This is multi-surgeon data which will reduce bias due to greater surgeon familiarity with certain techniques and includes the outcomes outside of a trial setting making it more applicable to routine clinical practice. Model development To effectively develop an AI model for different glaucoma surgeries, a rigorous data selection process was employed to ensure the integrity and relevance of the data used for training the model. The process unfolded in several systematic stages, each crucial for honing the quality and applicability of the dataset: 1. Initial Dataset Compilation : The primary dataset comprised of 1965 patient samples from the IGRS and 372 samples from a previously published study(10), each characterized by more than 75 features. These features included a mix of numeric, categorical, and ordinal data types, providing a comprehensive set of variables for initial analysis. 2. Feature Reduction Based on Clinical Relevance : To streamline the dataset and focus on clinically significant parameters, the number of features was reduced from 75 to 30. This reduction was based on the prevalence and recognized importance of these features in current clinical practice, ensuring that the dataset remained robust yet manageable and clinically pertinent. It also excluded the previous study dataset which had fewer features. 3. Univariate Correlation Analysis : A univariate analysis was conducted to identify the features with the strongest correlations to the outcomes of interest. Only features demonstrating a correlation coefficient greater than 75% were retained. This stringent criterion helped to refine the dataset to 10 highly relevant features, thereby enhancing the potential predictive power of the resulting AI model. 4. Handling Incomplete Data : The final stage of the data preparation involved scrutinizing the dataset for completeness. Samples with missing data were excluded, resulting in a final dataset of 1,725 complete samples, each with 10 selected features. This step was critical to ensure the accuracy and reliability of the AI model's outputs. By adhering to this detailed data selection process, the study ensures that the AI model is built on a foundation of high-quality, relevant data, thus maximizing its effectiveness and applicability in clinical settings for glaucoma surgery. This methodical approach not only enhances model performance but also aids in the interpretability and clinical integration of the AI tool. ANFIS Model The below table shows the categories of data that could either be fed into the model such as age or lens status or the outputs such as predicted intraocular pressure at one year. We used this to create four models the first three of which looked at baseline characteristics of the patient to predict a clinical parameter such as vision, intraocular pressure or number of medications at 1 year whilst the fourth predicted the optimum surgery: Model 1. VA12 Predictor (VA at 12 Months) a. Inputs: Features 1 through 6 and 10. b. Output: Feature 7. Model 2. IOP12 Predictor (IOP at 12 Months) a. Inputs: Features 1 through 6 and 10. b. Output: Feature 8. Model 3. Meds12 Predictor (Meds at 12 Months) a. Inputs: Features 1 through 6 and 10. b. Output: Feature 9. Model 4. Operations Classifier (Best Operation to choose) a. Inputs: Features 1, 2, 6, D(7, 3), D(8, 4), and D(9, 5). {Note: D(x, y) = Feature y – Feature x, and represents the require difference for that category of features} b. Output: Feature 10. For a more detailed description of how the model was designed please see appendix 1. Results The dataset contained 1,725 patients who had undergone glaucoma surgery with an average age of 67.9 years. The most common diagnosis was primary open angle glaucoma and the most performed glaucoma surgery was trabeculectomy. There were however also high numbers of minimally invasive glaucoma surgeries with the nine procedures included in the dataset all having more than 100 cases. Models 1 to 3 were able to take basic demographic and clinical parameters regarding the patient such as age, diagnosis and type of surgery to predict the clinical outcomes at one year of visual acuity, intraocular pressure and number of medications with a high degree of accuracy. Model 4 looked at baseline characteristics such as vision, age, lens status and diagnosis combined with clinical parameters at 1 year to determine the optimum surgery predicted by the model. This was achieved by the model matching the patient clinical characteristics to those in the database who had a successful surgery based on intraocular pressure reduction. Table 2 demonstrates that for all the models there was good correlation between the ANFIS model and the actual data and that this was statistically significant at p < 0.0001 for all models. The results show that the ANFIS model could be used as a clinical decision support tool as it allows the baseline clinical parameters to be inputted to accurately predict outcomes at one year for key factors such as intraocular pressure and number of glaucoma medications. We know from other studies on MIGS that the effect at 1 year is likely to be maintained with published data out to five years(5) although few studies have looked at a longer timeframe meaning long term results are awaited(11). This means that models 1 to 3 will be extremely useful for predicting whether a particular procedure is likely to achieve the desired clinical outcome. Model 4 shows the surgery that is most likely to be successful for lowering intraocular pressure based on patients who have had a good result in the international glaucoma surgery registry. All four of the models could therefore provide valuable information to support glaucoma surgeons, in combination with patients, in different clinical situations which are reviewed in the discussion. The models will be developed into an app to be easily accessible in clinical settings to provide decision support. The outlook of the App is shown in Figs. 2 and 3 with a high-level logic flow diagram to depict the functionality of the system. Discussion There are an increasing number of minimally invasive glaucoma procedures available but despite increasing evidence regarding their efficacy (4–6, 12) it is still unclear exactly where these surgeries fit into the glaucoma treatment pathway(13). In addition, there have been few trials comparing different types of MIGS making it hard for surgeons to know which to offer to patients in different clinical scenarios(7). Likewise, few studies have compared MIGS versus traditional glaucoma operations such as trabeculectomy and so there is significant variation in practice regarding which surgery is performed in different clinical situations. The use of the ANFIS model harnesses real-world data to help predict which surgical technique would be optimum based on individual patient characteristics and we have demonstrated can accurately predict clinical parameters at one year. Having this model available in an easy-to-use format such as an app would therefore be extremely helpful to tailor the surgical plan to the aims of the specific patient. For example, the surgeon in clinic, in conjunction with the patient, could use the model to look at whether each procedure is likely to achieve the target pressure and the likely medication burden. This is important as the goal can vary in different patients with some needing to reach a certain target pressure whilst for others it may be that they are intolerant to or non-compliant with eye drops(14) meaning that the optimum surgery can vary for two patients with the same clinical parameters. The model could therefore help to determine whether the required effect in terms of intraocular pressure or medication burden is likely to be achieved with each procedure and so which would be preferable. In model 4 the ANFIS model selects overall which is likely to be the optimum surgery based on results achieved by surgeons around the world. This decision support tool could therefore be useful for surgeons to use when deciding on a surgery and may lead them to consider other options or increase their confidence that the procedure they planned would be optimum for that specific patient. This is likely to be of particular interest currently as few trials compare MIGS directly(7) and many surgeons do not yet have extensive personal experience of different MIGS procedures. The ANFIS model has the advantage of being trained on real-world data and so can help guide surgeons based on actual outcomes in clinical practice, which may be different to the performance found in clinical trials. This is a powerful way to make evidence-based decisions and is without bias from commercial interests. A further advantage of using data from a live registry is that this involves multiple surgeons from different countries making it less prone to bias from outliers and allows the model to be regularly updated as new data or techniques become available. The ANFIS model is currently being developed into an app and therefore can easily be kept up to date as new data is added to the registry as well as providing an easily accessible route for surgeons to access the decision support. Implementing the ANFIS model into clinical practice does however come with some of the well documented concerns around the use of AI in healthcare such as the potential for bias and the effect on health equity(15). Whilst our study has demonstrated the model to perform well using IGSR.org data it is also unclear whether the model could incorporate other valuable sources of data such as from clinical trials which has the strength of demonstrating surgical outcomes following a defined protocol. In addition, whilst the ANFIS model performed well using data from the IGSR it is unclear whether it would still be accurate using other sources of data such as an individual surgeons’ own outcomes. This may be the most relevant when deciding on which procedure would be most beneficial for a patient particularly in the surgeon has less experience of some techniques than other surgeons entering data into the IGSR. Below are two clinical scenarios fed into the model to demonstrate how it could be used in clinical practice. The example on the left-hand side shows a 59-year-old phakic patient with reduced visual acuity. The suggested surgeries in this situation are both cataract surgery combined with minimally invasive glaucoma surgery as at 1 year this will likely lead to improved vision as well as reduced intraocular pressure and medication burden. The example on the right-hand side shows a pseudophakic patient with secondary glaucoma with very high pressure on four classes of medication. In this situation the model suggests two bleb forming surgeries with the predicted outcome at 1 year of no change in visual acuity but a large reduction in intraocular pressure and medication burden. This model is envisaged to be used as a decision support tool for surgeons hence providing more than one suggestion is helpful to consider different options and provides an additional suggested technique if either is not available in their surgical centre. In conclusion, this is an exciting time for the glaucoma surgeon with the proliferation of new techniques to try and preserve vision and as a result an increasing volume of glaucoma surgeries being performed(16) albeit with a decline in the number trabeculectomies performed in some countries(17). Whilst the emerging trial data has demonstrated the efficacy of MIGS we believe that an AI model trained on real-world data(18) has the potential to be an extremely valuable complimentary tool for surgeons to help select the optimum technique in specific clinical scenarios. The high degree accuracy achieved by the model also demonstrates the power of AI and how the rapid development in technology can be harnessed by the modern glaucoma surgeon to gain the best outcomes for their patients(19). References Jayaram H, Kolko M, Friedman DS, Gazzard G. Glaucoma: now and beyond. Lancet. 2023;402(10414):1788-801. Vinod K, Gedde SJ. Safety profile of minimally invasive glaucoma surgery. Curr Opin Ophthalmol. 2021;32(2):160-8. Xin C, Wang H, Wang N. Minimally Invasive Glaucoma Surgery: What Do We Know? Where Should We Go? Transl Vis Sci Technol. 2020;9(5):15. Ahmed IIK, Fea A, Au L, Ang RE, Harasymowycz P, Jampel HD, et al. A Prospective Randomized Trial Comparing Hydrus and iStent Microinvasive Glaucoma Surgery Implants for Standalone Treatment of Open-Angle Glaucoma: The COMPARE Study. Ophthalmology. 2020;127(1):52-61. Ahmed IIK, De Francesco T, Rhee D, McCabe C, Flowers B, Gazzard G, et al. Long-term Outcomes from the HORIZON Randomized Trial for a Schlemm's Canal Microstent in Combination Cataract and Glaucoma Surgery. Ophthalmology. 2022;129(7):742-51. Gołaszewska K, Konopińska J, Obuchowska I. Evaluation of the Efficacy and Safety of Canaloplasty and iStent Bypass Implantation in Patients with Open-Angle Glaucoma: A Review of the Literature. J Clin Med. 2021;10(21). Gillmann K, Mansouri K. Minimally Invasive Glaucoma Surgery: Where Is the Evidence? Asia Pac J Ophthalmol (Phila). 2020;9(3):203-14. Li Z, Wang L, Wu X, Jiang J, Qiang W, Xie H, et al. Artificial intelligence in ophthalmology: The path to the real-world clinic. Cell Rep Med. 2023;4(7):101095. Jeon S, Liu Y, Li JO, Webster D, Peng L, Ting D. AI papers in ophthalmology made simple. Eye (Lond). 2020;34(11):1947-9. Qidwai U, Sivapalan T, Ratnarajan G. iMIGS: An innovative AI based prediction system for selecting the best patient-specific glaucoma treatment. MethodsX. 2023;10:102209. StatPearls. 2024. Cantor L, Lindfield D, Ghinelli F, Świder AW, Torelli F, Steeds C, et al. Systematic Literature Review of Clinical, Economic, and Humanistic Outcomes Following Minimally Invasive Glaucoma Surgery or Selective Laser Trabeculoplasty for the Treatment of Open-Angle Glaucoma with or Without Cataract Extraction. Clin Ophthalmol. 2023;17:85-101. Chan PPM, Larson MD, Dickerson JE, Mercieca K, Koh VTC, Lim R, et al. Minimally Invasive Glaucoma Surgery: Latest Developments and Future Challenges. Asia Pac J Ophthalmol (Phila). 2023;12(6):537-64. Gawdzik B, Bukowska-Śluz I, Koziol AE, Mazur L. Synthesis and Characterization of Biodegradable Polymers Based on Glucose Derivatives. Materials (Basel). 2022;16(1). Abràmoff MD, Tarver ME, Loyo-Berrios N, Trujillo S, Char D, Obermeyer Z, et al. Considerations for addressing bias in artificial intelligence for health equity. NPJ Digit Med. 2023;6(1):170. Rathi S, Andrews CA, Greenfield DS, Stein JD. Trends in Glaucoma Surgeries Performed by Glaucoma Subspecialists versus Nonsubspecialists on Medicare Beneficiaries from 2008 through 2016. Ophthalmology. 2021;128(1):30-8. Boland MV, Corcoran KJ, Lee AY. Changes in Performance of Glaucoma Surgeries 1994 through 2017 Based on Claims and Payment Data for United States Medicare Beneficiaries. Ophthalmol Glaucoma. 2021;4(5):463-71. Kim HS, Lee S, Kim JH. Real-world Evidence versus Randomized Controlled Trial: Clinical Research Based on Electronic Medical Records. J Korean Med Sci. 2018;33(34):e213. Nair M, Tagare S, Venkatesh R, Odayappan A. Artificial intelligence in glaucoma. Indian J Ophthalmol. 2022;70(5):1868-9. Tables Tables are available in the Supplementary Files section. Additional Declarations There is no conflict of interest Supplementary Files ANFISsupplementarymaterial.docx Appendix ANFIStable1.docx ANFIStable2.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6244012","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":436699008,"identity":"05b632fe-e138-46af-8cb9-394c241d8447","order_by":0,"name":"Guy Mole","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzklEQVRIiWNgGAWjYDCCAxBKhoG9AUgZWBCvhYeBB8QykCBFi0QCiCZCC9/t49ekeXNsePgln1/d8KNAgoG/vTsBrxbJczll0rzb0ngkZ+eU3ewBOkzizNkNeLUYnOFJA2o5zGNwOyftBg9Qi4FELpFa7G+eSbv5hzgt7McgtkiwH7tNlC2SZ3iYLecC/SJxJofttoyBBA9Bv/CdYX944+02Gzn+9uPPbr75A2L04tcCjBEDJh4oA0wSUA4C7A8Yf0AZRKgeBaNgFIyCkQgA2G5FYGFtFzcAAAAASUVORK5CYII=","orcid":"","institution":"Oxford University Hospital NHS Foundation Trust","correspondingAuthor":true,"prefix":"","firstName":"Guy","middleName":"","lastName":"Mole","suffix":""},{"id":436699009,"identity":"04735fd7-1175-4b83-ab26-fb45a51bce0a","order_by":1,"name":"Uvais Qidwai","email":"","orcid":"https://orcid.org/0000-0002-9924-590X","institution":"Qatar University","correspondingAuthor":false,"prefix":"","firstName":"Uvais","middleName":"","lastName":"Qidwai","suffix":""},{"id":436699010,"identity":"800e411e-5708-4162-b0f7-6696668e62ad","order_by":2,"name":"Umair Qidwai","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Umair","middleName":"","lastName":"Qidwai","suffix":""},{"id":436699011,"identity":"3586b6d1-9526-4d21-ae0d-e5e826d2ae5b","order_by":3,"name":"Nathan Kerr","email":"","orcid":"https://orcid.org/0000-0001-7381-285X","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Nathan","middleName":"","lastName":"Kerr","suffix":""},{"id":436699012,"identity":"2aa01ebf-9bbe-4ed9-8c09-b012709a55e3","order_by":4,"name":"Keith Barton","email":"","orcid":"","institution":"Moorfields Eye Hospital, Glaucoma Unit","correspondingAuthor":false,"prefix":"","firstName":"Keith","middleName":"","lastName":"Barton","suffix":""},{"id":436699013,"identity":"dfb3c29b-3bdb-4d84-a038-c94b7bfd460f","order_by":5,"name":"Gokulan Ratnarajan","email":"","orcid":"","institution":"Queen Victoria Hospital NHS Foundation Trust","correspondingAuthor":false,"prefix":"","firstName":"Gokulan","middleName":"","lastName":"Ratnarajan","suffix":""}],"badges":[],"createdAt":"2025-03-17 11:15:54","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6244012/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6244012/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":81699737,"identity":"3cd1d895-61ce-4dae-bc92-3a68e161fcab","added_by":"auto","created_at":"2025-04-30 13:03:54","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":343109,"visible":true,"origin":"","legend":"\u003cp\u003ePRISMA diagram for data selection.\u003c/p\u003e","description":"","filename":"ANFISfigure1.png","url":"https://assets-eu.researchsquare.com/files/rs-6244012/v1/f76686e48b75e6481a6f87c2.png"},{"id":81699716,"identity":"4c106089-32cd-48fa-b5b1-487c82511276","added_by":"auto","created_at":"2025-04-30 13:03:53","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":382806,"visible":true,"origin":"","legend":"\u003cp\u003eDemonstrates how the app can be used in clinic to input the patients baseline data and predict the vision, intraocular pressure and number of medications at 12 months post-op with different operations\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6244012/v1/6dbe6578b005d3afd358933e.png"},{"id":81699718,"identity":"33940992-0795-4daf-a823-d3f5d822b848","added_by":"auto","created_at":"2025-04-30 13:03:54","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":358243,"visible":true,"origin":"","legend":"\u003cp\u003eDemonstrates how the app can be used to suggest an operation based on registry data with a specific patients’ clinical data\u003c/p\u003e","description":"","filename":"ANFISfigure3.png","url":"https://assets-eu.researchsquare.com/files/rs-6244012/v1/c27130404b54c1aaaa51ba32.png"},{"id":81699767,"identity":"9e6c659d-53ea-4b0d-9aeb-3da3ae0f692a","added_by":"auto","created_at":"2025-04-30 13:03:55","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":844320,"visible":true,"origin":"","legend":"\u003cp\u003eDemonstrates two clinical scenarios fed into the model and the suggested surgical approaches as well as the predicted clinical parameters at 1 year\u003c/p\u003e","description":"","filename":"ANFISfigure4.png","url":"https://assets-eu.researchsquare.com/files/rs-6244012/v1/399f0f54c9e7801d465feea8.png"},{"id":83754446,"identity":"5053dd7f-4fe9-499e-9bfe-6dfd7e2ed126","added_by":"auto","created_at":"2025-06-02 08:03:40","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2910251,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6244012/v1/16b970d7-98eb-4079-8ba7-2724fc15a224.pdf"},{"id":81699885,"identity":"1af1bdcb-5521-4981-9063-9a59487948cd","added_by":"auto","created_at":"2025-04-30 13:04:00","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1557911,"visible":true,"origin":"","legend":"Appendix","description":"","filename":"ANFISsupplementarymaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-6244012/v1/6a89ccb2e2eb011c69daa61b.docx"},{"id":81699612,"identity":"8cadfad2-f69e-4a02-bb69-0538fe330864","added_by":"auto","created_at":"2025-04-30 13:03:46","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":45487,"visible":true,"origin":"","legend":"","description":"","filename":"ANFIStable1.docx","url":"https://assets-eu.researchsquare.com/files/rs-6244012/v1/d9a59cf2f53b65a292ce77b4.docx"},{"id":81699860,"identity":"db158039-c3f8-48fb-a28b-fc32f9bab81a","added_by":"auto","created_at":"2025-04-30 13:03:58","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":20987,"visible":true,"origin":"","legend":"","description":"","filename":"ANFIStable2.docx","url":"https://assets-eu.researchsquare.com/files/rs-6244012/v1/1b7342bd47a262627627a832.docx"}],"financialInterests":"There is no conflict of interest","formattedTitle":"AI-powered Glaucoma Management: Predicting the Optimal Surgical Treatment","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe escalating global burden of glaucoma, the leading cause of irreversible blindness(1), necessitates innovation in both its medical management and surgical intervention. Recently Minimally Invasive Glaucoma Surgery (MIGS) has emerged as a promising group of techniques due to their ability to reduce intraocular pressure with a lower risk profile(2) compared to traditional surgical approaches(3) although this does vary depending on the mechanism of action(2). However, despite the emerging evidence base supporting the efficacy of MIGS(4\u0026ndash;6) the selection of appropriate candidates and choice of the optimum surgical option remains unclear, in part because few trials directly compare different MIGS against each other(7) or against traditional glaucoma filtering surgery.\u003c/p\u003e \u003cp\u003eArtificial Intelligence (AI) has been revolutionizing many industries and holds transformative potential in this domain, offering the potential to harness complex clinical real-world data to enhance decision-making processes. This paper explores the development of an AI model designed specifically for glaucoma surgery, combining the integration of machine learning algorithms with large-scale clinical datasets aiming to improve patient outcomes. We describe the accuracy of the model and how it could be deployed as a predictive tool that can personalize treatment plans, predict surgical outcomes, and ultimately improve the quality of life for patients with glaucoma. By leveraging detailed patient data\u0026mdash;including demographic variables, clinical parameters, and surgical histories\u0026mdash;this study underscores the significant contribution AI can make to precision medicine in glaucoma surgery but also has implications for other aspects of care in ophthalmology(8, 9).\u003c/p\u003e \u003cp\u003eThe strategy employed in this study uses the less data-hogging AI technique of a Neuro-Fuzzy model trained on real world data from the International Glaucoma Surgery Registry (IGSR) at IGSR.org. This anonymized registry is used by leading surgeons around the world as a valuable and constantly updating source of real-world data. Ultimately the aim of this work is for the model to act as a clinical decision support tool for surgeons, in collaboration with patients, to help them in choosing the optimum glaucoma surgery in specific clinical situations. As a result, we explore how the positive findings could be used in clinical practice and how the model could be further improved and developed in the future for new applications.\u003c/p\u003e"},{"header":"Methodology","content":"\u003cp\u003eThe complete data selection process used for this model is set out in the PRISMA diagram shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eDataset:\u003c/h2\u003e \u003cp\u003eThe data used to train the model is real-world from multiple surgical centers in different countries and which was recorded in the anonymous International Glaucoma Surgery Registry. This is multi-surgeon data which will reduce bias due to greater surgeon familiarity with certain techniques and includes the outcomes outside of a trial setting making it more applicable to routine clinical practice.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eModel development\u003c/h3\u003e\n\u003cp\u003eTo effectively develop an AI model for different glaucoma surgeries, a rigorous data selection process was employed to ensure the integrity and relevance of the data used for training the model. The process unfolded in several systematic stages, each crucial for honing the quality and applicability of the dataset:\u003c/p\u003e\n\u003cp\u003e\u003cspan\u003e\u003cstrong\u003e1. Initial Dataset Compilation\u003c/strong\u003e: The primary dataset comprised of 1965 patient samples from the IGRS and 372 samples from a previously published study(10), each characterized by more than 75 features. These features included a mix of numeric, categorical, and ordinal data types, providing a comprehensive set of variables for initial analysis.\u003cbr\u003e\u003c/span\u003e\u003cspan\u003e\u003cstrong\u003e2. Feature Reduction Based on Clinical Relevance\u003c/strong\u003e: To streamline the dataset and focus on clinically significant parameters, the number of features was reduced from 75 to 30. This reduction was based on the prevalence and recognized importance of these features in current clinical practice, ensuring that the dataset remained robust yet manageable and clinically pertinent. It also excluded the previous study dataset which had fewer features.\u003cbr\u003e\u003c/span\u003e\u003cspan\u003e\u003cstrong\u003e3. Univariate Correlation Analysis\u003c/strong\u003e: A univariate analysis was conducted to identify the features with the strongest correlations to the outcomes of interest. Only features demonstrating a correlation coefficient greater than 75% were retained. This stringent criterion helped to refine the dataset to 10 highly relevant features, thereby enhancing the potential predictive power of the resulting AI model.\u003cbr\u003e\u003c/span\u003e\u003cspan\u003e\u003cstrong\u003e4. Handling Incomplete Data\u003c/strong\u003e: The final stage of the data preparation involved scrutinizing the dataset for completeness. Samples with missing data were excluded, resulting in a final dataset of 1,725 complete samples, each with 10 selected features. This step was critical to ensure the accuracy and reliability of the AI model\u0026apos;s outputs.\u003cbr\u003e\u003c/span\u003e\u003c/p\u003e\n\u003cp\u003eBy adhering to this detailed data selection process, the study ensures that the AI model is built on a foundation of high-quality, relevant data, thus maximizing its effectiveness and applicability in clinical settings for glaucoma surgery. This methodical approach not only enhances model performance but also aids in the interpretability and clinical integration of the AI tool.\u003c/p\u003e\n\u003ch3\u003eANFIS Model\u003c/h3\u003e\n\u003cp\u003eThe below table shows the categories of data that could either be fed into the model such as age or lens status or the outputs such as predicted intraocular pressure at one year.\u003c/p\u003e\n\u003cp\u003eWe used this to create four models the first three of which looked at baseline characteristics of the patient to predict a clinical parameter such as vision, intraocular pressure or number of medications at 1 year whilst the fourth predicted the optimum surgery:\u003c/p\u003e\n\u003cp\u003eModel 1. VA12 Predictor (VA at 12 Months)\u003c/p\u003e\n\u003cp\u003ea. Inputs: Features 1 through 6 and 10.\u003c/p\u003e\n\u003cp\u003eb. Output: Feature 7.\u003c/p\u003e\n\u003cp\u003eModel 2. IOP12 Predictor (IOP at 12 Months)\u003c/p\u003e\n\u003cp\u003ea. Inputs: Features 1 through 6 and 10.\u003c/p\u003e\n\u003cp\u003eb. Output: Feature 8.\u003c/p\u003e\n\u003cp\u003eModel 3. Meds12 Predictor (Meds at 12 Months)\u003c/p\u003e\n\u003cp\u003ea. Inputs: Features 1 through 6 and 10.\u003c/p\u003e\n\u003cp\u003eb. Output: Feature 9.\u003c/p\u003e\n\u003cp\u003eModel 4. Operations Classifier (Best Operation to choose)\u003c/p\u003e\n\u003cp\u003ea. Inputs: Features 1, 2, 6, D(7, 3), D(8, 4), and D(9, 5). {Note: D(x, y) = Feature y \u0026ndash; Feature x, and represents the require difference for that category of features}\u003c/p\u003e\n\u003cp\u003eb. Output: Feature 10.\u003c/p\u003e\n\u003cp\u003eFor a more detailed description of how the model was designed please see appendix 1.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eThe dataset contained 1,725 patients who had undergone glaucoma surgery with an average age of 67.9 years. The most common diagnosis was primary open angle glaucoma and the most performed glaucoma surgery was trabeculectomy. There were however also high numbers of minimally invasive glaucoma surgeries with the nine procedures included in the dataset all having more than 100 cases.\u003c/p\u003e \u003cp\u003eModels 1 to 3 were able to take basic demographic and clinical parameters regarding the patient such as age, diagnosis and type of surgery to predict the clinical outcomes at one year of visual acuity, intraocular pressure and number of medications with a high degree of accuracy. Model 4 looked at baseline characteristics such as vision, age, lens status and diagnosis combined with clinical parameters at 1 year to determine the optimum surgery predicted by the model. This was achieved by the model matching the patient clinical characteristics to those in the database who had a successful surgery based on intraocular pressure reduction. Table\u0026nbsp;2 demonstrates that for all the models there was good correlation between the ANFIS model and the actual data and that this was statistically significant at p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001 for all models.\u003c/p\u003e \u003cp\u003eThe results show that the ANFIS model could be used as a clinical decision support tool as it allows the baseline clinical parameters to be inputted to accurately predict outcomes at one year for key factors such as intraocular pressure and number of glaucoma medications. We know from other studies on MIGS that the effect at 1 year is likely to be maintained with published data out to five years(5) although few studies have looked at a longer timeframe meaning long term results are awaited(11). This means that models 1 to 3 will be extremely useful for predicting whether a particular procedure is likely to achieve the desired clinical outcome. Model 4 shows the surgery that is most likely to be successful for lowering intraocular pressure based on patients who have had a good result in the international glaucoma surgery registry. All four of the models could therefore provide valuable information to support glaucoma surgeons, in combination with patients, in different clinical situations which are reviewed in the discussion.\u003c/p\u003e \u003cp\u003eThe models will be developed into an app to be easily accessible in clinical settings to provide decision support. The outlook of the App is shown in Figs.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e with a high-level logic flow diagram to depict the functionality of the system.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThere are an increasing number of minimally invasive glaucoma procedures available but despite increasing evidence regarding their efficacy (4\u0026ndash;6, 12) it is still unclear exactly where these surgeries fit into the glaucoma treatment pathway(13). In addition, there have been few trials comparing different types of MIGS making it hard for surgeons to know which to offer to patients in different clinical scenarios(7). Likewise, few studies have compared MIGS versus traditional glaucoma operations such as trabeculectomy and so there is significant variation in practice regarding which surgery is performed in different clinical situations. The use of the ANFIS model harnesses real-world data to help predict which surgical technique would be optimum based on individual patient characteristics and we have demonstrated can accurately predict clinical parameters at one year. Having this model available in an easy-to-use format such as an app would therefore be extremely helpful to tailor the surgical plan to the aims of the specific patient. For example, the surgeon in clinic, in conjunction with the patient, could use the model to look at whether each procedure is likely to achieve the target pressure and the likely medication burden. This is important as the goal can vary in different patients with some needing to reach a certain target pressure whilst for others it may be that they are intolerant to or non-compliant with eye drops(14) meaning that the optimum surgery can vary for two patients with the same clinical parameters. The model could therefore help to determine whether the required effect in terms of intraocular pressure or medication burden is likely to be achieved with each procedure and so which would be preferable.\u003c/p\u003e \u003cp\u003eIn model 4 the ANFIS model selects overall which is likely to be the optimum surgery based on results achieved by surgeons around the world. This decision support tool could therefore be useful for surgeons to use when deciding on a surgery and may lead them to consider other options or increase their confidence that the procedure they planned would be optimum for that specific patient. This is likely to be of particular interest currently as few trials compare MIGS directly(7) and many surgeons do not yet have extensive personal experience of different MIGS procedures.\u003c/p\u003e \u003cp\u003eThe ANFIS model has the advantage of being trained on real-world data and so can help guide surgeons based on actual outcomes in clinical practice, which may be different to the performance found in clinical trials. This is a powerful way to make evidence-based decisions and is without bias from commercial interests. A further advantage of using data from a live registry is that this involves multiple surgeons from different countries making it less prone to bias from outliers and allows the model to be regularly updated as new data or techniques become available. The ANFIS model is currently being developed into an app and therefore can easily be kept up to date as new data is added to the registry as well as providing an easily accessible route for surgeons to access the decision support. Implementing the ANFIS model into clinical practice does however come with some of the well documented concerns around the use of AI in healthcare such as the potential for bias and the effect on health equity(15). Whilst our study has demonstrated the model to perform well using IGSR.org data it is also unclear whether the model could incorporate other valuable sources of data such as from clinical trials which has the strength of demonstrating surgical outcomes following a defined protocol. In addition, whilst the ANFIS model performed well using data from the IGSR it is unclear whether it would still be accurate using other sources of data such as an individual surgeons\u0026rsquo; own outcomes. This may be the most relevant when deciding on which procedure would be most beneficial for a patient particularly in the surgeon has less experience of some techniques than other surgeons entering data into the IGSR.\u003c/p\u003e \u003cp\u003eBelow are two clinical scenarios fed into the model to demonstrate how it could be used in clinical practice. The example on the left-hand side shows a 59-year-old phakic patient with reduced visual acuity. The suggested surgeries in this situation are both cataract surgery combined with minimally invasive glaucoma surgery as at 1 year this will likely lead to improved vision as well as reduced intraocular pressure and medication burden. The example on the right-hand side shows a pseudophakic patient with secondary glaucoma with very high pressure on four classes of medication. In this situation the model suggests two bleb forming surgeries with the predicted outcome at 1 year of no change in visual acuity but a large reduction in intraocular pressure and medication burden. This model is envisaged to be used as a decision support tool for surgeons hence providing more than one suggestion is helpful to consider different options and provides an additional suggested technique if either is not available in their surgical centre.\u003c/p\u003e \u003cp\u003eIn conclusion, this is an exciting time for the glaucoma surgeon with the proliferation of new techniques to try and preserve vision and as a result an increasing volume of glaucoma surgeries being performed(16) albeit with a decline in the number trabeculectomies performed in some countries(17). Whilst the emerging trial data has demonstrated the efficacy of MIGS we believe that an AI model trained on real-world data(18) has the potential to be an extremely valuable complimentary tool for surgeons to help select the optimum technique in specific clinical scenarios. The high degree accuracy achieved by the model also demonstrates the power of AI and how the rapid development in technology can be harnessed by the modern glaucoma surgeon to gain the best outcomes for their patients(19).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eJayaram H, Kolko M, Friedman DS, Gazzard G. Glaucoma: now and beyond. Lancet. 2023;402(10414):1788-801.\u003c/li\u003e\n\u003cli\u003eVinod K, Gedde SJ. Safety profile of minimally invasive glaucoma surgery. Curr Opin Ophthalmol. 2021;32(2):160-8.\u003c/li\u003e\n\u003cli\u003eXin C, Wang H, Wang N. Minimally Invasive Glaucoma Surgery: What Do We Know? Where Should We Go? Transl Vis Sci Technol. 2020;9(5):15.\u003c/li\u003e\n\u003cli\u003eAhmed IIK, Fea A, Au L, Ang RE, Harasymowycz P, Jampel HD, et al. A Prospective Randomized Trial Comparing Hydrus and iStent Microinvasive Glaucoma Surgery Implants for Standalone Treatment of Open-Angle Glaucoma: The COMPARE Study. Ophthalmology. 2020;127(1):52-61.\u003c/li\u003e\n\u003cli\u003eAhmed IIK, De Francesco T, Rhee D, McCabe C, Flowers B, Gazzard G, et al. Long-term Outcomes from the HORIZON Randomized Trial for a Schlemm\u0026apos;s Canal Microstent in Combination Cataract and Glaucoma Surgery. Ophthalmology. 2022;129(7):742-51.\u003c/li\u003e\n\u003cli\u003eGołaszewska K, Konopińska J, Obuchowska I. Evaluation of the Efficacy and Safety of Canaloplasty and iStent Bypass Implantation in Patients with Open-Angle Glaucoma: A Review of the Literature. J Clin Med. 2021;10(21).\u003c/li\u003e\n\u003cli\u003eGillmann K, Mansouri K. Minimally Invasive Glaucoma Surgery: Where Is the Evidence? Asia Pac J Ophthalmol (Phila). 2020;9(3):203-14.\u003c/li\u003e\n\u003cli\u003eLi Z, Wang L, Wu X, Jiang J, Qiang W, Xie H, et al. Artificial intelligence in ophthalmology: The path to the real-world clinic. Cell Rep Med. 2023;4(7):101095.\u003c/li\u003e\n\u003cli\u003eJeon S, Liu Y, Li JO, Webster D, Peng L, Ting D. AI papers in ophthalmology made simple. Eye (Lond). 2020;34(11):1947-9.\u003c/li\u003e\n\u003cli\u003eQidwai U, Sivapalan T, Ratnarajan G. iMIGS: An innovative AI based prediction system for selecting the best patient-specific glaucoma treatment. MethodsX. 2023;10:102209.\u003c/li\u003e\n\u003cli\u003eStatPearls. 2024.\u003c/li\u003e\n\u003cli\u003eCantor L, Lindfield D, Ghinelli F, Świder AW, Torelli F, Steeds C, et al. Systematic Literature Review of Clinical, Economic, and Humanistic Outcomes Following Minimally Invasive Glaucoma Surgery or Selective Laser Trabeculoplasty for the Treatment of Open-Angle Glaucoma with or Without Cataract Extraction. Clin Ophthalmol. 2023;17:85-101.\u003c/li\u003e\n\u003cli\u003eChan PPM, Larson MD, Dickerson JE, Mercieca K, Koh VTC, Lim R, et al. Minimally Invasive Glaucoma Surgery: Latest Developments and Future Challenges. Asia Pac J Ophthalmol (Phila). 2023;12(6):537-64.\u003c/li\u003e\n\u003cli\u003eGawdzik B, Bukowska-Śluz I, Koziol AE, Mazur L. Synthesis and Characterization of Biodegradable Polymers Based on Glucose Derivatives. Materials (Basel). 2022;16(1).\u003c/li\u003e\n\u003cli\u003eAbr\u0026agrave;moff MD, Tarver ME, Loyo-Berrios N, Trujillo S, Char D, Obermeyer Z, et al. Considerations for addressing bias in artificial intelligence for health equity. NPJ Digit Med. 2023;6(1):170.\u003c/li\u003e\n\u003cli\u003eRathi S, Andrews CA, Greenfield DS, Stein JD. Trends in Glaucoma Surgeries Performed by Glaucoma Subspecialists versus Nonsubspecialists on Medicare Beneficiaries from 2008 through 2016. Ophthalmology. 2021;128(1):30-8.\u003c/li\u003e\n\u003cli\u003eBoland MV, Corcoran KJ, Lee AY. Changes in Performance of Glaucoma Surgeries 1994 through 2017 Based on Claims and Payment Data for United States Medicare Beneficiaries. Ophthalmol Glaucoma. 2021;4(5):463-71.\u003c/li\u003e\n\u003cli\u003eKim HS, Lee S, Kim JH. Real-world Evidence versus Randomized Controlled Trial: Clinical Research Based on Electronic Medical Records. J Korean Med Sci. 2018;33(34):e213.\u003c/li\u003e\n\u003cli\u003eNair M, Tagare S, Venkatesh R, Odayappan A. Artificial intelligence in glaucoma. Indian J Ophthalmol. 2022;70(5):1868-9.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTables are available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6244012/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6244012/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eIntroduction: Glaucoma is the leading cause of irreversible blindness worldwide, and due to changing demographics leading to increased prevalence, pressure on ophthalmic services is growing rapidly in many countries. Recently there has been a rapid increase in new surgical techniques to prevent sight loss from glaucoma with the introduction of Minimally Invasive Glaucoma Surgery (MIGS). This is a relatively new set of techniques with increasing evidence regarding efficacy however it is not yet clear which glaucoma surgery is the optimum procedure to perform in different clinical scenarios.\u003c/p\u003e \u003cp\u003eMethods: We developed an Adaptive Neuro Fuzzy Inference System (ANFIS) AI model to help surgeons decide which surgical technique would likely have the best outcome for an individual patient depending on core clinical parameters such as vision and intraocular pressure. The model was also able to accurately predict clinical outcomes such as vision, intraocular pressure and number of medications at 1 year.\u003c/p\u003e \u003cp\u003eResults: The ANFIS model had a very high degree of accuracy both in predicting clinical parameters such as vision and intraocular pressure 1 year after surgery and in determining the optimum surgery in different clinical scenarios.\u003c/p\u003e \u003cp\u003eDiscussion: With the increasing array of available MIGS procedures as well as traditional glaucoma surgery, AI could provide a powerful tool to help surgeons decide, in collaboration with their patients, on the optimum procedure. As the training data comes from an international registry, and so represents real world results across different surgeons and surgical centers, this makes it a powerful tool to help surgeons to practice evidence-based medicine whilst harnessing these new techniques to treat patients with glaucoma.\u003c/p\u003e","manuscriptTitle":"AI-powered Glaucoma Management: Predicting the Optimal Surgical Treatment","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-30 12:10:34","doi":"10.21203/rs.3.rs-6244012/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"c572d97c-ec27-4231-9d62-3356bb7b5b0e","owner":[],"postedDate":"April 30th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":46498114,"name":"Health sciences/Health care/Health services"},{"id":46498115,"name":"Health sciences/Diseases/Eye diseases/Optic nerve diseases"},{"id":46498116,"name":"Health sciences/Medical research/Outcomes research"}],"tags":[],"updatedAt":"2025-06-02T07:55:32+00:00","versionOfRecord":[],"versionCreatedAt":"2025-04-30 12:10:34","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6244012","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6244012","identity":"rs-6244012","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}