Search

Purpose: To evaluate the capabilities of Chat Generative Pre-Trained Transformer (ChatGPT), as a large language model (LLM), for diagnosing glaucoma using the Ocular Hypertension Treatment Study (OHTS) dataset, and comparing the diagnostic capability of ChatGPT 3.5 and ChatGPT 4.0. Design: Prospective data collection study. Participants: A total of 3170 eyes of 1585 subjects from the OHTS were included in this study. Methods: We selected demographic, clinical, ocular, visual field, optic nerve head photo, and history of disease parameters of each participant and developed case reports by converting tabular data into textual format based on information from both eyes of all subjects. We then developed a procedure using the application programming interface of ChatGPT, a LLM-based chatbot, to automatically input prompts into a chat box. This was followed by querying 2 different generations of ChatGPT (versions 3.5 and 4.0) regarding the underlying diagnosis of each subject. We then evaluated the output responses based on several objective metrics. Main Outcome Measures: Area under the receiver operating characteristic curve (AUC), accuracy, specificity, sensitivity, and F1 score. Results: Chat Generative Pre-Trained Transformer 3.5 achieved AUC of 0.74, accuracy of 66%, specificity of 64%, sensitivity of 85%, and F1 score of 0.72. Chat Generative Pre-Trained Transformer 4.0 obtained AUC of 0.76, accuracy of 87%, specificity of 90%, sensitivity of 61%, and F1 score of 0.92. Conclusions: The accuracy of ChatGPT 4.0 in diagnosing glaucoma based on input data from OHTS was promising. The overall accuracy of ChatGPT 4.0 was higher than ChatGPT 3.5. However, ChatGPT 3.5 was found to be more sensitive than ChatGPT 4.0. In its current forms, ChatGPT may serve as a useful tool in exploring disease status of ocular hypertensive eyes when specific data are available for analysis. In the future, leveraging LLMs with multimodal capabilities, allowing for integration of imaging and diagnostic testing as part of the analyses, could further enhance diagnostic capabilities and enhance diagnostic accuracy. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

OverviewThis study evaluates the diagnostic accuracy of a multimodal large language model (LLM), ChatGPT-4, in recognizing glaucoma using color fundus photographs (CFPs) with a benchmark dataset and without prior training or fine tuning.MethodsThe publicly accessible Retinal Fundus Glaucoma Challenge “REFUGE” dataset was utilized for analyses. The input data consisted of the entire 400 image testing set. The task involved classifying fundus images into either ‘Likely Glaucomatous’ or ‘Likely Non-Glaucomatous’. We constructed a confusion matrix to visualize the results of predictions from ChatGPT-4, focusing on accuracy of binary classifications (glaucoma vs non-glaucoma).ResultsChatGPT-4 demonstrated an accuracy of 90% with a 95% confidence interval (CI) of 87.06%-92.94%. The sensitivity was found to be 50% (95% CI: 34.51%-65.49%), while the specificity was 94.44% (95% CI: 92.08%-96.81%). The precision was recorded at 50% (95% CI: 34.51%-65.49%), and the F1 Score was 0.50.ConclusionChatGPT-4 achieved relatively high diagnostic accuracy without prior fine tuning on CFPs. Considering the scarcity of data in specialized medical fields, including ophthalmology, the use of advanced AI techniques, such as LLMs, might require less data for training compared to other forms of AI with potential savings in time and financial resources. It may also pave the way for the development of innovative tools to support specialized medical care, particularly those dependent on multimodal data for diagnosis and follow-up, irrespective of resource constraints.

Abstract Introduction The purpose of this study was to evaluate the capabilities of large language models such as Chat Generative Pretrained Transformer (ChatGPT) to diagnose glaucoma based on specific clinical case descriptions with comparison to the performance of senior ophthalmology resident trainees. Methods We selected 11 cases with primary and secondary glaucoma from a publicly accessible online database of case reports. A total of four cases had primary glaucoma including open-angle, juvenile, normal-tension, and angle-closure glaucoma, while seven cases had secondary glaucoma including pseudo-exfoliation, pigment dispersion glaucoma, glaucomatocyclitic crisis, aphakic, neovascular, aqueous misdirection, and inflammatory glaucoma. We input the text of each case detail into ChatGPT and asked for provisional and differential diagnoses. We then presented the details of 11 cases to three senior ophthalmology residents and recorded their provisional and differential diagnoses. We finally evaluated the responses based on the correct diagnoses and evaluated agreements. Results The provisional diagnosis based on ChatGPT was correct in eight out of 11 (72.7%) cases and three ophthalmology residents were correct in six (54.5%), eight (72.7%), and eight (72.7%) cases, respectively. The agreement between ChatGPT and the first, second, and third ophthalmology residents were 9, 7, and 7, respectively. Conclusions The accuracy of ChatGPT in diagnosing patients with primary and secondary glaucoma, using specific case examples, was similar or better than senior ophthalmology residents. With further development, ChatGPT may have the potential to be used in clinical care settings, such as primary care offices, for triaging and in eye care clinical practices to provide objective and quick diagnoses of patients with glaucoma.

AIM: To compare intraocular pressure (IOP) readings obtained with Perkins tonometry, iCare Home, iCare 200, and Tonopen to IOP readings obtained with the manometer of a perfusion system to assess the accuracy and reproducibility of each method of tonometry at set pressures. METHODS: The IOP of human cadaveric eyes (n=2) was measured using a manometer inserted into the eye through the optic nerve. IOP measurements were obtained using a Perkins tonometer, iCare Home, iCare 200, and Tonopen. These measurements were compared to set point IOP measurements of a manometer to determine accuracy and reproducibility of each device. RESULTS: Mean IOP readings obtained with the Perkins tonometer compared to manometer readings demonstrated a difference of -1.0±5.0 mm Hg (P=0.45), indicating a lower reading on average than manometery although not significant. Mean IOP difference between iCare 200 and manometer was 5.3±2.2 mm Hg (P

AIM: To report the long-term outcomes of combined excisional goniotomy and manual small incision cataract surgery (MSICS). METHODS: This is a retrospective case series of patients with open angle glaucoma and visually significant cataracts that underwent combined excisional goniotomy and MSICS with one-year follow-up. The medical history, demographic information, and clinical characteristics of each case were recorded. Data regarding changes in vision, intraocular pressure (IOP), the number of glaucoma medications, and the evolution of the disease after surgery were reported. RESULTS: Three patients, with open angle glaucoma and cataracts underwent combined excisional goniotomy and MSICS without adverse events. All patients had improvement in vision compared to baseline measurements. The range of IOP at baseline was from 14 to 18 mm Hg and decrease to a range of 10 to 14 mm Hg after one year of follow-up. Additionally, two patients also decreased their dependence on IOP-lowering medications at the last follow up visit with one patient maintaining baseline level of medication use. CONCLUSION: A combination of excisional goniotomy and MSICS illustrates both the safety and efficacy to treat patients with visually significant cataract and glaucoma. This procedure allows for a more cost-effective surgical approach that matches the needs of resource strained territories around the globe.

AIM: To assess the relationship between axial length (AL) and intraocular lens (IOL) rotation among eyes receiving a toric IOL and subsequently entered into an online toric back-calculator database. METHODS: Retrospective analysis of data collected online via astigmatismfix.com, a freely available online toric back-calculator where surgeons enter pre- and post-operative information to help manage residual postoperative astigmatism. Included records were deemed valid with entry of AL and IOL orientation between January 2017 and March 2019. Rotation was determined by a difference of ≥5° between pre-operative intended IOL orientation and actual post-operative IOL orientation. Frequency and magnitude of rotation are presented with means and associated standard deviation (SD). Linear regression models of this association are presented. RESULTS: Records of 6752 eyes were included in the analysis, of which 74.8% were determined to have a rotated IOL. The magnitude of rotation increased with each millimeter (mm) increase in AL with a mean rotation of 13.3° (SD: 12.8°) for eyes with AL 20-20.9 mm and a maximum mean rotation of 30.6° (SD: 30.3°) among eyes with AL 29-29.9 mm. General linear modeling demonstrated a significant association (P

Purpose: This trial was designed to determine if artificial intelligence (AI)-supported diabetic retinopathy (DR) screening improved referral uptake in Rwanda. Design: The Rwanda Artificial Intelligence for Diabetic Retinopathy Screening (RAIDERS) study was an investigator-masked, parallel-group randomized controlled trial. Participants: Patients ≥ 18 years of age with known diabetes who required referral for DR based on AI interpretation. Methods: The RAIDERS study screened for DR using retinal imaging with AI interpretation implemented at 4 facilities from March 2021 through July 2021. Eligible participants were assigned randomly (1:1) to immediate feedback of AI grading (intervention) or communication of referral advice after human grading was completed 3 to 5 days after the initial screening (control). Main Outcome Measures: Difference between study groups in the rate of presentation for referral services within 30 days of being informed of the need for a referral visit. Results: Of the 823 clinic patients who met inclusion criteria, 275 participants (33.4%) showed positive findings for referable DR based on AI screening and were randomized for inclusion in the trial. Study participants (mean age, 50.7 years; 58.2% women) were randomized to the intervention (n = 136 [49.5%]) or control (n = 139 [50.5%]) groups. No significant intergroup differences were found at baseline, and main outcome data were available for analyses for 100% of participants. Referral adherence was statistically significantly higher in the intervention group (70/136 [51.5%]) versus the control group (55/139 [39.6%]; P = 0.048), a 30.1% increase. Older age (odds ratio [OR], 1.04; 95% confidence interval [CI], 1.02–1.05; P < 0.0001), male sex (OR, 2.07; 95% CI, 1.22–3.51; P = 0.007), rural residence (OR, 1.79; 95% CI, 1.07–3.01; P = 0.027), and intervention group (OR, 1.74; 95% CI, 1.05–2.88; P = 0.031) were statistically significantly associated with acceptance of referral in multivariate analyses. Conclusions: Immediate feedback on referral status based on AI-supported screening was associated with statistically significantly higher referral adherence compared with delayed communications of results from human graders. These results provide evidence for an important benefit of AI screening in promoting adherence to prescribed treatment for diabetic eye care in sub-Saharan Africa.

AIM: To quantify intraoperative and postoperative complications in complex phacoemulsification cataract extraction (phacoemulsification) with iris manipulation compared to non-complex and complex phacoemulsification without iris manipulation. METHODS: All phacoemulsification cases at the University of Colorado between January 1, 2014, and June 30, 2017 were included. Exclusion criteria for the primary outcome of intraoperative complications were planned combination surgery and eyes with less than 28d follow-up. Exclusion criteria for the secondary outcomes of postoperative complications were unplanned additional surgery, and chronic steroid eye drop use prior to surgery. Data including sex, race/ethnicity, surgery length, visual acuity, intraoperative and postoperative complications, and intraocular pressures (IOP) were collected and analyzed utilizing general linear and Logistic regression modeling. RESULTS: The medical records of 5772 eyes were reviewed (500 complex without iris manipulation, 367 with iris manipulation). The number of any intraoperative complication in the complex with iris manipulation and complex without iris manipulation groups was 15 (4.1%) and 26 (5.2%), respectively, compared to 41 (0.8%) in the non-complex group. Postoperative inflammation was found in 135 (2.8%) non-complex cases, 20 (4.1%) complex cases without iris manipulation, and 20 (5.6%) complex cases with iris manipulation. The adjusted odds ratio of postoperative inflammation in phacoemulsification with iris manipulation compared to non-complex was 2.3 (95%CI: 1.3-4.0, P=0.005). The rate of IOP spikes >10 mm Hg was significantly greater in cases with iris manipulation (P=0.001). CONCLUSION: Complex cases have more intraoperative complications. However, only complex cases with iris manipulation led to increase rates of postoperative inflammation and IOP spikes >10 mm Hg.

Glaucoma, a leading cause of irreversible blindness, affects more than 64 million people worldwide and is expected to grow in number due to the aging global population and enhanced methods of detection. Although topical therapies are often effective when used as prescribed, the drawbacks of current medical management methods include poor patient adherence, local and systemic side effects, and in some cases, limited therapeutic efficacy. Novel ocular drug delivery platforms promise to deliver differentiated drug formulations with targeted delivery leveraging patient-independent administration. Several platforms are in various stages of development with promising pre-clinical and clinical data. The Bimatoprost Sustained Release (SR) intracameral implant was approved in the United States in March of 2020, making it the first long-term injectable therapy available for the treatment of glaucoma. This review aims to provide an update on novel sustained release drug delivery systems that are available today as well as those that might be commercialized in coming years.

Abstract Purpose To quantify the effects of modified Ahmed glaucoma valves® (AGV) with anti-fibrotic plate coatings or a plate surface micro-pattern on outflow resistance and tissue response. Methods Twelve New Zealand rabbits were divided into four groups: commercially available AGV implants (n = 3), AGV with hydrophilic coating (n = 3), AGV with heparin coating (n = 3), and AGV with a plate surface micro-pattern (n = 3). After 6 weeks, the anterior chamber silicone tube was cannulated in situ and perfused with 2.5 μL/min of saline. The pressures were recorded with a perfusion system to measure outflow resistance. The rabbits were then euthanized followed by enucleation of all eyes for bleb histological analyses. Results Hydrostatic pressures were significantly lower in AGVs with the hydrophilic plate coating (mean difference −9.6 mm Hg; p

Purpose: To remind eye care providers of the importance of obtaining vertical rasters in OCT evaluation of patients with suspected hypotony maculopathy. Observations: OCT with vertical rasters may identify chorioretinal folds that are missed by ophthalmoscopic examination, fundus photography, and traditional OCT with horizontal rasters alone. Conclusions and importance: In patients with low IOP and decreased vision, OCT imaging with horizonal and vertical rasters should be obtained to diagnose hypotony maculopathy and monitor response to treatment.

Abstract Background Glaucoma patients undergoing phacoemulsification alone have a higher rate of refractive surprise compared to patients without glaucoma. This risk is further increased with combined filtering procedures. Indeed, there are few and conflicting reports on the effect of combined phacoemulsification and micro-invasive glaucoma surgery (MIGS). Here, we look at refractive outcomes of glaucoma patients undergoing phacoemulsification with and without Kahook Dual Blade (KDB) goniotomy. Methods Retrospective chart review of 385 glaucomatous eyes of 281 patients, which underwent either phacoemulsification alone (n = 309) or phacoemulsification with KDB goniotomy (n = 76, phaco-KDB) at the University of Colorado. The main outcome was refractive surprise defined as the difference in target and postoperative refraction spherical equivalent greater than ±0.5 Diopter (D). Results Refractive surprise greater than ±0.5 D occurred in 26.3% of eyes in the phaco-KDB group and 36.2% in the phacoemulsification group (p = 0.11). Refractive surprise greater than ±1.0 D occurred in 6.6% for the phaco-KDB group and 9.7% for the phacoemulsification group (p = 0.08). There was no significant difference in risk of refractive surprise when pre-operative IOP, axial length, keratometry or performance of KDB goniotomy were assessed in univariate analyses. Conclusion There was no difference between refractive outcomes of glaucomatous patients undergoing phacoemulsification with or without KDB goniotomy.

To compare the histological effects of trans-scleral cyclophotocoagulation (TCP) performed with two different probes, the G-probe (IRIDEX Medical Instruments, Mountain View, CA, USA) and the Ciliprobe (Katalyst Surgical, Chesterfield, MO, USA). TCP was performed on two human cadaver eyes from the same corpse. The vertical meridian was marked and opposite sides were treated using either the G-probe or Ciliprobe. The first eye was treated with each probe at 2000ms/2000 mW and the second eye at 3000ms/1500 mW. Histological examination revealed separation and loss of the non-pigmented ciliary epithelium as well as vacuolization in all sections for both probes and settings. Changes to the non-pigmented ciliary epithelium treated at 3000ms/1500 mW were similar between the two probes. A slightly more complete separation of the non-pigmented epithelium was noted on the Ciliprobe treated sections as compared to the G-probe treated sections in the eye treated at 2000ms/2000 mW. Therefore, in human cadaver eyes, both the G-probe and Ciliprobe produced separation, vacuolization, and loss of the non-pigmented ciliary epithelium at two different, clinically utilized settings.

AIM: To report on the efficacy of combined endoscopic cyclophotocoagulation (ECP) and phacoemulsification cataract extraction (PCE) with intraocular lens placement for reduction of intraocular pressure (IOP) and medication burden in glaucoma. METHODS: A retrospective case review of 91 eyes (73 patients) with glaucoma and cataract that underwent combined PCE/ECP surgery was performed. Baseline demographic and ocular characteristics were recorded, as well as intraocular pressure, number of glaucoma medications, and visual acuity postoperatively with 12-month follow-up. Treatment failure was defined as less than 20% reduction in IOP from baseline on two consecutive visits (at 1, 3, 6, or 12mo postoperatively), IOP ≥21 mm Hg or ≤5 mm Hg on two consecutive visits, or additional glaucoma surgery performed within 12mo after PCE/ECP. RESULTS: Overall, mean medicated IOP was reduced from 16.65 mm Hg at baseline to 13.38 mm Hg at 12mo (P

Spectral domain optical coherence tomography (SOCT) is an important tool for the diagnosis and management of glaucoma. In an effort to help automate the detection of progressive structural glaucomatous optic nerve damage, Carl Zeiss Meditec has developed Guided Progression Analysis (GPA) software in conjunction with the Cirrus SOCT (Carl Zeiss Meditec, Dublin, Calif., USA). Here, we report a case in which the GPA merged information from 2 different patients, and another case in which data from alternate eyes was compared, making the final analysis invalid.

Michael B Horsley, Malik Y KahookRocky Mountain Lions Eye Institute, Department of Ophthalmology, University of Colorado Denver, Aurora, CO, USAPurpose: To quantify changes in tear break-up time (TBUT), corneal staining and ocular surface disease index (OSDI) in glaucoma patients after switching therapy from latanoprost with 0.02% benzalkonium chloride (BAK) to travoprost with sofZia™.Methods: Prospective consecutive case series evaluating patients before and 8 weeks after switching from latanoprost with BAK to travoprost with sofZia™ in patients with baseline TBUT less than 6 seconds.Results: Forty eyes of 20 consecutive patients using latanoprost with BAK were switched to travoprost with sofZia™. Mean TBUT prior to starting travoprost was 2.02 ± 0.71 seconds and increased to 6.34 ± 1.31 seconds 8 weeks after the switch (p < 0.001). Mean inferior corneal staining scores decreased from 2.40 ± 0.87 to 1.38 ± 0.59 (p < 0.001). Mean OSDI scores decreased from 26.31 ± 8.25 to 16.56 ± 6.19 (p

Lawrence Y Ho1, Larissa Camejo1, Malik Y Kahook2, Robert Noecker11UMPC Eye Center, University of Pittsburgh School of Medicine, Pittsburgh, PA 15213, USA; 2Rocky Mountain Lions Institute, University of Colorado, CO, USAAbstract: We studied the effects of audible and visual alarms on adherence with a recommended dosing regimen in the management of glaucoma. Forty-two patients were begun on therapy with the Travatan® Dosing Aid (TDA) and randomly divided into two observation groups-one with visual and audible alarm functions turned on and the other with alarms off. Dosing information was analyzed for mean rates of adherence, missed days, and dosing at the wrong time. Twenty patients were randomized to the TDA alarm on group and 22 to the alarm off group. The rates of adherence were 87.9% and 79.7% (p = 0.02), rates of missed dosing were 7.6% and 14.4% (p = 0.03), and rates of dosing at the incorrect times were 7.1% and 9.8% (p = 0.19), respectively for alarm on versus alarm off groups. In the alarm on group, the adherence rate was significantly higher and proportion of missed dosing was significantly lower. It is still yet to be determined whether there is a relationship between adherence and progression of glaucoma.Keywords: adherence, glaucoma, Travatan® Dosing Aid, audible alarms, visual alarms

Purpose. To compare morphologic changes in human trabecular meshwork (TM) after selective laser trabeculoplasty (SLT) and argon laser trabeculoplasty (ALT). Design. Laboratory evaluation of ex vivo human eye TM after laser trabeculoplasty. Methods. Corneoscleral rims from human cadaver eyes were sectioned and treated with varying powers of either SLT or ALT. Specimens were examined using light microscopy, scanning electron microscopy (SEM), and transmission electron microscopy (TEM). Results. TEM of SLT at all powers resulted in disrupted TM cells with cracked and extracellular pigment granules. SEM of SLT samples treated at high power revealed tissue destruction with scrolling of trabecular beams. SEM of ALT-treated tissue showed increasing destruction with exposure to higher power. The presence or absence of “champagne” bubbles during SLT did not alter the histologic findings. Conclusions. SLT-treated human TM revealed disruption of TM cells with cracked, extracellular pigment granules, particularly at higher treatment powers. Tissue scrolling was noted at very high SLT energy levels. ALT-treated tissue showed significant damage to both the superficial and deeper TM tissues in a dose-dependent fashion. Further studies are needed to guide titration of treatment power to maximize the IOP-lowering effect while minimizing both energy delivered and damage to target tissues.

Malik Y Kahook1, Larissa Camejo2, Robert J Noecker21Department of Ophthalmology, The Rocky Mountain Lions Eye Institute, University of Colorado Health Sciences Center, Aurora, CO, USA; 2UPMC Eye Center, Eye and Ear Institute, University of Pittsburgh School of Medicine, Pittsburgh, PA, USAAbstract: Use of topical steroids is an important component of postoperative care after filtration surgery. Efficacy of postoperative medications is affected by patient adherence and physical limitations in the elderly population often prohibit proper dosing of ophthalmic drops. We describe a technique for the use of intraoperative retrobulbar triamcinolone acetonide in trabeculectomy surgery and report on postoperative outcomes. This technique appears safe and may be an attractive method of delivering a steroid depot at the time of trabeculectomy.Keywords: glaucoma, trabeculectomy, steroids

We review multiphoton microscopy (MPM) including two-photon autofluorescence (2PAF), second harmonic generation (SHG), third harmonic generation (THG), fluorescence lifetime (FLIM), and coherent anti-Stokes Raman Scattering (CARS) with relevance to clinical applications in ophthalmology. The different imaging modalities are discussed highlighting the particular strength that each has for functional tissue imaging. MPM is compared with current clinical ophthalmological imaging techniques such as reflectance confocal microscopy, optical coherence tomography, and fluorescence imaging. In addition, we discuss the future prospects for MPM in disease detection and clinical monitoring of disease progression, understanding fundamental disease mechanisms, and real-time monitoring of drug delivery.