The eye provides novel insights into general health, as well as pathogenesis and development of systemic diseases. In the past decade, growing evidence has demonstrated that the eye's structure and function mirror multiple systemic health conditions, especially in cardiovascular diseases, neurodegenerative disorders, and kidney impairments. This has given rise to the field of oculomics-the application of ophthalmic biomarkers to understand mechanisms, detect and predict disease. The development of this field has been accelerated by three major advances: 1) the availability and widespread clinical adoption of high-resolution and non-invasive ophthalmic imaging ("hardware"); 2) the availability of large studies to interrogate associations ("big data"); 3) the development of novel analytical methods, including artificial intelligence (AI) ("software"). Oculomics offers an opportunity to enhance our understanding of the interplay between the eye and the body, while supporting development of innovative diagnostic, prognostic, and therapeutic tools. These advances have been further accelerated by developments in AI, coupled with large-scale linkage datasets linking ocular imaging data with systemic health data. Oculomics also enables the detection, screening, diagnosis, and monitoring of many systemic health conditions. Furthermore, oculomics with AI allows prediction of the risk of systemic diseases, enabling risk stratification, opening up new avenues for prevention or individualized risk prediction and prevention, facilitating personalized medicine. In this review, we summarise current concepts and evidence in the field of oculomics, highlighting the progress that has been made, remaining challenges, and the opportunities for future research.

Optical Coherence Tomography (OCT) has revolutionized the diagnosis and management of retinal diseases, offering high-resolution, cross-sectional imaging that aids in early detection and continuous monitoring. However, traditional OCT devices are limited to clinical settings and require a technician to operate, which poses accessibility challenges such as a lack of appointment availability, patient and family burden of frequent transportation, and heightened healthcare costs, especially when treatable pathology is undetected. With the increasing global burden of retinal conditions such as age-related macular degeneration (AMD) and diabetic retinopathy, there is a critical need for improved accessibility in the detection of retinal diseases. Advances in biomedical engineering have led to innovations such as portable models, community-based systems, and artificial intelligence-enabled image analysis. The SightSync OCT is a community-based, technician-free device designed to enhance accessibility while ensuring secure data transfer and high-quality imaging (6 × 6 mm resolution, 80,000 A-scans/s). With its compact design and potential for remote interpretation, SightSync widens the possibility for community-based screening for vision-threatening retinal diseases. By integrating innovations in OCT imaging, the future of monitoring for retinal disease can be transformed to reduce barriers to care and improve patient outcomes. This article discusses the evolution of OCT technology, its role in the diagnosis and management of retinal diseases, and how novel engineering solutions like SightSync OCT are transforming accessibility in retinal imaging.

To evaluate the efficacy of artificial intelligence (AI) in screening for diabetic retinopathy (DR) using fundus images and optical coherence tomography (OCT) in comparison to traditional screening methods.

This systematic review was registered with PROSPERO (ID: CRD42024560750). Systematic searches were conducted in PubMed Medline, Cochrane Central, ScienceDirect, and Web of Science using keywords such as "diabetic retinopathy," "screening," and "artificial intelligence." Only studies published in English from 2019 to July 22, 2024, were considered. We also manually reviewed the reference lists of relevant reviews. Two independent reviewers assessed the risk of bias using the QUADAS-2 tool, resolving disagreements through discussion with the principal investigator. Meta-analysis was performed using MetaDiSc software (version 1.4). To calculate combined sensitivity, specificity, summary receiver operating characteristic (SROC) plots, forest plots, and subgroup analyses were performed according to clinician type (ophthalmologists vs. retina specialists) and imaging modality (fundus images vs. fundus images + OCT).

18 studies were included. Meta-analysis showed that AI systems demonstrated superior diagnostic performance compared to doctors, with the pooled sensitivity, specificity, diagnostic odds ratio, and Cochrane Q index of the AI being 0.877, 0.906, 0.94, and 153.79 accordingly. The Fagan nomogram analysis further confirmed the strong diagnostic value of AI. Subgroup analyses revealed that factors like imaging modality, and doctor expertise can influence diagnostic performance.

AI systems have demonstrated strong diagnostic performance in detecting diabetic retinopathy, with sensitivity and specificity comparable to or exceeding traditional clinicians.

Metabolic syndrome is leading to an increased risk of diabetes and cardiovascular disease. Our study developed a model using retinal image data from fundus photographs taken during comprehensive health check-ups to classify metabolic syndrome. The model achieved an AUC of 0.7752 (95% CI: 0.7719-0.7786) using retinal images, and an AUC of 0.8725 (95% CI: 0.8669-0.8781) when combining retinal images with basic clinical features. Furthermore, we propose a method to improve the interpretability of the relationship between retinal image features and metabolic syndrome by visualizing metabolic syndrome-related areas in retinal images. The results highlight the potential of retinal images in classifying metabolic syndrome.

Diabetic retinopathy constitutes a major complication associated with diabetes mellitus, resulting in visual impairment and blindness on a global scale. The pathophysiology of DR is characterized by intricate interactions among metabolic, hemodynamic, and inflammatory pathways, which include the activation of the polyol pathway, the accumulation of advanced glycation end products, the overactivation of protein kinase C, dysregulation of the renin-angiotensin-aldosterone system, and retinal neurodegeneration. This review investigates the classification, complex pathophysiology, and therapeutic modalities for DR, encompassing conventional interventions such as anti-VEGF agents, aldose reductase inhibitors, angiotensin receptor blockers, laser photocoagulation, and vitrectomy. Innovative treatments, including advanced anti-VEGF agents, neuroprotective strategies, gene and stem cell therapies, and advancements in drug delivery systems, exhibit considerable transformative potential. Furthermore, integrating artificial intelligence for early detection and modulation of inflammatory pathways signifies cutting-edge progress in the field. By integrating contemporary knowledge and prospective avenues, this review underscores the significance of comprehending the multifaceted nature of DR and the advancements in its therapeutic approaches. The objective is to bridge the gaps between research findings and clinical application, thereby providing a comprehensive resource to enhance outcomes and quality of life for individuals impacted by DR.

The nonmydriatic, noncontact, and rapid acquisition features of ultra-widefield fundus (UWF) imaging create an invaluable tool for pediatric retinal screening in primary care. This review assesses the landscape of pediatric eye screening, identifies gaps in diagnosing a range of pediatric retinal conditions, and discusses potential uses of UWF imaging for retinal screening. The standards for pediatric eye screening in primary care include red reflex testing, direct ophthalmoscopy, external ocular exam, instrument-based screening, and visual acuity testing. These tests fail to diagnose several treatable retinal diseases. In this gap, UWF retinal imaging provides a panoramic view of the retinal landscape, allowing for a more comprehensive examination. For several pediatric retinal conditions (eg, retinal detachment, retinopathy of prematurity, Coats' disease, familial exudative vitreoretinopathy, Stargardt disease, ocular toxocariasis), UWF retinal imaging provides the high spatial resolution necessary for reliable diagnosis, expediting time to treatment while maintaining low false positive rates. [Ophthalmic Surg Lasers Imaging Retina 2025;56:232-239.].

Artificial intelligence (AI) is gaining widespread traction in ophthalmology, with multiple screening and diagnostic tools already being approved by U. S. and EU authorities. However, the adoption of these tools among medical professionals and their acceptance among patients is still questionable. This narrative review analyses the current literature on stakeholder perspectives on the integration of AI in ophthalmology, with a focus on comparing views across different global healthcare contexts. A PubMed search was conducted for original research articles published between January 1, 2015 and August 31, 2024. The analysis revealed different levels of acceptance for different AI applications among different stakeholder groups. Ophthalmologists and optometrists generally showed positive attitudes toward AI as an adjunct tool, while patients expressed mixed views, appreciating potential benefits while expressing concerns about a lack of transparency in the integration of AI into healthcare. This review reveals a complex landscape of stakeholder perspectives on AI in ophthalmology, highlighting the need for tailored approaches to AI implementation that address specific concerns and consider different healthcare contexts. The findings underscore the importance of collaborative efforts to develop context-specific, effective AI solutions in ophthalmology.

The "black box" nature of many artificial intelligence (AI) models has limited their adoption in real-world ophthalmologic practices. Our lab developed an AI model for detecting the presence of a choroidal melanocytic lesion (CML) in colour fundus images. The purpose of this article is to investigate whether there are known clinical features of CMLs that are associated with false-negative (FN) classifications from the model to aid in validation and increase its interpretability.

A retrospective cohort study of CML patients was performed. A total of 388 fundus images from 194 patients with (n = 194) and without (n = 194) CMLs collected through routine clinical assessment were used to train an AI model. The model's classification (lesion present/lesion absent) of the images with CMLs, as well as CML characteristics, demographics, and risk factors for uveal melanoma (UM) were extracted. Logistic regression models were used to test for associations between the FN classifications and these characteristics.

The AI model returned 150 true-positive classifications and 44 FN classifications (23%) for CML eyes. Thinner lesions were more likely to be missed by the model (p = 0.026), resulting in a FN classification. The presence of imaging risk factors for UM was not shown to have any statistically significant relationships with a FN classification.

The results from this study demonstrate that the FN classifications for CML fundus image classifications from our AI model are not associated with the presence of imaging risk factors for UM but are influenced by thinness of the lesion.

To evaluate the use of an autonomous artificial intelligence (AI)-based device to screen for diabetic retinopathy (DR) and to evaluate the frequency of diabetes mellitus (DM) and DR in an under-resourced population served by the Stanford Belize Vision Clinic (SBVC).

The records of all patients from 2017 to 2024 were collected and analyzed, dividing the study into two time periods: Pre-AI (before June 2022, prior to the implementation of the LumineticsCore® device at SBVC) and Post-AI (from June 2022 to the present) and subdivided into post-COVID19 and pre-COVID19 periods. Patients were categorized based on self-reported past medical history (PMH) as DM positive (diagnosed DM) and DM negative (no PMH of DM). AI camera outcomes included: negative for more than mild DR (MTMDR), positive for MTMDR, and insufficient exam quality.

A total of 1897 patients with a mean age of 47.6 years were included. The gradability of encounters by the AI device was 89.1%. The frequency of DR detection increased significantly in the Post-AI period (55/639) compared to the Pre-AI period (38/1258), including during the COVID-19 pandemic. The mean age of DR diagnosis was significantly lower in the Post-AI period (44.1 years) compared to Pre-AI period (60.7 years) among DM negative patients. There was a significant association between having DR and hypertension. Additionally, the detection rate of DM increased in the Post-AI period compared to Pre-AI period.

Autonomous AI-based screening significantly improves the detection of patients with DR in areas with limited healthcare resources by reducing dependence on on-field ophthalmologists. This innovative approach can be seamlessly integrated into primary care settings, with technicians capturing images quickly and efficiently within just a few minutes. This study demonstrates the effectiveness of autonomous AI in identifying patients with both DR and DM, as well as associated high-burden diseases such as hypertension, across various age ranges.

This study aimed to compare general ophthalmologists, retina specialists, and Aireen AI screening system with the clinical reference standard of a three-member high-level expert committee for diabetic retinopathy (DR) in the evaluation of fundus images for DR.

The study was designed as a diagnostic, multicenter, cross-sectional, non-randomized diagnostic study. The cohort included in the clinical investigation consisted of 1274 patients with diabetes mellitus (DM) type I or II. Each patient underwent one-field fundus photography using a non-mydriatic camera to assess findings of DR. One hundred and nineteen subjects (9.3%) were excluded from the clinical investigation based on Aireen system assessment. In the clinical investigation, all images were assessed at three independent levels of evaluation: 1) general ophthalmologists (GO) - without subspecialty training in the retina; 2) retina specialists (RS); and 3) system Aireen. In cases where there may be disagreements amongst groups, the image is referred for assessment by the Diabetic Retinopathy Board (DRB).

The overall prevalence of any DR was 31.9% (368 cases out of 1154 DM), according to the DRB. Overall concordance between AI system Aireen and GO and RS assessments in the detection of DR from fundus photography occurred in 734 cases (63.6%). The number of disagreements between Aireen system, GO and RS evaluation occurred in 420 (36.4%) cases. Sensitivity for GO was 87.0% (95% CI: 83.6; 90.4), for RS was 82.9% (95% CI: 79.1; 86.7), and for AI system Aireen was 92.1% (95% CI: 89.3; 94.9). Specificity was 76.5% (95% CI: 73.5; 79.5), 81.2% (95% CI: 78.5; 83.9), and 90.7% (95% CI: 88.7; 92.7) for GO, RS and AI system Aireen, respectively.

This real-world study illustrates the potential use of AI system Aireen in screening for DR. It exhibits higher sensitivity and specificity compared to telemedicine evaluation of one field fundus image.

Screening for diabetic retinopathy (DR) by ophthalmologists is costly and labour-intensive. Artificial Intelligence (AI) for automated DR detection could be a clinically and economically alternative. We assessed the performance of a confocal fundus imaging system (DRSplus, Centervue SpA), coupled with an AI algorithm (RetCAD, Thirona B.V.) in a real-world setting.

45° non-mydriatic retinal images from 506 patients with diabetes were graded both by an ophthalmologist and by the AI algorithm, according to the International Clinical Diabetic Retinopathy severity scale. Less than moderate retinopathy (DR scores 0, 1) was defined as non-referable, while more severe stages were defined as referable retinopathy. The gradings were then compared both at eye-level and patient-level. Key metrics included sensitivity, specificity all measured with a 95% Confidence Interval.

The percentage of ungradable eyes according to the AI was 2.58%. The performances of the AI algorithm for detecting referable DR were 97.18% sensitivity, 93.73% specificity at eye-level and 98.70% sensitivity and 91.06% specificity at patient-level.

DRSplus paired with RetCAD represents a reliable DR screening solution in a real-world setting. The high sensitivity of the system ensures that almost all patients requiring medical attention for DR are referred to an ophthalmologist for further evaluation.

To compare the performance of Medios (offline) and EyeArt (online) artificial intelligence (AI) algorithms for detecting diabetic retinopathy (DR) on images captured using fundus-on-smartphone photography in a remote outreach field setting.

In June, 2019 in the Yucatan Peninsula, 248 patients, many of whom had chronic visual impairment, were screened for DR using two portable Remidio fundus-on-phone cameras, and 2130 images obtained were analyzed, retrospectively, by Medios and EyeArt. Screening performance metrics also were determined retrospectively using masked image analysis combined with clinical examination results as the reference standard.

A total of 129 patients were determined to have some level of DR; 119 patients had no DR. Medios was capable of evaluating every patient with a sensitivity (95% confidence intervals [CIs]) of 94% (88%-97%) and specificity of 94% (88%-98%). Owing primarily to photographer error, EyeArt evaluated 156 patients with a sensitivity of 94% (86%-98%) and specificity of 86% (77%-93%). In a head-to-head comparison of 110 patients, the sensitivities of Medios and EyeArt were 99% (93%-100%) and 95% (87%-99%). The specificities for both were 88% (73%-97%).

Medios and EyeArt AI algorithms demonstrated high levels of sensitivity and specificity for detecting DR when applied in this real-world field setting. Both programs should be considered in remote, large-scale DR screening campaigns where immediate results are desirable, and in the case of EyeArt, online access is possible.

Glaucoma, a leading cause of blindness, requires accurate early detection. We present an AI-based Glaucoma Screening (AI-GS) network comprising six lightweight deep learning models (total size: 110 MB) that analyze fundus images to identify early structural signs such as optic disc cupping, hemorrhages, and nerve fiber layer defects. The segmentation of the optic cup and disc closely matches that of expert ophthalmologists. AI-GS achieved a sensitivity of 0.9352 (95% CI 0.9277-0.9435) at 95% specificity. In real-world testing, sensitivity dropped to 0.5652 (95% CI 0.5218-0.6058) at ~0.9376 specificity (95% CI 0.9174-0.9562) for the standalone binary glaucoma classification model, whereas the full AI-GS network maintained higher sensitivity (0.8053, 95% CI 0.7704-0.8382) with good specificity (0.9112, 95% CI 0.8887-0.9356). The sub-models in AI-GS, with enhanced capabilities in detecting early glaucoma-related structural changes, drive these improvements. With low computational demands and tunable detection parameters, AI-GS promises widespread glaucoma screening, portable device integration, and improved understanding of disease progression.

Diabetic retinopathy (DR) remains a leading cause of vision impairment and blindness among individuals with diabetes, necessitating innovative approaches to screening and management. This editorial explores the transformative potential of artificial intelligence (AI) and machine learning (ML) in revolutionizing DR care. AI and ML technologies have demonstrated remarkable advancements in enhancing the accuracy, efficiency, and accessibility of DR screening, helping to overcome barriers to early detection. These technologies leverage vast datasets to identify patterns and predict disease progression with unprecedented precision, enabling clinicians to make more informed decisions. Furthermore, AI-driven solutions hold promise in personalizing management strategies for DR, incorporating predictive analytics to tailor interventions and optimize treatment pathways. By automating routine tasks, AI can reduce the burden on healthcare providers, allowing for a more focused allocation of resources towards complex patient care. This review aims to evaluate the current advancements and applications of AI and ML in DR screening, and to discuss the potential of these technologies in developing personalized management strategies, ultimately aiming to improve patient outcomes and reduce the global burden of DR. The integration of AI and ML in DR care represents a paradigm shift, offering a glimpse into the future of ophthalmic healthcare.

Background: Polycystic ovary syndrome (PCOS) is the most prevalent endocrinopathy in women of reproductive age, characterized by a broad spectrum of clinical, metabolic, and ultrasound findings. Over time, ultrasound has evolved into a cornerstone for diagnosing polycystic ovarian morphology (PCOM), thanks to advances in probe technology, 3D imaging, and novel stromal markers. The recent incorporation of artificial intelligence (AI) further enhances diagnostic precision by reducing operator-related variability. Methods: We conducted a narrative review of English-language articles in PubMed and Embase using the keywords "PCOS", "polycystic ovary syndrome", "ultrasound", "3D ultrasound", and "ovarian stroma". Studies on diagnostic criteria, imaging modalities, stromal assessment, and machine-learning algorithms were prioritized. Additional references were identified via citation screening. Results: Conventional 2D ultrasound remains essential in clinical practice, with follicle number per ovary (FNPO) and ovarian volume (OV) functioning as primary diagnostic criteria. However, sensitivity and specificity values vary significantly depending on probe frequency, cut-off thresholds (≥12, ≥20, or ≥25 follicles), and patient characteristics (e.g., adolescence, obesity). Three-dimensional (3D) ultrasound and Doppler techniques refine PCOS diagnosis by enabling automated follicle measurements, stromal/ovarian area ratio assessments, and evaluation of vascular indices correlating strongly with hyperandrogenism. Meanwhile, AI-driven ultrasound analysis has emerged as a promising tool for minimizing observer bias and validating advanced metrics (e.g., SA/OA ratio) that may overcome traditional limitations of stroma-based criteria. Conclusions: The continual evolution of ultrasound, encompassing higher probe frequencies, 3D enhancements, and now AI-assisted algorithms, has expanded our ability to characterize PCOM accurately. Nevertheless, challenges such as operator dependency and inter-observer variability persist despite standardized protocols; the integration of AI holds promise in further enhancing diagnostic accuracy. Future directions should focus on robust AI training datasets, multicenter validation, and age-/BMI-specific cut-offs to optimize the balance between sensitivity and specificity, ultimately facilitating earlier and more precise PCOS diagnoses.

Autonomous artificial intelligence (AI) for pediatric diabetic retinal disease (DRD) screening has demonstrated safety, effectiveness, and the potential to enhance health equity and clinician productivity. We examined the cost-effectiveness of an autonomous AI strategy versus a traditional eye care provider (ECP) strategy during the initial year of implementation from a health system perspective. The incremental cost-effectiveness ratio (ICER) was the main outcome measure. Compared to the ECP strategy, the base-case analysis shows that the AI strategy results in an additional cost of $242 per patient screened to a cost saving of $140 per patient screened, depending on health system size and patient volume. Notably, the AI screening strategy breaks even and demonstrates cost savings when a pediatric endocrine site screens 241 or more patients annually. Autonomous AI-based screening consistently results in more patients screened with greater cost savings in most health system scenarios.

Artificial intelligence (AI) is increasingly being used to diagnose complications of diabetes. Artificial intelligence is technology that enables computers and machines to simulate human intelligence and solve complicated problems. In this article, we address current and likely future applications for AI to be applied to diabetes and its complications, including pharmacoadherence to therapy, diagnosis of hypoglycemia, diabetic eye disease, diabetic kidney diseases, diabetic neuropathy, diabetic foot ulcers, and heart failure in diabetes.Artificial intelligence is advantageous because it can handle large and complex datasets from a variety of sources. With each additional type of data incorporated into a clinical picture of a patient, the calculation becomes increasingly complex and specific. Artificial intelligence is the foundation of emerging medical technologies; it will power the future of diagnosing diabetes complications.

The rapid advancement of artificial intelligence and machine learning (AI/ML) technologies in healthcare presents significant opportunities for enhancing patient care through innovative diagnostic tools, monitoring systems, and personalized treatment plans. However, these innovative advancements might result in regulatory challenges given recent Supreme Court decisions that impact the authority of regulatory agencies like the Food and Drug Administration (FDA). This paper explores the implications of regulatory uncertainty for the healthcare industry related to balancing innovation in biotechnology and biocomputing with ensuring regulatory uniformity and patient safety. We examine key Supreme Court cases, including Loper Bright Enterprises v. Raimondo, Relentless, Inc. v. Department of Commerce, and Corner Post, Inc. v. Board of Governors of the Federal Reserve System, and their impact on the Chevron doctrine. We also discuss other relevant cases to highlight shifts in judicial approaches to agency deference and regulatory authority that might affect how science is handled in regulatory spaces, including how biocomputing and other health sciences are governed, how scientific facts are applied in policymaking, and how scientific expertise guides decision making. Through a detailed analysis, we assess the potential impact of regulatory uncertainty in healthcare. Additionally, we provide recommendations for the medical community on navigating these challenges.

AbstractAs artificial intelligence (AI) becomes increasingly incorporated into the workflow of clinical practice, it will raise numerous ethical issues and lead to related clinical ethics consultations to address those issues. However, many of the ethical issues associated with AI feature fundamental distinctions from those currently encountered in clinical ethics consultations. Despite potential differences in the types of ethical issues precipitated by AI, little attention has been given to how to approach these issues when they need to be addressed in clinical ethics consultations. In this article, we provide a walkthrough on adapting clinical ethics consultations to look at these issues through an AI lens, which will allow us to recognize essential information and develop targeted questions that guide consultations toward appropriate decisions. We then provide three case studies exploring hypothetical scenarios based on real AI systems and how a clinical ethicist might guide the discussion of ethical issues presented by AI in each scenario. Following the case studies, we further discuss clinical AI in the context of clinical ethics consultations and conclude with a call for more attention to this area of increasing importance.

Introduction Diabetic retinopathy (DR) is a leading cause of blindness globally, emphasizing the urgent need for efficient diagnostic tools. Machine learning, particularly convolutional neural networks (CNNs), has shown promise in automating the diagnosis of retinal conditions with high accuracy. This study evaluates two CNN models, VGG16 and InceptionV3, for classifying retinal optical coherence tomography (OCT) images into four categories: normal, choroidal neovascularization, diabetic macular edema (DME), and drusen. Methods Using 83,000 OCT images across four categories, the CNNs were trained and tested via Python-based libraries, including TensorFlow and Keras. Metrics such as accuracy, sensitivity, and specificity were analyzed with confusion matrices and performance graphs. Comparisons of dataset sizes evaluated the impact on model accuracy with tools deployed on JupyterLab. Results VGG16 and InceptionV3 achieved accuracy between 85% and 95%, with VGG16 peaking at 94% and outperforming InceptionV3 (92%). Larger datasets improved sensitivity by 7% and accuracy across all categories, with the highest performance for normal and drusen classifications. Metrics like sensitivity and specificity positively correlated with dataset size. Conclusions The study confirms CNNs' potential in retinal diagnostics, achieving high classification accuracy. Limitations included reliance on grayscale images and computational intensity, which hindered finer distinctions. Future work should integrate data augmentation, patient-specific variables, and lightweight architectures to optimize performance for clinical use, reducing costs and improving outcomes.

To assess the feasibility, implementation, and patient experience of autonomous artificial intelligence-based diabetic retinopathy detection models.

This was a prospective cohort study where consenting adult participants previously diagnosed with diabetes were screened for diabetic retinopathy using retinal imaging with autonomous artificial intelligence (AI) interpretation at their routine primary care appointment from December 2022 through October 2023 in Thunder Bay, Ontario. Demographic (age, sex, race) and clinical (type and duration of diabetes, last reported eye exam) data were collected using a data collection form. A 5-point Likert scale questionnaire was completed by participants to assess patient experience following the AI exam.

Among the 202 participants (38.6% women) with a mean age of 70.8 ± 11.7 years included in the study and screened by AI, the exam was successfully completed by 93.6% (n = 189), with only 1.5% (n = 3) requiring dilating eyedrops. The most common reason for an unsuccessful exam was small pupils with patient refusal for dilating eyedrops (n = 4). Among the participants with successful eye exams, 22.2% (n = 42) had referable diabetic retinopathy detected and were referred to see an ophthalmologist; 32/42 (76.0%) of these attended their ophthalmologist appointment. A total of 184 participants completed the satisfaction questionnaire; the mean score (out of 5) for satisfaction with the addition of an eye exam to their primary care visit was 4.8 ± 0.6.

Screening for diabetic retinopathy using autonomous artificial intelligence in a primary care setting is feasible and acceptable. This approach has significant advantages for both physicians and patients while achieving very high patient satisfaction.

Artificial Intelligence (AI) tools are gaining popularity in healthcare. OpenAI released ChatGPT on November 30, 2022. ChatGPT is a language model that comprehends and generates human language, providing instant data analysis and recommendations. This is particularly significant in the dynamic field of physiotherapy, where its integration has the potential to enhance healthcare efficiency.

This study aims to evaluate whether ChatGPT-3.5 (free version) provides consistent and accurate clinical responses, its ability to imitate human clinical reasoning in simple and complex scenarios, and its capability to produce a differential diagnosis.

Two studies were conducted using the ChatGPT-3.5. Study 1 evaluated the consistency and accuracy of ChatGPT's responses in clinical assessment using ten user-participants who submitted the phrase "Which are the main steps for a completed physiotherapy assessment?" Study 2 assessed ChatGPT's differential diagnostic ability using published case studies by 2 independent participants. The case reports consisted of one simple and one complex scenario.

Study 1 underscored the variability in ChatGPT's responses, which ranged from comprehensive to concise. Notably, essential steps such as re-assessment and subjective examination were omitted in 30% and 40% of the responses, respectively. In Study 2, ChatGPT demonstrated its capability to develop evidence-based clinical reasoning, particularly evident in simple clinical scenarios. Question phrasing significantly impacted the generated answers.

This study highlights the potential benefits of using ChatGPT in healthcare. It also provides a balanced perspective on ChatGPT's strengths and limitations and emphasizes the importance of using AI tools in a responsible and informed manner.

Clinical medicine requires the integration of various forms of patient data including demographics, symptom characteristics, electrocardiogram findings, laboratory values, biomarker levels, and imaging studies. Decision-making on the optimal management should be based on a high probability that the envisaged treatment is appropriate, provides benefit, and bears no or little potential harm. To that end, personalized risk-benefit considerations should guide the management of individual patients to achieve optimal results. These basic clinical tasks have become more and more challenging with the massively growing data now available; artificial intelligence and machine learning (AI/ML) can provide assistance for clinicians by obtaining and comprehensively preparing the history of patients, analysing face and voice and other clinical features, by integrating laboratory results, biomarkers, and imaging. Furthermore, AI/ML can provide a comprehensive risk assessment as a basis of optimal acute and chronic care. The clinical usefulness of AI/ML algorithms should be carefully assessed, validated with confirmation datasets before clinical use, and repeatedly re-evaluated as patient phenotypes change. This review provides an overview of the current data revolution that has changed and will continue to change the face of clinical medicine radically, if properly used, to the benefit of physicians and patients alike.

The opaqueness of artificial intelligence (AI) algorithms decision processes limit their application in healthcare. Our objective was to explore discrepancies in heatmaps originated from slightly different retinal images from the same eyes of individuals with diabetes, to gain insights into the deep learning (DL) decision process.

Pairs of retinal images from the same eyes of individuals with diabetes, composed of images obtained before and after pupil dilation, underwent automatic analysis by a convolutional neural network for the presence of diabetic retinopathy (DR), output being a score ranging from 0 to 1. Gradient-based Class Activation Maps (GradCam) allowed visualization of activated areas. Pairs of images with discordant DL scores or outputs within the pair were objectively compared to the concordant pairs, regarding the sum of activations of Class Activation Mapping (CAM), the number of activated areas, and DL score differences. Heatmaps of discordant pairs were also qualitatively assessed.

Algorithmic performance for the detection of DR attained 89.8% sensitivity, 96.3% specificity and area under the receiver operating characteristic (ROC) curve of 0.95. Out of 210 comparable pairs of images, 20 eyes and 10 eyes were considered discordant according to DL score difference and regarding DL output, respectively. Comparison of concordant versus discordant groups showed statistically significant differences for all objective variables. Qualitative analysis pointed to subtle differences in image quality within discordant pairs.

The successfully established relationship among objective parameters extracted from heatmaps and DL output discrepancies reinforces the role of heatmaps for DL explainability, fostering acceptance of DL systems for clinical use.

Primary angle closure glaucoma (PACG) is a major cause of visual impairment, particularly in Asia. Although effective screening tools are necessary, the current gold standard is complex and time-consuming, requiring extensive expertise. Artificial intelligence has introduced new opportunities for innovation in ophthalmic imaging. Anterior chamber depth (ACD) is a key risk factor for angle closure and has been suggested as a quick screening parameter for PACG. This study aims to develop an AI algorithm to quantitatively predict ACD from anterior segment photographs captured using a portable smartphone slit-lamp microscope. We retrospectively collected 204,639 frames from 1586 eyes, with ACD values obtained by anterior-segment OCT. We developed two models, (Model 1) diagnosable frame extraction and (Model 2) ACD estimation, using SWSL ResNet as the machine learning model. Model 1 achieved an accuracy of 0.994. Model 2 achieved an MAE of 0.093 ± 0.082 mm, an MSE of 0.123 ± 0.170 mm, and a correlation of R = 0.953. Furthermore, our model's estimation of the risk for angle closure showed a sensitivity of 0.943, specificity of 0.902, and an area under the curve (AUC) of 0.923 (95%CI: 0.878-0.968). We successfully developed a high-performance ACD estimation model, laying the groundwork for predicting other quantitative measurements relevant to PACG screening.

Federally qualified health centers (FQHCs) are federally funded community health clinics that provide comprehensive care to underserved populations, making them potential opportunities to offer eye care and address unmet health care needs. Evaluating the presence of eye care services at FQHCs in Florida is important in understanding and addressing possible gaps in care for the state's large uninsured and underserved populations.

To determine whether FQHCs in Florida are currently offering eye care services, where they are available, what services are being offered, and who provides them.

This study used a cross-sectional design conducted within 1 year (from November 2023 to February 2024). FQHCs listed in the US Health Resources and Services Administration database were contacted by telephone to inquire about the presence of eye care services. The FQHCs were located in both urban and rural areas in Florida to assess accessibility of eye care services in the state. School-based health centers and nonophthalmic specialty care health centers were excluded. A total of 437 FQHCs were included.

Primary outcomes included the presence of eye care services, types of services offered, clinician type (optometrists or ophthalmologists), frequency of services, and availability of pediatric services.

Among 437 FQHCs contacted, only 39 (8.9%) reported offering eye care services. These services primarily included vision examinations, glasses prescriptions, and dilated eye examinations. Optometrists were the primary providers of services at all clinics, with no clinics reporting care by ophthalmologists. The frequency of services varied considerably, ranging from daily to bimonthly. Thirty-seven (94.9%) of the 39 clinics offered pediatric eye care services.

The low prevalence of FQHCs with eye care services and the absence of ophthalmologist-provided care highlight a gap in access to eye care for underserved populations in Florida. These findings support investigations into implementing eye care services and interventions at FQHCs that might enhance access and equity in eye care.

Artificial intelligence (AI) has been applied in healthcare for diagnosis, treatments, disease management, and for studying underlying mechanisms and disease complications in diseases like diabetes and metabolic disorders. This review is a comprehensive overview of various applications of AI in the healthcare system for managing diabetes. A literature search was conducted on PubMed to locate studies integrating AI in the diagnosis, treatment, management and prevention of diabetes. As diabetes is now considered a pandemic now so employing AI and machine learning approaches can be applied to limit diabetes in areas with higher prevalence. Machine learning algorithms can visualize big datasets, and make predictions. AI-powered mobile apps and the closed-loop system automated glucose monitoring and insulin delivery can lower the burden on insulin. AI can help identify disease markers and potential risk factors as well. While promising, AI's integration in the medical field is still challenging due to privacy, data security, bias, and transparency. Overall, AI's potential can be harnessed for better patient outcomes through personalized treatment.

Computer vision is a powerful tool in medical image analysis, supporting the early detection and classification of eye diseases. Diabetic retinopathy (DR), a severe eye disease secondary to diabetes, accompanies several early signs of eye-threatening conditions, such as microaneurysms (MAs), hemorrhages (HEMOs), and exudates (EXs), which have been widely studied and targeted as objects to be detected by computer vision models. In this work, we tested the performances of the state-of-the-art YOLOv8 and YOLOv9 architectures on DR fundus features segmentation without coding experience or a programming background. We took one hundred DR images from the public MESSIDOR database, manually labelled and prepared them for pixel segmentation, and tested the detection abilities of different model variants. We increased the diversity of the training sample by data augmentation, including tiling, flipping, and rotating the fundus images. The proposed approaches reached an acceptable mean average precision (mAP) in detecting DR lesions such as MA, HEMO, and EX, as well as a hallmark of the posterior pole of the eye, such as the optic disc. We compared our results with related works in the literature involving different neural networks. Our results are promising, but far from being ready for implementation into clinical practice. Accurate lesion detection is mandatory to ensure early and correct diagnoses. Future works will investigate lesion detection further, especially MA segmentation, with improved extraction techniques, image pre-processing, and standardized datasets.

This study aims to assess systemic risk factors in diabetes mellitus (DM) patients and predict diabetic retinopathy (DR) using a Random Forest (RF) classification model.

We included DM patients presenting to the retina clinic for first-time DR screening. Data on age, gender, diabetes type, treatment history, DM control status, family history, pregnancy history, and systemic comorbidities were collected. DR and sight-threatening DR (STDR) were diagnosed via a dilated fundus examination. The dataset was split 80:20 into training and testing sets. The RF model was trained to detect DR and STDR separately, and its performance was evaluated using misclassification rates, sensitivity, and specificity.

Data from 1416 DM patients were analyzed. The RF model was trained on 1132 (80%) patients. The misclassification rates were 0% for DR and ~20% for STDR in the training set. External testing on 284 (20%) patients showed 100% accuracy, sensitivity, and specificity for DR detection. For STDR, the model achieved 76% (95% CI-70.7%-80.7%) accuracy, 53% (95% CI-39.2%-66.6%) sensitivity, and 80% (95% CI-74.6%-84.7%) specificity.

The RF model effectively predicts DR in DM patients using systemic risk factors, potentially reducing unnecessary referrals for DR screening. However, further validation with diverse datasets is necessary to establish its reliability for clinical use.

Cardiac autonomic neuropathy (CAN) in diabetes mellitus (DM) is independently associated with cardiovascular (CV) events and CV death. Diagnosis of this complication of DM is time-consuming and not routinely performed in the clinical practice, in contrast to fundus retinal imaging which is accessible and routinely performed. Whether artificial intelligence (AI) utilizing retinal images collected through diabetic eye screening can provide an efficient diagnostic method for CAN is unknown.

This was a single center, observational study in a cohort of patients with DM as a part of the Cardiovascular Disease in Patients with Diabetes: The Silesia Diabetes-Heart Project (NCT05626413). To diagnose CAN, we used standard CV autonomic reflex tests. In this analysis we implemented AI-based deep learning techniques with non-mydriatic 5-field color fundus imaging to identify patients with CAN. Two experiments have been developed utilizing Multiple Instance Learning and primarily ResNet 18 as the backbone network. Models underwent training and validation prior to testing on an unseen image set.

In an analysis of 2275 retinal images from 229 patients, the ResNet 18 backbone model demonstrated robust diagnostic capabilities in the binary classification of CAN, correctly identifying 93% of CAN cases and 89% of non-CAN cases within the test set. The model achieved an area under the receiver operating characteristic curve (AUCROC) of 0.87 (95% CI 0.74-0.97). For distinguishing between definite or severe stages of CAN (dsCAN), the ResNet 18 model accurately classified 78% of dsCAN cases and 93% of cases without dsCAN, with an AUCROC of 0.94 (95% CI 0.86-1.00). An alternate backbone model, ResWide 50, showed enhanced sensitivity at 89% for dsCAN, but with a marginally lower AUCROC of 0.91 (95% CI 0.73-1.00).

AI-based algorithms utilising retinal images can differentiate with high accuracy patients with CAN. AI analysis of fundus images to detect CAN may be implemented in routine clinical practice to identify patients at the highest CV risk.

This is a part of the Silesia Diabetes-Heart Project (Clinical-Trials.gov Identifier: NCT05626413).

Diabetic retinopathy (DR) remains a leading cause of vision loss worldwide, with early detection critical for preventing irreversible damage. This review explores the current landscape and future directions of artificial intelligence (AI)-enhanced detection of DR from fundus images. Recent advances in deep learning and computer vision have enabled AI systems to analyze retinal images with expert-level accuracy, potentially transforming DR screening. Key developments include convolutional neural networks achieving high sensitivity and specificity in detecting referable DR, multi-task learning approaches that can simultaneously detect and grade DR severity, and lightweight models enabling deployment on mobile devices. While these AI systems show promise in improving the efficiency and accessibility of DR screening, several challenges remain. These include ensuring generalizability across diverse populations, standardizing image acquisition and quality, addressing the "black box" nature of complex models, and integrating AI seamlessly into clinical workflows. Future directions in the field encompass explainable AI to enhance transparency, federated learning to leverage decentralized datasets, and the integration of AI with electronic health records and other diagnostic modalities. There is also growing potential for AI to contribute to personalized treatment planning and predictive analytics for disease progression. As the technology continues to evolve, maintaining a focus on rigorous clinical validation, ethical considerations, and real-world implementation will be crucial for realizing the full potential of AI-enhanced DR detection in improving global eye health outcomes.

Chronic elevation of blood glucose at first causes relatively minor changes to the neural and vascular components of the retina. As the duration of hyperglycemia persists, the nature and extent of damage increases and becomes readily detectable. While this second, overt manifestation of diabetic retinopathy (DR) has been studied extensively, what prevents maximal damage from the very start of hyperglycemia remains largely unexplored. Recent studies indicate that diabetes (DM) engages mitochondria-based defense during the retinopathy-resistant phase, and thereby enables the retina to remain healthy in the face of hyperglycemia. Such resilience is transient, and its deterioration results in progressive accumulation of retinal damage. The concepts that co-emerge with these discoveries set the stage for novel intellectual and therapeutic opportunities within the DR field. Identification of biomarkers and mediators of protection from DM-mediated damage will enable development of resilience-based therapies that will indefinitely delay the onset of DR.

Valid and updated clinical indicators can serve as important tools in assessing and improving eyecare delivery.

Indicators for diabetic eyecare in Australia were previously developed from guidelines published before 2013 and then used to assess the appropriateness of care delivery through a nationwide patient record card audit (the iCareTrack study). To reflect emerging evidence and contemporary practice, this study aimed to update clinical indicators for optometric care for people with type 2 diabetes in Australia.

Forty-five candidate indicators, including existing iCareTrack and new indicators derived from nine high-quality evidence-based guidelines, were generated. A two-round modified Delphi process where expert panel members rated the impact, acceptability, and feasibility of the indicators on a 9-point scale and voted for inclusion or exclusion of the candidate indicators was used. Consensus on inclusion was reached when the median scores for impact, acceptability, and feasibility were ≥7 and >75% of experts voted for inclusion.

Thirty-two clinical indicators with high acceptability, impact and feasibility ratings (all median scores: 9) were developed. The final indicators were related to history taking (n = 12), physical examination (n = 8), recall period (n = 5), referral (n = 5), and patient education/communication (n = 2). Most (14 of 15) iCareTrack indicators were retained either in the original format or with modifications. New indicators included documenting the type of diabetes, serum lipid level, pregnancy, systemic medications, nephropathy, Indigenous status, general practitioner details, pupil examination, intraocular pressure, optical coherence tomography, diabetic retinopathy grading, recall period for high-risk diabetic patients without retinopathy, referral of high-risk proliferative retinopathy, communication with the general practitioner, and patient education.

A set of 32 updated diabetic eyecare clinical indicators was developed based on contemporary evidence and expert consensus. These updated indicators inform the development of programs to assess and enhance the eyecare delivery for people with diabetes in Australia.

Precision medicine (PM), also termed stratified, individualised, targeted, or personalised medicine, embraces a rapidly expanding area of research, knowledge, and practice. It brings together two emerging health technologies to deliver better individualised care: the many "-omics" arising from increased capacity to understand the human genome and "big data" and data analytics, including artificial intelligence (AI). PM has the potential to transform an individual's health, moving from population-based disease prevention to more personalised management. There is however a tension between the two, with a real risk that this will exacerbate health inequalities and divert funds and attention from basic healthcare requirements leading to worse health outcomes for many. All areas of medicine should consider how this will affect their practice, with PM now strongly encouraged and supported by government initiatives and research funding. In this review, we discuss examples of PM in current practice and its emerging applications in primary care, such as clinical prediction tools that incorporate genomic markers and pharmacogenomic testing. We look towards potential future applications and consider some key questions for PM, including evidence of its real-world impact, its affordability, the risk of exacerbating health inequalities, and the computational and storage challenges of applying PM technologies at scale.

Purpose The purpose of this study is to assess the accuracy of and bias in recommendations for oculoplastic surgeons from three artificial intelligence (AI) chatbot systems. Methods ChatGPT, Microsoft Bing Balanced, and Google Bard were asked for recommendations for oculoplastic surgeons practicing in 20 cities with the highest population in the United States. Three prompts were used: "can you help me find (an oculoplastic surgeon)/(a doctor who does eyelid lifts)/(an oculofacial plastic surgeon) in (city)." Results A total of 672 suggestions were made between (oculoplastic surgeon; doctor who does eyelid lifts; oculofacial plastic surgeon); 19.8% suggestions were excluded, leaving 539 suggested physicians. Of these, 64.1% were oculoplastics specialists (of which 70.1% were American Society of Ophthalmic Plastic and Reconstructive Surgery (ASOPRS) members); 16.1% were general plastic surgery trained, 9.0% were ENT trained, 8.8% were ophthalmology but not oculoplastics trained, and 1.9% were trained in another specialty. 27.7% of recommendations across all AI systems were female. Conclusions Among the chatbot systems tested, there were high rates of inaccuracy: up to 38% of recommended surgeons were nonexistent or not practicing in the city requested, and 35.9% of those recommended as oculoplastic/oculofacial plastic surgeons were not oculoplastics specialists. Choice of prompt affected the result, with requests for "a doctor who does eyelid lifts" resulting in more plastic surgeons and ENTs and fewer oculoplastic surgeons. It is important to identify inaccuracies and biases in recommendations provided by AI systems as more patients may start using them to choose a surgeon.

This is a systematic review of 25 publications on the topics of the prevalence and cost of diabetic retinopathy (DR) in Trinidad and Tobago, the cost of traditional methods of screening for DR, and the use and cost of artificial intelligence (AI) in screening for DR. Analysis of these publications was used to identify and make estimates for how resources allocated to ophthalmology in public health systems in Trinidad and Tobago can be more efficiently utilized by employing AI in diagnosing treatable DR. DR screening was found to be an effective method of detecting the disease. Screening was found to be a universally cost-effective method of disease prevention and for altering the natural history of the disease in the spectrum of low-middle to high-income economies, such as Rwanda, Thailand, China, South Korea, and Singapore. AI and deep learning systems were found to be clinically superior to, or as effective as, human graders in areas where they were deployed, indicating that the systems are clinically safe. They have been shown to improve access to diabetic retinal screening, improve compliance with screening appointments, and prove to be cost-effective, especially in rural areas. Trinidad and Tobago, which is estimated to be disproportionately more affected by the burden of DR when projected out to the mid-21st century, stands to save as much as US$60 million annually from the implementation of an AI-based system to screen for DR versus conventional manual grading.

To study the role of artificial intelligence (AI) to identify key risk factors for diabetic retinopathy (DR) screening and develop recommendations based on clinician and large language model (LLM) based AI platform opinions for newly detected diabetes mellitus (DM) cases.

Five clinicians and three AI applications were given 20 AI-generated hypothetical case scenarios to assess DR screening timing. We calculated inter-rater agreements between clinicians, AI-platforms, and the "majority clinician response" (defined as the maximum number of identical responses provided by the clinicians) and "majority AI-platform" (defined as the maximum number of identical responses among the 3 distinct AI). Scoring was used to identify risk factors of different severity. Three, two, and one points were given to risk factors requiring screening immediately, within a year, and within five years, respectively. After calculating a cumulative screening score, categories were assigned.

Clinicians, AI platforms, and the "majority clinician response" and "majority AI response" had fair inter-rater reliability (k value: 0.21-0.40). Uncontrolled DM and systemic co-morbidities required immediate screening, while family history of DM and a co-existing pregnancy required screening within a year. The absence of these risk factors required screening within 5 years of DM diagnosis. Screening scores in this study were between 0 and 10. Cases with screening scores of 0-2 needed screening within 5 years, 3-5 within 1 year, and 6-12 immediately.

Based on the findings of this study, AI could play a critical role in DR screening of newly diagnosed DM patients by developing a novel DR screening score. Future studies would be required to validate the DR screening score before it could be used as a reference in real-life clinical situations.

Not applicable.

Early detection and treatment of diabetic retinopathy (DR) are critical for decreasing the risk of vision loss and preventing blindness. Community vision screenings may play an important role, especially in communities at higher risk for diabetes. To address the need for increased DR detection and referrals, we evaluated the use of artificial intelligence (AI) for screening DR.

Patient images of 124 eyes were obtained using a 45° Canon Non-Mydriatic CR-2 Plus AF retinal camera in the Department of Endocrinology Clinic (Newark, NJ) and in a community screening event (Newark, NJ). Images were initially classified by an onsite grader and uploaded for analysis by EyeArt, a cloud-based AI software developed by Eyenuk (California, USA). The images were also graded by an off-site retina specialist. Using Fleiss kappa analysis, a correlation was investigated between the three grading systems, the AI, onsite grader, and a US board-certified retina specialist, for a diagnosis of DR and referral pattern.

The EyeArt results, onsite grader, and the retina specialist had a 79% overall agreement on the diagnosis of DR: 86 eyes with full agreement, 37 eyes with agreement between two graders, 1 eye with full disagreement. The kappa value for concordance on a diagnosis was 0.69 (95% CI 0.61-0.77), indicating substantial agreement. Referral patterns by EyeArt, the onsite grader, and the ophthalmologist had an 85% overall agreement: 96 eyes with full agreement, 28 eyes with disagreement. The kappa value for concordance on "whether to refer" was 0.70 (95% CI 0.60-0.80), indicating substantial agreement. Using the board-certified retina specialist as the gold standard, EyeArt had an 81% accuracy (101/124 eyes) for diagnosis and 83% accuracy (103/124 eyes) in referrals. For referrals, the sensitivity of EyeArt was 74%, specificity was 87%, positive predictive value was 72%, and negative predictive value was 88%.

This retrospective cross-sectional analysis offers insights into use of AI in diabetic screenings and the significant role it will play in automated detection of DR. The EyeArt readings were beneficial with some limitations in a community screening environment. These limitations included a decreased accuracy in the presence of cataracts and the functional cost of EyeArt uploads in a community setting.

This study aims to compare artificial intelligence (AI) systems applied in diabetic retinopathy (DR) teleophthalmology screening, currently deployed systems, fairness initiatives and the challenges for implementation.

The review included articles retrieved from PubMed/Medline/EMBASE literature search strategy regarding telemedicine, DR and AI. The screening criteria included human articles in English, Portuguese or Spanish and related to telemedicine and AI for DR screening. The author's affiliations and the study's population income group were classified according to the World Bank Country and Lending Groups.

The literature search yielded a total of 132 articles, and nine were included after full-text assessment. The selected articles were published between 2004 and 2020 and were grouped as telemedicine systems, algorithms, economic analysis and image quality assessment. Four telemedicine systems that perform a quality assessment, image preprocessing and pathological screening were reviewed. A data and post-deployment bias assessment are not performed in any of the algorithms, and none of the studies evaluate the social impact implementations. There is a lack of representativeness in the reviewed articles, with most authors and target populations from high-income countries and no low-income country representation.

Telemedicine and AI hold great promise for augmenting decision-making in medical care, expanding patient access and enhancing cost-effectiveness. Economic studies and social science analysis are crucial to support the implementation of AI in teleophthalmology screening programs. Promoting fairness and generalizability in automated systems combined with telemedicine screening programs is not straightforward. Improving data representativeness, reducing biases and promoting equity in deployment and post-deployment studies are all critical steps in model development.

Diabetic retinopathy (DR) is one of the most common retinal vascular diseases and one of the main causes of blindness worldwide. Early detection and treatment can effectively delay vision decline and even blindness in patients with DR. In recent years, artificial intelligence (AI) models constructed by machine learning and deep learning (DL) algorithms have been widely used in ophthalmology research, especially in diagnosing and treating ophthalmic diseases, particularly DR. Regarding DR, AI has mainly been used in its diagnosis, grading, and lesion recognition and segmentation, and good research and application results have been achieved. This study summarizes the research progress in AI models based on machine learning and DL algorithms for DR diagnosis and discusses some limitations and challenges in AI research.

To evaluate the accuracy and bias of ophthalmologist recommendations made by three AI chatbots, namely ChatGPT 3.5 (OpenAI, San Francisco, CA, USA), Bing Chat (Microsoft Corp., Redmond, WA, USA), and Google Bard (Alphabet Inc., Mountain View, CA, USA). This study analyzed chatbot recommendations for the 20 most populous U.S. cities.

Each chatbot returned 80 total recommendations when given the prompt "Find me four good ophthalmologists in (city)." Characteristics of the physicians, including specialty, location, gender, practice type, and fellowship, were collected. A one-proportion z-test was performed to compare the proportion of female ophthalmologists recommended by each chatbot to the national average (27.2% per the Association of American Medical Colleges (AAMC)). Pearson's chi-squared test was performed to determine differences between the three chatbots in male versus female recommendations and recommendation accuracy.

Female ophthalmologists recommended by Bing Chat (1.61%) and Bard (8.0%) were significantly less than the national proportion of 27.2% practicing female ophthalmologists (p<0.001, p<0.01, respectively). ChatGPT recommended fewer female (29.5%) than male ophthalmologists (p<0.722). ChatGPT (73.8%), Bing Chat (67.5%), and Bard (62.5%) gave high rates of inaccurate recommendations. Compared to the national average of academic ophthalmologists (17%), the proportion of recommended ophthalmologists in academic medicine or in combined academic and private practice was significantly greater for all three chatbots.

This study revealed substantial bias and inaccuracy in the AI chatbots' recommendations. They struggled to recommend ophthalmologists reliably and accurately, with most recommendations being physicians in specialties other than ophthalmology or not in or near the desired city. Bing Chat and Google Bard showed a significant tendency against recommending female ophthalmologists, and all chatbots favored recommending ophthalmologists in academic medicine.