Effect of childhood atropine treatment on adult choroidal thickness using sequential deep learning-enabled segmentation.

PURPOSE: To describe choroidal thickness measurements using a sequential deep learning segmentation in adults who received childhood atropine treatment for myopia control. DESIGN: Prospective, observational study. METHODS: Choroidal thickness was measured by swept-source optical coherence tomography in adults who received childhood atropine, segmented using a sequential deep learning approach. RESULTS: Of 422 eyes, 94 (22.3 %) had no previous exposure to atropine treatment, while 328 (77.7 %) had received topical atropine during childhood. After adjusting for age, sex, and axial length, childhood atropine exposure was associated with a thicker choroid by 32.1 µm (95 % CI, 9.2-55.0; P = 0.006) in the inner inferior, 23.5 µm (95 % CI, 1.9-45.1; P = 0.03) in the outer inferior, 21.8 µm (95 % CI, 0.76-42.9; P = 0.04) in the inner nasal, and 21.8 µm (95 % CI, 2.6-41.0; P = 0.03) in the outer nasal. Multivariable analysis, adjusted for age, sex, atropine use, and axial length, showed an independent association between central subfield choroidal thickness and the incidence of tessellated fundus (P < 0.001; OR, 0.97; 95 % CI, 0.96-0.98). CONCLUSIONS: This study demonstrated that short-term (2-4 years) atropine treatment during childhood was associated with an increase in choroidal thickness of 20-40 µm in adulthood (10-20 years later), after adjusting for age, sex, and axial length. We also observed an independent association between eyes with thicker central choroidal measurements and reduced incidence of tessellated fundus. Our study suggests that childhood exposure to atropine treatment may affect choroidal thickness in adulthood.

A Competition for the Diagnosis of Myopic Maculopathy by Artificial Intelligence Algorithms.

Importance: Myopic maculopathy (MM) is a major cause of vision impairment globally. Artificial intelligence (AI) and deep learning (DL) algorithms for detecting MM from fundus images could potentially improve diagnosis and assist screening in a variety of health care settings. Objectives: To evaluate DL algorithms for MM classification and segmentation and compare their performance with that of ophthalmologists. Design, Setting, and Participants: The Myopic Maculopathy Analysis Challenge (MMAC) was an international competition to develop automated solutions for 3 tasks: (1) MM classification, (2) segmentation of MM plus lesions, and (3) spherical equivalent (SE) prediction. Participants were provided 3 subdatasets containing 2306, 294, and 2003 fundus images, respectively, with which to build algorithms. A group of 5 ophthalmologists evaluated the same test sets for tasks 1 and 2 to ascertain performance. Results from model ensembles, which combined outcomes from multiple algorithms submitted by MMAC participants, were compared with each individual submitted algorithm. This study was conducted from March 1, 2023, to March 30, 2024, and data were analyzed from January 15, 2024, to March 30, 2024. Exposure: DL algorithms submitted as part of the MMAC competition or ophthalmologist interpretation. Main Outcomes and Measures: MM classification was evaluated by quadratic-weighted κ (QWK), F1 score, sensitivity, and specificity. MM plus lesions segmentation was evaluated by dice similarity coefficient (DSC), and SE prediction was evaluated by R2 and mean absolute error (MAE). Results: The 3 tasks were completed by 7, 4, and 4 teams, respectively. MM classification algorithms achieved a QWK range of 0.866 to 0.901, an F1 score range of 0.675 to 0.781, a sensitivity range of 0.667 to 0.778, and a specificity range of 0.931 to 0.945. MM plus lesions segmentation algorithms achieved a DSC range of 0.664 to 0.687 for lacquer cracks (LC), 0.579 to 0.673 for choroidal neovascularization, and 0.768 to 0.841 for Fuchs spot (FS). SE prediction algorithms achieved an R2 range of 0.791 to 0.874 and an MAE range of 0.708 to 0.943. Model ensemble results achieved the best performance compared to each submitted algorithms, and the model ensemble outperformed ophthalmologists at MM classification in sensitivity (0.801; 95% CI, 0.764-0.840 vs 0.727; 95% CI, 0.684-0.768; P = .006) and specificity (0.946; 95% CI, 0.939-0.954 vs 0.933; 95% CI, 0.925-0.941; P = .009), LC segmentation (DSC, 0.698; 95% CI, 0.649-0.745 vs DSC, 0.570; 95% CI, 0.515-0.625; P < .001), and FS segmentation (DSC, 0.863; 95% CI, 0.831-0.888 vs DSC, 0.790; 95% CI, 0.742-0.830; P < .001). Conclusions and Relevance: In this diagnostic study, 15 AI models for MM classification and segmentation on a public dataset made available for the MMAC competition were validated and evaluated, with some models achieving better diagnostic performance than ophthalmologists.

Generative artificial intelligence and ethical considerations in health care: a scoping review and ethics checklist.

The widespread use of Chat Generative Pre-trained Transformer (known as ChatGPT) and other emerging technology that is powered by generative artificial intelligence (GenAI) has drawn attention to the potential ethical issues they can cause, especially in high-stakes applications such as health care, but ethical discussions have not yet been translated into operationalisable solutions. Furthermore, ongoing ethical discussions often neglect other types of GenAI that have been used to synthesise data (eg, images) for research and practical purposes, which resolve some ethical issues and expose others. We did a scoping review of the ethical discussions on GenAI in health care to comprehensively analyse gaps in the research. To reduce the gaps, we have developed a checklist for comprehensive assessment and evaluation of ethical discussions in GenAI research. The checklist can be integrated into peer review and publication systems to enhance GenAI research and might be useful for ethics-related disclosures for GenAI-powered products and health-care applications of such products and beyond.

Artificial Intelligence-Based Disease Activity Monitoring to Personalized Neovascular Age-Related Macular Degeneration Treatment: A Feasibility Study.

Purpose: To evaluate the performance of a disease activity (DA) model developed to detect DA in participants with neovascular age-related macular degeneration (nAMD). Design: Post hoc analysis. Participants: Patient dataset from the phase III HAWK and HARRIER (H&H) studies. Methods: An artificial intelligence (AI)-based DA model was developed to generate a DA score based on measurements of OCT images and other parameters collected from H&H study participants. Disease activity assessments were classified into 3 categories based on the extent of agreement between the DA model's scores and the H&H investigators' decisions: agreement ("easy"), disagreement ("noisy"), and close to the decision boundary ("difficult"). Then, a panel of 10 international retina specialists ("panelists") reviewed a sample of DA assessments of these 3 categories that contributed to the training of the final DA model. A panelists' majority vote on the reviewed cases was used to evaluate the accuracy, sensitivity, and specificity of the DA model. Main Outcome Measures: The DA model's performance in detecting DA compared with the DA assessments made by the investigators and panelists' majority vote. Results: A total of 4472 OCT DA assessments were used to develop the model; of these, panelists reviewed 425, categorized as "easy" (17.2%), "noisy" (20.5%), and "difficult" (62.4%). False-positive and false negative rates of the DA model's assessments decreased after changing the assessment in some cases reviewed by the panelists and retraining the DA model. Overall, the DA model achieved 80% accuracy. For "easy" cases, the DA model reached 96% accuracy and performed as well as the investigators (96% accuracy) and panelists (90% accuracy). For "noisy" cases, the DA model performed similarly to panelists and outperformed the investigators (84%, 86%, and 16% accuracies, respectively). The DA model also outperformed the investigators for "difficult" cases (74% and 53% accuracies, respectively) but underperformed the panelists (86% accuracy) owing to lower specificity. Subretinal and intraretinal fluids were the main clinical parameters driving the DA assessments made by the panelists. Conclusions: These results demonstrate the potential of using an AI-based DA model to optimize treatment decisions in the clinical setting and in detecting and monitoring DA in patients with nAMD. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Latest developments of generative artificial intelligence and applications in ophthalmology.

The emergence of generative artificial intelligence (AI) has revolutionized various fields. In ophthalmology, generative AI has the potential to enhance efficiency, accuracy, personalization and innovation in clinical practice and medical research, through processing data, streamlining medical documentation, facilitating patient-doctor communication, aiding in clinical decision-making, and simulating clinical trials. This review focuses on the development and integration of generative AI models into clinical workflows and scientific research of ophthalmology. It outlines the need for development of a standard framework for comprehensive assessments, robust evidence, and exploration of the potential of multimodal capabilities and intelligent agents. Additionally, the review addresses the risks in AI model development and application in clinical service and research of ophthalmology, including data privacy, data bias, adaptation friction, over interdependence, and job replacement, based on which we summarized a risk management framework to mitigate these concerns. This review highlights the transformative potential of generative AI in enhancing patient care, improving operational efficiency in the clinical service and research in ophthalmology. It also advocates for a balanced approach to its adoption.

Cybersecurity in the generative artificial intelligence era.

Generative Artificial Intelligence (GenAI) are algorithms capable of generating original content. The ability of GenAI to learn and generate novel outputs alike human cognition has taken the world by storm and ushered in a new era. In this review, we explore the role of GenAI in healthcare, including clinical, operational, and research applications, and delve into the cybersecurity risks of this technology. We discuss risks such as data privacy risks, data poisoning attacks, the propagation of bias, and hallucinations. In this review, we recommend risk mitigation strategies to enhance cybersecurity in GenAI technologies and further explore the use of GenAI as a tool in itself to enhance cybersecurity across the various AI algorithms. GenAI is emerging as a pivotal catalyst across various industries including the healthcare domain. Comprehending the intricacies of this technology and its potential risks will be imperative for us to fully capitalise on the benefits that GenAI can bring.

Digital tools in heart failure: addressing unmet needs.

This Series paper provides an overview of digital tools in heart failure care, encompassing screening, early diagnosis, treatment initiation and optimisation, and monitoring, and the implications these tools could have for research. The current medical environment favours the implementation of digital tools in heart failure due to rapid advancements in technology and computing power, unprecedented global connectivity, and the paradigm shift towards digitisation. Despite available effective therapies for heart failure, substantial inadequacies in managing the condition have hindered improvements in patient outcomes, particularly in low-income and middle-income countries. As digital health tools continue to evolve and exert a growing influence on both clinical care and research, establishing clinical frameworks and supportive ecosystems that enable their effective use on a global scale is crucial.

Comparison of the Quality of Discharge Letters Written by Large Language Models and Junior Clinicians: Single-Blinded Study.

BACKGROUND: Discharge letters are a critical component in the continuity of care between specialists and primary care providers. However, these letters are time-consuming to write, underprioritized in comparison to direct clinical care, and are often tasked to junior doctors. Prior studies assessing the quality of discharge summaries written for inpatient hospital admissions show inadequacies in many domains. Large language models such as GPT have the ability to summarize large volumes of unstructured free text such as electronic medical records and have the potential to automate such tasks, providing time savings and consistency in quality. OBJECTIVE: The aim of this study was to assess the performance of GPT-4 in generating discharge letters written from urology specialist outpatient clinics to primary care providers and to compare their quality against letters written by junior clinicians. METHODS: Fictional electronic records were written by physicians simulating 5 common urology outpatient cases with long-term follow-up. Records comprised simulated consultation notes, referral letters and replies, and relevant discharge summaries from inpatient admissions. GPT-4 was tasked to write discharge letters for these cases with a specified target audience of primary care providers who would be continuing the patient's care. Prompts were written for safety, content, and style. Concurrently, junior clinicians were provided with the same case records and instructional prompts. GPT-4 output was assessed for instances of hallucination. A blinded panel of primary care physicians then evaluated the letters using a standardized questionnaire tool. RESULTS: GPT-4 outperformed human counterparts in information provision (mean 4.32, SD 0.95 vs 3.70, SD 1.27; P=.03) and had no instances of hallucination. There were no statistically significant differences in the mean clarity (4.16, SD 0.95 vs 3.68, SD 1.24; P=.12), collegiality (4.36, SD 1.00 vs 3.84, SD 1.22; P=.05), conciseness (3.60, SD 1.12 vs 3.64, SD 1.27; P=.71), follow-up recommendations (4.16, SD 1.03 vs 3.72, SD 1.13; P=.08), and overall satisfaction (3.96, SD 1.14 vs 3.62, SD 1.34; P=.36) between the letters generated by GPT-4 and humans, respectively. CONCLUSIONS: Discharge letters written by GPT-4 had equivalent quality to those written by junior clinicians, without any hallucinations. This study provides a proof of concept that large language models can be useful and safe tools in clinical documentation.

Leveraging a clinical emergency department dataset to estimate two-dose COVID-19 vaccine effectiveness and duration of protection in Canada.

BACKGROUND: During the COVID-19 pandemic, clinical care shifted toward virtual and Emergency Department care. We explored the feasibility of mRNA vaccine effectiveness (VE) estimation against SARS-CoV-2-related Emergency Department visits and hospitalizations using prospectively collected Emergency Department data. METHODS: We estimated two-dose VE using a test-negative design and data from 10 participating sites of the Canadian COVID-19 Emergency Department Rapid Response Network (CCEDRRN). We included Emergency Department patients presenting with COVID-19 symptoms and nucleic acid amplification testing for SARS-CoV-2 between July 19 and December 31, 2021. We excluded patients with unclear vaccination and one or more than 2 vaccine doses by their Emergency Department visit. RESULTS: Among 3,405 eligible patients, adjusted two-dose mRNA VE against SARS-CoV-2-related Emergency Department visits was 93.3 % (95 % CI 87.9-96.3 %) between 7-55 days, sustained over 80 % through 139 days post-vaccination. In stratified analyses, VE was similar among patients with select immune-compromising conditions, chronic kidney disease, lung disease, unstable housing, and reported illicit substance use. CONCLUSIONS: Two-dose mRNA VE against SARS-CoV-2-related Emergency Department visit was high and sustained, including among vulnerable subgroups. Compared to administrative datasets, active Emergency Department enrolment enables standardization for testing access and indication and supports separate VE assessment among special population subgroups. Compared to other active enrolment settings, Emergency Departments more consistently function during crises when alternate healthcare sectors become variably closed. TRIAL REGISTRATION: Clinicaltrials.gov, NCT0470294.

Novel artificial intelligence algorithms for diabetic retinopathy and diabetic macular edema.

BACKGROUND: Diabetic retinopathy (DR) and diabetic macular edema (DME) are major causes of visual impairment that challenge global vision health. New strategies are needed to tackle these growing global health problems, and the integration of artificial intelligence (AI) into ophthalmology has the potential to revolutionize DR and DME management to meet these challenges. MAIN TEXT: This review discusses the latest AI-driven methodologies in the context of DR and DME in terms of disease identification, patient-specific disease profiling, and short-term and long-term management. This includes current screening and diagnostic systems and their real-world implementation, lesion detection and analysis, disease progression prediction, and treatment response models. It also highlights the technical advancements that have been made in these areas. Despite these advancements, there are obstacles to the widespread adoption of these technologies in clinical settings, including regulatory and privacy concerns, the need for extensive validation, and integration with existing healthcare systems. We also explore the disparity between the potential of AI models and their actual effectiveness in real-world applications. CONCLUSION: AI has the potential to revolutionize the management of DR and DME, offering more efficient and precise tools for healthcare professionals. However, overcoming challenges in deployment, regulatory compliance, and patient privacy is essential for these technologies to realize their full potential. Future research should aim to bridge the gap between technological innovation and clinical application, ensuring AI tools integrate seamlessly into healthcare workflows to enhance patient outcomes.

Clinical performance of automated machine learning: A systematic review.

Introduction: Automated machine learning (autoML) removes technical and technological barriers to building artificial intelligence models. We aimed to summarise the clinical applications of autoML, assess the capabilities of utilised platforms, evaluate the quality of the evidence trialling autoML, and gauge the performance of autoML platforms relative to conventionally developed models, as well as each other. Method: This review adhered to a prospectively registered protocol (PROSPERO identifier CRD42022344427). The Cochrane Library, Embase, MEDLINE and Scopus were searched from inception to 11 July 2022. Two researchers screened abstracts and full texts, extracted data and conducted quality assessment. Disagreement was resolved through discussion and if required, arbitration by a third researcher. Results: There were 26 distinct autoML platforms featured in 82 studies. Brain and lung disease were the most common fields of study of 22 specialties. AutoML exhibited variable performance: area under the receiver operator characteristic curve (AUCROC) 0.35-1.00, F1-score 0.16-0.99, area under the precision-recall curve (AUPRC) 0.51-1.00. AutoML exhibited the highest AUCROC in 75.6% trials; the highest F1-score in 42.3% trials; and the highest AUPRC in 83.3% trials. In autoML platform comparisons, AutoPrognosis and Amazon Rekognition performed strongest with unstructured and structured data, respectively. Quality of reporting was poor, with a median DECIDE-AI score of 14 of 27. Conclusion: A myriad of autoML platforms have been applied in a variety of clinical contexts. The performance of autoML compares well to bespoke computational and clinical benchmarks. Further work is required to improve the quality of validation studies. AutoML may facilitate a transition to data-centric development, and integration with large language models may enable AI to build itself to fulfil user-defined goals.

Application of deep learning algorithms in classification and localization of implant cutout for the postoperative hip.

OBJECTIVE: This study aims to explore the feasibility of employing convolutional neural networks for detecting and localizing implant cutouts on anteroposterior pelvic radiographs. MATERIALS AND METHODS: The research involves the development of two Deep Learning models. Initially, a model was created for image-level classification of implant cutouts using 40191 pelvic radiographs obtained from a single institution. The radiographs were partitioned into training, validation, and hold-out test datasets in a 6/2/2 ratio. Performance metrics including the area under the receiver operator characteristics curve (AUROC), sensitivity, and specificity were calculated using the test dataset. Additionally, a second object detection model was trained to localize implant cutouts within the same dataset. Bounding box visualizations were generated on images predicted as cutout-positive by the classification model in the test dataset, serving as an adjunct for assessing algorithm validity. RESULTS: The classification model had an accuracy of 99.7%, sensitivity of 84.6%, specificity of 99.8%, AUROC of 0.998 (95% CI: 0.996, 0.999) and AUPRC of 0.774 (95% CI: 0.646, 0.880). From the pelvic radiographs predicted as cutout-positive, the object detection model could achieve 95.5% localization accuracy on true positive images, but falsely generated 14 results from the 15 false-positive predictions. CONCLUSION: The classification model showed fair accuracy for detection of implant cutouts, while the object detection model effectively localized cutout. This serves as proof of concept of using a deep learning-based approach for classification and localization of implant cutouts from pelvic radiographs.

Optical coherence tomography choroidal enhancement using generative deep learning.

Spectral-domain optical coherence tomography (SDOCT) is the gold standard of imaging the eye in clinics. Penetration depth with such devices is, however, limited and visualization of the choroid, which is essential for diagnosing chorioretinal disease, remains limited. Whereas swept-source OCT (SSOCT) devices allow for visualization of the choroid these instruments are expensive and availability in praxis is limited. We present an artificial intelligence (AI)-based solution to enhance the visualization of the choroid in OCT scans and allow for quantitative measurements of choroidal metrics using generative deep learning (DL). Synthetically enhanced SDOCT B-scans with improved choroidal visibility were generated, leveraging matching images to learn deep anatomical features during the training. Using a single-center tertiary eye care institution cohort comprising a total of 362 SDOCT-SSOCT paired subjects, we trained our model with 150,784 images from 410 healthy, 192 glaucoma, and 133 diabetic retinopathy eyes. An independent external test dataset of 37,376 images from 146 eyes was deployed to assess the authenticity and quality of the synthetically enhanced SDOCT images. Experts' ability to differentiate real versus synthetic images was poor (47.5% accuracy). Measurements of choroidal thickness, area, volume, and vascularity index, from the reference SSOCT and synthetically enhanced SDOCT, showed high Pearson's correlations of 0.97 [95% CI: 0.96-0.98], 0.97 [0.95-0.98], 0.95 [0.92-0.98], and 0.87 [0.83-0.91], with intra-class correlation values of 0.99 [0.98-0.99], 0.98 [0.98-0.99], and 0.95 [0.96-0.98], 0.93 [0.91-0.95], respectively. Thus, our DL generative model successfully generated realistic enhanced SDOCT data that is indistinguishable from SSOCT images providing improved visualization of the choroid. This technology enabled accurate measurements of choroidal metrics previously limited by the imaging depth constraints of SDOCT. The findings open new possibilities for utilizing affordable SDOCT devices in studying the choroid in both healthy and pathological conditions.

A comparative study of an on premise AutoML solution for medical image classification.

Automated machine learning (AutoML) allows for the simplified application of machine learning to real-world problems, by the implicit handling of necessary steps such as data pre-processing, feature engineering, model selection and hyperparameter optimization. This has encouraged its use in medical applications such as imaging. However, the impact of common parameter choices such as the number of trials allowed, and the resolution of the input images, has not been comprehensively explored in existing literature. We therefore benchmark AutoKeras (AK), an open-source AutoML framework, against several bespoke deep learning architectures, on five public medical datasets representing a wide range of imaging modalities. It was found that AK could outperform the bespoke models in general, although at the cost of increased training time. Moreover, our experiments suggest that a large number of trials and higher resolutions may not be necessary for optimal performance to be achieved.

Replacing the postoperative week 1 visit after routine phacoemulsification with a telephone consult.

OBJECTIVE: To assess the safety of replacing the postoperative week 1 (POW1) clinic visit with a nurse-conducted telephone call. DESIGN: Retrospective observational study that included cases from January 2019 to June 2021. PARTICIPANTS: Patients who had undergone uncomplicated phacoemulsification surgery with an unremarkable postoperative day 1 (POD1) examination. METHODS: All patients were seen in clinic on POD1 by an ophthalmologist. They then had a telephone conversation with a nurse at POW1 and subsequently an in-person postoperative month 1 (POM1) clinic consultation with an ophthalmologist. Main outcome measure was the incidence of unexpected management changes related to cataract surgery within POM1. Data also were collected on the reasons for unscheduled patient-initiated visits, additional procedures or medications, and postoperative visual acuity worse than 6/12 at POM1. RESULTS: Of the 20,475 patients, 541 patients (2.64%) had an unexpected management change within POM1. There were 565 patients (2.76%) who had self-initiated unscheduled visits between POD1 to POM1. There were 23 patients (0.11%) who required additional surgery within POM1 and 1 patient (0.005%) with endophthalmitis. The most common indication for additional surgical procedures was retained lens material (7 patients, 30.43%). Visual acuity was worse than 6/12 in 1,199 patients (6.22%), with the most common causes attributed to preexisting ocular conditions. CONCLUSIONS: These results suggest that replacing the POW1 visit with a nurse-conducted telephone consult for patients who have undergone uncomplicated phacoemulsification surgery and had a normal POD1 consultation is safe.

Large language models approach expert-level clinical knowledge and reasoning in ophthalmology: A head-to-head cross-sectional study.

Large language models (LLMs) underlie remarkable recent advanced in natural language processing, and they are beginning to be applied in clinical contexts. We aimed to evaluate the clinical potential of state-of-the-art LLMs in ophthalmology using a more robust benchmark than raw examination scores. We trialled GPT-3.5 and GPT-4 on 347 ophthalmology questions before GPT-3.5, GPT-4, PaLM 2, LLaMA, expert ophthalmologists, and doctors in training were trialled on a mock examination of 87 questions. Performance was analysed with respect to question subject and type (first order recall and higher order reasoning). Masked ophthalmologists graded the accuracy, relevance, and overall preference of GPT-3.5 and GPT-4 responses to the same questions. The performance of GPT-4 (69%) was superior to GPT-3.5 (48%), LLaMA (32%), and PaLM 2 (56%). GPT-4 compared favourably with expert ophthalmologists (median 76%, range 64-90%), ophthalmology trainees (median 59%, range 57-63%), and unspecialised junior doctors (median 43%, range 41-44%). Low agreement between LLMs and doctors reflected idiosyncratic differences in knowledge and reasoning with overall consistency across subjects and types (p>0.05). All ophthalmologists preferred GPT-4 responses over GPT-3.5 and rated the accuracy and relevance of GPT-4 as higher (p<0.05). LLMs are approaching expert-level knowledge and reasoning skills in ophthalmology. In view of the comparable or superior performance to trainee-grade ophthalmologists and unspecialised junior doctors, state-of-the-art LLMs such as GPT-4 may provide useful medical advice and assistance where access to expert ophthalmologists is limited. Clinical benchmarks provide useful assays of LLM capabilities in healthcare before clinical trials can be designed and conducted.

Cost-effectiveness and cost-utility of a digital technology-driven hierarchical healthcare screening pattern in China.

Utilization of digital technologies for cataract screening in primary care is a potential solution for addressing the dilemma between the growing aging population and unequally distributed resources. Here, we propose a digital technology-driven hierarchical screening (DH screening) pattern implemented in China to promote the equity and accessibility of healthcare. It consists of home-based mobile artificial intelligence (AI) screening, community-based AI diagnosis, and referral to hospitals. We utilize decision-analytic Markov models to evaluate the cost-effectiveness and cost-utility of different cataract screening strategies (no screening, telescreening, AI screening and DH screening). A simulated cohort of 100,000 individuals from age 50 is built through a total of 30 1-year Markov cycles. The primary outcomes are incremental cost-effectiveness ratio and incremental cost-utility ratio. The results show that DH screening dominates no screening, telescreening and AI screening in urban and rural China. Annual DH screening emerges as the most economically effective strategy with 341 (338 to 344) and 1326 (1312 to 1340) years of blindness avoided compared with telescreening, and 37 (35 to 39) and 140 (131 to 148) years compared with AI screening in urban and rural settings, respectively. The findings remain robust across all sensitivity analyses conducted. Here, we report that DH screening is cost-effective in urban and rural China, and the annual screening proves to be the most cost-effective option, providing an economic rationale for policymakers promoting public eye health in low- and middle-income countries.

Ethical and regulatory challenges of large language models in medicine.

With the rapid growth of interest in and use of large language models (LLMs) across various industries, we are facing some crucial and profound ethical concerns, especially in the medical field. The unique technical architecture and purported emergent abilities of LLMs differentiate them substantially from other artificial intelligence (AI) models and natural language processing techniques used, necessitating a nuanced understanding of LLM ethics. In this Viewpoint, we highlight ethical concerns stemming from the perspectives of users, developers, and regulators, notably focusing on data privacy and rights of use, data provenance, intellectual property contamination, and broad applications and plasticity of LLMs. A comprehensive framework and mitigating strategies will be imperative for the responsible integration of LLMs into medical practice, ensuring alignment with ethical principles and safeguarding against potential societal risks.

Deep learning for detection of Fuchs endothelial dystrophy from widefield specular microscopy imaging: a pilot study.

BACKGROUND: To describe the diagnostic performance of a deep learning (DL) algorithm in detecting Fuchs endothelial corneal dystrophy (FECD) based on specular microscopy (SM) and to reliably detect widefield peripheral SM images with an endothelial cell density (ECD) > 1000 cells/mm2. METHODS: Five hundred and forty-seven subjects had SM imaging performed for the central cornea endothelium. One hundred and seventy-three images had FECD, while 602 images had other diagnoses. Using fivefold cross-validation on the dataset containing 775 central SM images combined with ECD, coefficient of variation (CV) and hexagonal endothelial cell ratio (HEX), the first DL model was trained to discriminate FECD from other images and was further tested on an external set of 180 images. In eyes with FECD, a separate DL model was trained with 753 central/paracentral SM images to detect SM with ECD > 1000 cells/mm2 and tested on 557 peripheral SM images. Area under curve (AUC), sensitivity and specificity were evaluated. RESULTS: The first model achieved an AUC of 0.96 with 0.91 sensitivity and 0.91 specificity in detecting FECD from other images. With an external validation set, the model achieved an AUC of 0.77, with a sensitivity of 0.69 and specificity of 0.68 in differentiating FECD from other diagnoses. The second model achieved an AUC of 0.88 with 0.79 sensitivity and 0.78 specificity in detecting peripheral SM images with ECD > 1000 cells/mm2. CONCLUSIONS: Our pilot study developed a DL model that could reliably detect FECD from other SM images and identify widefield SM images with ECD > 1000 cells/mm2 in eyes with FECD. This could be the foundation for future DL models to track progression of eyes with FECD and identify candidates suitable for therapies such as Descemet stripping only.