Development and Evaluation of Aeyeconsult: A Novel Ophthalmology Chatbot Leveraging Verified Textbook Knowledge and GPT-4.

OBJECTIVE: There has been much excitement on the use of large language models (LLMs) such as ChatGPT in ophthalmology. However, LLMs are limited in that they are trained on unverified information and do not cite their sources. This paper highlights a new methodology to create a generative AI chatbot to answer eye care related questions which uses only verified ophthalmology textbooks as data and cites its sources. SETTING: Yale School of Medicine Department of Ophthalmology and Visual Science. DESIGN/METHODS: Aeyeconsult, an ophthalmology chatbot, was developed using GPT-4 (the LLM used to power the publicly available chatbot ChatGPT-4), LangChain, and Pinecone. Ophthalmology textbooks were processed into embeddings and stored in Pinecone. User queries were similarly converted, compared to stored embeddings, and GPT-4 generated responses. The interface was adapted from public code. Both Aeyeconsult and ChatGPT-4 were tested on the same 260 questions from OphthoQuestions.com, with the first response from Aeyeconsult and ChatGPT-4 recorded as the answer. RESULTS: Aeyeconsult outperformed ChatGPT-4 on the OKAP dataset, with 83.4% correct answers compared to 69.2% (p = 0.0118). Aeyeconsult also had fewer instances of no answer and multiple answers. Both systems performed best in General Medicine, with Aeyeconsult achieving 96.2% accuracy. Aeyeconsult's weakest performance was in Clinical Optics at 68.1%, but it still outperformed ChatGPT-4 in this category (45.5%). CONCLUSION: LLMs may be useful in answering ophthalmology questions but their trustworthiness and accuracy is limited due to training on unverified internet data and lack of source citation. We used a new methodology, using verified ophthalmology textbooks as source material and providing citations, to mitigate these issues, resulting in a chatbot more accurate than ChatGPT-4 in answering OKAPs style questions.

Retinal Ischemic Perivascular Lesions in Individuals With Atrial Fibrillation.

Background We previously demonstrated that retinal ischemic perivascular lesions (RIPLs), which are indicative of ischemia in the middle retina, may be a biomarker of ischemic cardiovascular disease. In this study, we sought to determine the relationship between RIPLs and atrial fibrillation, a common source of cardiac emboli. Methods and Results In this case-control study, we identified individuals between the ages of 50 and 90 years who had undergone macular spectral domain optical coherence tomography imaging. Individuals with atrial fibrillation were identified, and age- and sex-matched individuals from the same pool, but without a diagnosis of atrial fibrillation, were selected as controls. Spectral domain optical coherence tomography scans were reviewed by 3 independent and masked observers for presence of RIPLs. The relationship between RIPLs and atrial fibrillation was analyzed using multivariable logistic regression models. There were 106 and 91 subjects with and without atrial fibrillation, respectively. The percentage of subjects with RIPLs was higher in the atrial fibrillation group compared with the control group (57.5% versus 37.4%; P=0.005). After adjusting for age, sex, smoking history, hypertension, diabetes, coronary artery disease, carotid stenosis, stroke, and myocardial infarction, the presence of RIPLs was significantly associated with atrial fibrillation, with an odds ratio of 1.91 (95% CI, 1.01-3.59). Conclusions RIPLs are significantly associated with atrial fibrillation, independent of underlying ischemic heart disease or cardiovascular risk factors. This association may inform the diagnostic cardiovascular workup for individuals with RIPLs incidentally detected on optical coherence tomography scan of the macula.

Retinotopic degeneration of the retina and optic tracts in autosomal dominant Alzheimer's disease.

INTRODUCTION: We investigated the correlation between retinal thickness and optic tract integrity in subjects with autosomal dominant Alzheimer's disease (ADAD) causing mutations. METHODS: Retinal thicknesses and diffusion tensor images (DTI) were obtained using optical coherence tomography and magnetic resonance imaging, respectively. The association between retinal thickness and DTI measures was adjusted for age, sex, retinotopy, and correlation between eyes. RESULTS: Optic tract mean diffusivity and axial diffusivity were negatively correlated with retinotopically defined ganglion cell inner plexiform thickness (GCIPL). Fractional anisotropy was negatively correlated with retinotopically defined retinal nerve fiber layer thickness. There was no correlation between outer nuclear layer (ONL) thickness and any DTI measure. DISCUSSION: In ADAD, GCIPL thickness is significantly associated with retinotopic optic tract DTI measures even in minimally symptomatic subjects. Similar associations were not present with ONL thickness or when ignoring retinotopy. We provide in vivo evidence for optic tract changes resulting from ganglion cell pathology in ADAD.

Characterization of spastic paraplegia in a family with a novel PSEN1 mutation.

Spastic paraparesis has been described to occur in 13.7% of PSEN1 mutations and can be the presenting feature in 7.5%. In this paper, we describe a family with a particularly young onset of spastic paraparesis due to a novel mutation in PSEN1 (F388S). Three affected brothers underwent comprehensive imaging protocols, two underwent ophthalmological evaluations and one underwent neuropathological examination after his death at age 29. Age of onset was consistently at age 23 with spastic paraparesis, dysarthria and bradyphrenia. Pseudobulbar affect followed with progressive gait problems leading to loss of ambulation in the late 20s. Cerebrospinal fluid levels of amyloid-ß, tau and phosphorylated tau and florbetaben PET were consistent with Alzheimer's disease. Flortaucipir PET showed an uptake pattern atypical for Alzheimer's disease, with disproportionate signal in posterior brain areas. Diffusion tensor imaging showed decreased mean diffusivity in widespread areas of white matter but particularly in areas underlying the peri-Rolandic cortex and in the corticospinal tracts. These changes were more severe than those found in carriers of another PSEN1 mutation, which can cause spastic paraparesis at a later age (A431E), which were in turn more severe than among persons carrying autosomal dominant Alzheimer's disease mutations not causing spastic paraparesis. Neuropathological examination confirmed the presence of cotton wool plaques previously described in association with spastic parapresis and pallor and microgliosis in the corticospinal tract with severe amyloid-ß pathology in motor cortex but without unequivocal disproportionate neuronal loss or tau pathology. In vitro modelling of the effects of the mutation demonstrated increased production of longer length amyloid-ß peptides relative to shorter that predicted the young age of onset. In this paper, we provide imaging and neuropathological characterization of an extreme form of spastic paraparesis occurring in association with autosomal dominant Alzheimer's disease, demonstrating robust diffusion and pathological abnormalities in white matter. That the amyloid-ß profiles produced predicted the young age of onset suggests an amyloid-driven aetiology though the link between this and the white matter pathology remains undefined.

Abnormal retinal capillary blood flow in autosomal dominant Alzheimer's disease.

INTRODUCTION: This study characterizes retinal capillary blood flow in subjects with autosomal dominant Alzheimer's disease (ADAD)-causing mutations. METHODS: Carriers of PSEN1 or APP mutations were prospectively recruited and split into early-stage (ES) and late-stage (LS) groups. Controls were normal subjects and non-carriers from the at-risk group. Capillary blood flow was quantified using an optical coherence tomography angiography-based measure of erythrocyte flux through capillary segments. Statistical analyses were adjusted for correlation between two eyes of the same subject. RESULTS: ES carriers had significantly greater capillary blood flow than controls and LS carriers. ES and LS carriers had significantly greater capillary blood flow heterogeneity than controls. There was no difference between capillary blood flow of LS carriers and controls. DISCUSSION: ES ADAD carriers demonstrate increased retinal capillary blood flow and flow heterogeneity compared to controls. These findings support the hypothesis that increased perfusion is a pathophysiologic feature of presymptomatic stages of ADAD.

Past, present and future role of retinal imaging in neurodegenerative disease.

Retinal imaging technology is rapidly advancing and can provide ever-increasing amounts of information about the structure, function and molecular composition of retinal tissue in humans in vivo. Most importantly, this information can be obtained rapidly, non-invasively and in many cases using Food and Drug Administration-approved devices that are commercially available. Technologies such as optical coherence tomography have dramatically changed our understanding of retinal disease and in many cases have significantly improved their clinical management. Since the retina is an extension of the brain and shares a common embryological origin with the central nervous system, there has also been intense interest in leveraging the expanding armamentarium of retinal imaging technology to understand, diagnose and monitor neurological diseases. This is particularly appealing because of the high spatial resolution, relatively low-cost and wide availability of retinal imaging modalities such as fundus photography or OCT compared to brain imaging modalities such as magnetic resonance imaging or positron emission tomography. The purpose of this article is to review and synthesize current research about retinal imaging in neurodegenerative disease by providing examples from the literature and elaborating on limitations, challenges and future directions. We begin by providing a general background of the most relevant retinal imaging modalities to ensure that the reader has a foundation on which to understand the clinical studies that are subsequently discussed. We then review the application and results of retinal imaging methodologies to several prevalent neurodegenerative diseases where extensive work has been done including sporadic late onset Alzheimer's Disease, Parkinson's Disease and Huntington's Disease. We also discuss Autosomal Dominant Alzheimer's Disease and cerebrovascular small vessel disease, where the application of retinal imaging holds promise but data is currently scarce. Although cerebrovascular disease is not generally considered a neurodegenerative process, it is both a confounder and contributor to neurodegenerative disease processes that requires more attention. Finally, we discuss ongoing efforts to overcome the limitations in the field and unmet clinical and scientific needs.

Impaired layer specific retinal vascular reactivity among diabetic subjects.

PURPOSE: To investigate layer specific retinal vascular reactivity (RVR) in capillaries of diabetic subjects without DR or with only mild non-proliferative diabetic retinopathy (NPDR). METHODS: A previously described nonrebreathing apparatus was used to deliver room air, 5% CO2, or 100% O2 to 41 controls and 22 diabetic subjects (with mild or no NPDR) while simultaneously acquiring fovea-centered 3x3mm2 Swept-Source Optical Coherence Tomography Angiography (SS-OCTA) images. Vessel skeleton density (VSD) and vessel diameter index (VDI) were calculated for each gas condition for the superficial retinal layer (SRL) and deep retinal layer (DRL). The superficial layer analysis excluded arterioles and venules. Data analysis was performed using mixed factorial analysis of covariance stratified by diabetic status. All models were adjusted for age, gender, and hypertension, and statistical significance for multiple comparisons from posthoc comparisons were defined at p<0.017. RESULTS: Among controls, there was a significant difference in capillary VSD between all gas conditions (p<0.001). This difference was present in both the SRL and DRL. Among diabetics, there was no significant difference in response to CO2 conditions in the SRL (p = 0.072), and a blunted response to both CO2 (p = 0.9) and O2 in the DRL (p = 0.019). A significant gas effect was detected in the capillary VDI in the SRL of controls (p = 0.001), which was driven by higher VDI in the oxygen condition compared to that of carbon dioxide. CONCLUSIONS: Impairment in RVR in diabetic subjects is characterized by a paradoxical response to CO2 in both the SRL and DRL as well as an attenuated response to O2 in the DRL. These layer and gas specific impairments in diabetics seem to occur early in the disease and to be driven primarily at the capillary level.