About

Benjamin J. Kim, MD, is an Associate Professor of Ophthalmology at the Hospital of the University of Pennsylvania and an Attending Ophthalmologist at the Philadelphia Veterans Administration Hospital. He is a member of the Penn Medicine Neuroscience Center and serves as Vice-Chair for Clinical Operations at the Scheie Eye Institute within the Department of Ophthalmology at the University of Pennsylvania. His clinical expertise includes age-related macular degeneration, diabetic retinopathy, retinal detachment, macular hole, and macular pucker. His research focuses on age-related macular degeneration, diabetic retinopathy, geographic atrophy, choroidal neovascularization, retinal neovascularization, clinical trials, and retinal imaging.

Research topics

• Medicine
• Biology
• Genetics
• Neuroscience
• Immunology
• Pediatrics
• Surgery
• Internal medicine
• Ophthalmology
• Bioinformatics

Selected publications

• Association of diabetes severity with cognitive function in US adults: a cross-sectional analysis of the AI-READI multicentre cohort

  BMJ Open · 2026-03-01

  articleOpen access

  OBJECTIVES: To evaluate whether type 2 diabetes mellitus (T2DM) presence and severity are associated with differences in global and domain-specific cognitive function among US adults, using standardised Montreal Cognitive Assessment (MoCA) testing. DESIGN: Cross-sectional study SETTING: Three U.S academic medical centres participating in the Artificial Intelligence-Ready and Equitable Atlas for Diabetes Insights (AI-READI) study. PARTICIPANTS: Adults aged ≥40 years enrolled in the AI-READI cross-sectional study at three US academic medical centres were eligible. The study excluded individuals with type 1 diabetes, pregnancy or inability to speak, read and understand English. For this secondary analysis, 1067 participants from the first publicly released AI-READI data set who had MoCA data and assigned glycaemic status were included. Participants were classified into four prespecified glycaemic groups: controls without diabetes (n=371), pre-diabetes (n=239), medication-controlled type 2 diabetes (n=323), and insulin-dependent type 2 diabetes (n=129). PRIMARY AND SECONDARY OUTCOME MEASURES: The primary outcome was global cognitive function measured by the MoCA total score. Secondary outcomes included MoCA domain scores and the prevalence of cognitive impairment, defined as MoCA<26. RESULTS: Significant differences in MoCA total scores were observed across glycaemic groups (p<0.001), with the lowest mean score (24.0) among insulin-dependent individuals and highest among controls (25.8), and the significant difference remained in multivariable analyses (p=0.02). Among MoCA subdomains, mean abstraction scores were significantly lower (1.76) in insulin-dependent T2DM group than other glycaemic groups (mean score 1.87-1.89) in multivariable analysis (p=0.03). The prevalence of cognitive impairment increased from 39% in controls to 58% in the insulin-dependent group. In multivariable analysis, increasing diabetes severity was associated with higher risk of cognitive impairment with adjusted OR 1.85 (95% CI: 1.11 to 3.07) for insulin-dependent T2DM and 1.50 (95% CI: 1.05 to 2.15) for medication-controlled T2DM. CONCLUSIONS: Individuals with more advanced T2DM, particularly those on insulin, had significantly higher risk of cognitive impairment. These findings support routine cognitive screening in patients with T2DM, especially those on insulin therapy. Early identification of cognitive impairment may improve diabetes management and cognitive outcomes.

  PublisherOA PDFDOI

• Eventual Atrophy-Associated Vision Loss from a Choroidal Nevus: 13 Years of Follow-Up

  Ophthalmology Retina · 2026-04-01

  articleSenior author
  PublisherDOI

• MULTICENTER STUDY OF FACTORS ASSOCIATED WITH VISUAL AND ANATOMIC OUTCOMES OF SUPRACHOROIDAL HEMORRHAGE

  Retina · 2025-09-09

  articleSenior author

  PURPOSE: To report outcomes of suprachoroidal hemorrhage (SCH). METHODS: Retrospective nonrandomized study of eyes with SCH from two sites (January 1, 2013-December 31, 2022). The primary outcome was the 6-month change in visual acuity (VA). Multivariable analysis was performed, as well as a comparison of matched eyes with and without systemic steroids. RESULTS: Overall, 143 eyes of 143 patients (mean age 70.8 years, 52.4% male) were included, with 72 perioperative, 24 traumatic, and 47 spontaneous SCH cases. The mean (SD) presenting VA was 2.07 (0.92) logMAR. 87 (60.8%) were managed nonsurgically, 24 (16.8%) underwent drainage, and 32 (22.4%) underwent drainage and vitrectomy; 36 (25.2%) received systemic steroids. At 6 months, the mean (SD) change in VA from presentation was -0.41 (0.84) logMAR. 102 eyes (71.3%) achieved anatomic success (complete retinal attachment). Concurrent retinal detachment was associated with worse VA change and anatomic success in multivariable analysis ( P < 0.05). In the matching analysis, eyes receiving systemic steroids were more likely to achieve ≥ 3-line gain in VA than matched eyes without systemic steroids (77.8% vs. 55.3%, P = 0.047). CONCLUSION: The visual prognosis of eyes with SCH remains guarded. Systemic steroids may be associated with a modest benefit for visual outcomes. Concurrent retinal detachment portends worse outcomes.

  PublisherDOI

• Characteristics of Study Design and Statistical Analysis in OCT-Based Studies of Neurodegeneration

  Ophthalmology Science · 2025-10-09

  articleOpen accessSenior author

  Purpose: To evaluate the design and statistical analysis of recently published clinical studies of retina OCT of neurodegenerative diseases. Design: Review of 134 publications. Methods: Clinical and translational publications from 2013-2023 involving human OCT studies were reviewed. Publications were restricted to the top 10 journals of either ophthalmology or neurology based on the SCImago Journal Rank. Data were extracted and analyzed regarding study design characteristics, disease studied, imaging characteristics, and statistical analysis method. Main Outcome Measures: Characteristics of study design and statistical analysis. Results: Of the 134 studies, the most commonly investigated diseases were multiple sclerosis (47.8%), Alzheimer disease (31.3%), and Parkinson disease (14.2%). One hundred three (76.9%) studies were cross-sectional, and 31 (23.1%) studies were longitudinal. Of the 42 Alzheimer disease studies, 19.0% used cerebrospinal fluid biomarkers, 26.2% used positron emission tomography imaging, and 4.8% used genetic data to confirm the diagnosis. The peripapillary retinal nerve fiber layer thickness (95 studies) and macular retina layer thickness (96 studies) were the most commonly studied OCT measurements; 69 studies looked at both of these measurements. For 25 (18.7%) of the studies, no statement of image quality criteria or exclusion of subjects or eyes because of image quality criteria was identified. One hundred seventeen (87.3%) studies performed bilateral eye imaging, but only 89 (66.4%) studies performed bilateral image analysis. Of the 89 studies with bilateral eye image analysis, 53 (59.6%) performed analysis at the eye level by combining data from the left and right eye, making each eye (rather than each subject) the unit of analysis. Of the 53 studies performing analysis at the eye level for data from both eyes, the intereye correlation was adjusted in 38 (71.7%) studies, while 15 (28.3%) studies did not account for the intereye correlation. Thirty-seven (27.6%) studies corrected for multiple comparisons. Conclusion: There may be opportunities for improvement in the study design and statistical analysis of OCT studies of neurodegenerative diseases. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

  PublisherOA PDFDOI

• Evaluation of ChatGPT-4 in detecting referable diabetic retinopathy using single fundus images

  AJO International · 2025-03-19 · 5 citations

  articleOpen access

  • ChatGPT-4 is unable to adequately read and interpret normal and mild NPDR images limiting utility in low-risk settings. • ChatGPT-4 shows high sensitivity for identifying referable diabetic retinopathy but lacks specificity. • ChatGPT-4 needs improvement in multimodal image fund of knowledge in order to improve real-world applicability. Evaluate ChatGPT-4′s ability to identify referable diabetic retinopathy (DR) from single fundus images. A cross-sectional study comparing ChatGPT-4′s versus retina specialists’ identification of more than mild DR (mtmDR) and vision-threatening DR (VTDR). Images in equal proportions of normal, mild, moderate, and severe nonproliferative DR (NPDR), proliferative DR (PDR), and blurry images with and without suspected PDR were presented to a panel of blinded retina specialists who identified images as readable or unreadable, and potentially as mtmDR or VTDR. These images were also submitted to ChatGPT-4 three times with a standardized prompt regarding mtmDR and VTDR. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated for ChatGPT-4′s responses regarding mtmDR and VTDR as compared to the retina specialists majority determination. Retina specialists read 158/180 prompts (87.7 %) with excellent interrater reliability while ChatGPT-4 read 132/180 (73.33 %) of the image prompts. For mtmDR, ChatGPT-4 demonstrated a sensitivity of 96.2 %, specificity of 19.1 %, PPV of 69.1 %, and NPV of 72.7 %. Overall, 90.9 % of prompts read by ChatGPT-4 were labeled as mtmDR. For VTDR, ChatGPT-4 demonstrated a 63.0 % sensitivity, 62.5 % specificity, 71.9 % PPV, and 52.6 % NPV compared to retina specialists. ChatGPT-4 labeled 51.5 % of read images as VTDR. Overall referability was 66.6 % for retina specialists and 93.3 % for ChatGPT-4. While ChatGPT-4 demonstrates promise in identifying moderate-to-severe DR, its limited specificity and tendency to overcall disease reduce its current utility as a screening tool.

  PublisherDOI

• Retinal imaging and tissue analysis for frontotemporal degeneration: recent advances and challenges for biomarker development

  Journal of Neurology Neurosurgery & Psychiatry · 2025-06-06 · 1 citations

  reviewOpen accessSenior author

  Frontotemporal degeneration (FTD) is a group of neurodegenerative disorders affecting behaviour, language and executive functions. FTD is a common cause of early-onset dementia, but there are no FDA-approved treatments or established biomarkers for diagnosing and tracking these conditions, making early and accurate diagnosis challenging during life. Recent advances in retinal imaging, particularly through technologies like optical coherence tomography (OCT), have emerged as promising tools for identifying potential biomarkers for FTD and related neurodegenerative diseases. The retina, being an accessible extension of the central nervous system, has shown abnormalities that might serve as indicators of forms of FTD. Retinal imaging has revealed changes such as thinning of specific retinal layers that could correlate with molecular forms of FTD, Alzheimer's disease and other neurodegenerative diseases. These advances highlight the potential of retinal imaging to not only aid in diagnosis but also differentiate between various neurodegenerative conditions. Emerging data on retinal tissue analysis with immunohistochemistry and other techniques further support the potential of retinal biomarkers, though further studies are required to validate and refine these findings. Future advancements in retinal imaging technologies, along with longitudinal and autopsy-validated studies, are crucial for enhancing diagnostic capabilities and understanding FTD-related pathologies within the retina.

  PublisherOA PDFDOI

• Choroidal evaluation of FTLD-Tau and biomarker-determined Alzheimer’s disease

  Scientific Reports · 2025-07-01

  articleOpen access1st authorCorresponding

  Frontotemporal lobar degeneration with tauopathy (FTLD-Tau) can present clinically similar to Alzheimer's disease but lacks a biomarker. Alzheimer's disease has been associated with choroidal thinning compared to controls. We compared the choroid of 25 probable FTLD-Tau (pFTLD-Tau) patients (42 eyes), 26 biomarker-determined probable Alzheimer's disease neuropathologic change (pADNC) patients (49 eyes), and 53 normal controls (80 eyes). Cerebrospinal fluid biomarkers determined presence of ADNC. All pFTLD-Tau patients had a syndrome highly associated with FTLD-Tau. Optical coherence tomography was performed with masked manual choroidal thickness (CT) measurements. With Image J, binarized images determined the choroidal vascularity index (CVI). Linear regression with generalized estimating equations to account for inter-eye correlation was performed. For pFTLD-Tau, pADNC, and controls, the subfoveal CT was 308.9, 286.0, and 301.5 μm, and CVI was 0.72, 0.72, and 0.73, respectively (all p > 0.05 for each group comparison). Adjusting for demographics, the CT and CVI were not significantly different between groups, including 13 CT measurement locations (all p > 0.05). Among pADNC patients, an exploratory analysis found a correlation between CVI and disease duration (Pearson r = 0.32, p = 0.04). We found no significant difference of CT or CVI between pFTLD-Tau, pADNC, and controls. Additional studies are warranted to evaluate how CVI relates to ADNC.

  PublisherOA PDFDOI

• Evaluating the Application of Artificial Intelligence and Ambient Listening to Generate Medical Notes in Vitreoretinal Clinic Encounters

  Clinical ophthalmology · 2025-06-01 · 3 citations

  articleOpen access

  Purpose: Analyze the application of large language models (LLM) to listen to and generate medical documentation in vitreoretinal clinic encounters. Subjects: Two publicly available large language models, Google Gemini 1.0 Pro and Chat GPT 3.5. Methods: Patient-physician dialogues simulating vitreoretinal clinic scenarios were scripted to simulate real-world encounters and recorded for standardization. Two artificial intelligence engines were given the audio files to transcribe the dialogue and produce medical documentation of the encounters. Similarity of the dialogue and LLM transcription was assessed using an online comparability tool. A panel of practicing retina specialists evaluated each generated medical note. Main Outcome Measures: The number of discrepancies and overall similarity of LLM text compared to scripted patient-physician dialogues, and scoring on the physician documentation quality instrument-9 (PDQI-9) of each medical note by five retina specialists. Results: On average, the documentation produced by AI engines scored 81.5% of total possible points in documentation quality. Similarity between pre-formed dialogue scripts and transcribed encounters was higher for ChatGPT (96.5%) compared to Gemini (90.6%, p<0.01). The mean total PDQI-9 score among all encounters from ChatGPT 3.5 (196.2/225, 87.2%) was significantly greater than Gemini 1.0 Pro (170.4/225, 75.7%, p=0.002). Conclusion: The authors report the aptitude of two popular LLMs (ChatGPT 3.5 and Google Gemini 1.0 Pro) in generating medical notes based on audio recordings of scripted vitreoretinal clinical encounters using a validated medical documentation tool. Artificial intelligence can produce quality vitreoretinal clinic encounter medical notes after listening to patient-physician dialogues despite case complexity and missing encounter variables. The performance of these engines was satisfactory but sometimes included fabricated information. We demonstrate the potential utility of LLMs in reducing the documentation burden on physicians and potentially streamlining patient care.

  PublisherOA PDFDOI

• Incidence and risk factors of perioperative suprachoroidal hemorrhage: A systematic review and meta-analysis

  Survey of Ophthalmology · 2024-10-03 · 1 citations

  reviewOpen accessSenior author
  PublisherOA PDFDOI

• FACTORS ASSOCIATED WITH OUTCOMES OF SUPRACHOROIDAL HEMORRHAGE

  Retina · 2024-03-06 · 1 citations

  reviewOpen accessSenior author

  PURPOSE: To determine factors associated with visual and anatomic outcomes of suprachoroidal hemorrhage in studies published between 1990 and 2022. METHODS: Individual participant data systematic review. The protocol was prospectively registered on Open Science Framework ( https://osf.io/69v3q/ ). PubMed, EMBASE, Web of Science, and Google Scholar were searched for peer-reviewed studies of suprachoroidal hemorrhage with ≥3 patients published between January 1, 1990, and September 1, 2022. The primary outcome was the change in logarithm of the minimum angle of resolution visual acuity from the time of suprachoroidal hemorrhage diagnosis to last follow-up. RESULTS: Four hundred thirteen eyes from 49 studies were included, with mean (SD) age 60.8 (22.4) years and mean (SD) follow-up of 13.8 (12.6) months. Among 145 eyes with at least 6 months of follow-up, the mean (SD) gain in visual acuity was -0.98 (0.89) logarithm of the minimum angle of resolution. In multivariable regression, treatment with systemic steroids was associated with greater improvement in logarithm of the minimum angle of resolution visual acuity (adjusted mean [SE] -1.29 [0.09] vs. -0.16 [0.30] for no systemic steroids; P < 0.001) and greater odds of achieving anatomic success (adjusted OR 10.59, 95% confidence interval 2.59-43.3; P = 0.001). CONCLUSION: The use of systemic steroids was associated with better visual and anatomic outcomes for suprachoroidal hemorrhage.

  PublisherOA PDFDOI

Frequent coauthors

• Gui‐Shuang Ying

  University of Pennsylvania

  81 shared

• Anton M. Kolomeyer

  63 shared

• Tomas S. Alemán

  University of Pennsylvania

  61 shared

• Ebenezer Daniel

  University of Pennsylvania

  46 shared

• Maureen G. Maguire

  Penn Presbyterian Medical Center

  45 shared

• Wei Pan

  Central South University

  43 shared

• Juan E. Grunwald

  41 shared

• Alexander J. Brucker

  Duke University

  41 shared