About

Dirk Uwe Bartsch is an Adjunct Professor of Ophthalmology at UC San Diego. His educational background includes a Vordiplom in electrical engineering from TH Darmstadt, and he earned his M.S. and Ph.D. degrees in bioengineering from the University of California San Diego. His research activities focus on multimodal retina image alignment and applications, with significant involvement in projects related to high-resolution ocular imaging, artificial intelligence-enhanced optical coherence tomography analysis, and retinal disease diagnostics. He has contributed to numerous studies utilizing advanced imaging techniques and AI to improve diagnosis and treatment of retinal conditions, including age-related macular degeneration and choroidal neovascularization.

Research topics

• Surgery
• Ophthalmology
• Medicine
• Pathology
• Biology
• Internal medicine

Selected publications

• Artificial intelligence based assessment of treatment response in wet age related macular degeneration using paired OCT angiography

  Scientific Reports · 2026-04-01

  articleOpen access

  Artificial intelligence (AI) has shown promise in retinal imaging, yet its application to longitudinal optical coherence tomography angiography (OCTA) remains limited. This study developed and evaluated an AI model for classifying treatment response in neovascular age-related macular degeneration (nAMD) using paired OCTA images acquired before and after anti-VEGF therapy. In this retrospective cohort study, paired OCTA en-face images and corresponding OCT B-scans were collected for each treatment course. OCTA image pairs were manually segmented and aligned for AI input, while ground-truth labels (Improved, Unchanged, Worsened) were determined based on structural OCT findings and clinical visual acuity outcomes. After quality exclusion, 1033 OCTA pairs were included and divided into training, validation, and independent testing subsets. Two experienced retina specialists graded all OCTA pairs for comparison. On the test set, the AI model achieved an overall accuracy of 82.08%, with class-specific accuracies of 74.29% (worsened), 81.48% (unchanged), and 88.64% (improved). In contrast, overall human grading accuracy was 61.40%. Human graders were significantly more likely to misclassify treatment response than the AI model for all groups (odds ratio = 2.88; 95% CI 1.68–4.92; p < 0.0001). These findings demonstrate that AI-based paired OCTA analysis can provide a more accurate and objective assessment of treatment response in nAMD.

  PublisherDOI

• Characterization of Pentosan Polysulfate Sodium Maculopathy With Fluorescence Lifetime Imaging.

  Open Access CRIS of the University of Bern · 2026-05-15

  articleOpen access

  Purpose Long-term use of pentosan polysulfate sodium (PPS) for interstitial cystitis, has been associated with a sight-threatening pigmentary maculopathy. This study evaluated fluorescence lifetime imaging ophthalmoscopy (FLIO) in PPS maculopathy patients and correlated findings with fundus autofluorescence (FAF) and optical coherence tomography (OCT).Methods This prospective study included patients with confirmed PPS maculopathy at the University of California, San Diego. All patients underwent same-day multimodal imaging (FLIO, fundus photography, FAF, OCT). Disease severity was graded using updated PPS maculopathy criteria. FLIO data were compared to age-, sex- and lens-status-matched controls.Results Nineteen eyes from 10 PPS patients (mean age 71.2) and 19 eyes from 11 age-sex matched normal controls (mean age 71.5) were included. Both groups had seven pseudophakic (PP) and four phakic (P) patients. PPS maculopathy patients showed characteristic FAF changes and RPE hypertrophy. Disease severity did not significantly differ within or between phakic and pseudophakic groups. Mean FLIO lifetimes in PPS eyes were significantly longer than in controls (P: 905 ps v.s 500 ps, p&lt;0.019), PP: 557 ps vs. 371 ps, p&lt;0.044). Lifetimes also varied significantly with maculopathy severity (p&lt;0.05). In 11 eyes, parafoveal foci with very short lifetimes (273-394 ps) were detected and differed significantly from central values (p=0.030).Conclusion FLIO detected significant differences in fluorescence lifetimes between PPS maculopathy patients and controls. Discrete very short lifetime foci may represent early or distinct disease features. FLIO is a promising non-invasive tool for diagnosing and monitoring PPS maculopathy.

  PublisherDOI

• Characterization of Pentosan Polysulfate Sodium Maculopathy With Fluorescence Lifetime Imaging

  Retina · 2026-05-15

  article

  PURPOSE: Long-term use of pentosan polysulfate sodium (PPS) for interstitial cystitis, has been associated with a sight-threatening pigmentary maculopathy. This study evaluated fluorescence lifetime imaging ophthalmoscopy (FLIO) in PPS maculopathy patients and correlated findings with fundus autofluorescence (FAF) and optical coherence tomography (OCT). METHODS: This prospective study included patients with confirmed PPS maculopathy at the University of California, San Diego. All patients underwent same-day multimodal imaging (FLIO, fundus photography, FAF, OCT). Disease severity was graded using updated PPS maculopathy criteria. FLIO data were compared to age-, sex- and lens-status-matched controls. RESULTS: Nineteen eyes from 10 PPS patients (mean age 71.2) and 19 eyes from 11 age-sex matched normal controls (mean age 71.5) were included. Both groups had seven pseudophakic (PP) and four phakic (P) patients. PPS maculopathy patients showed characteristic FAF changes and RPE hypertrophy. Disease severity did not significantly differ within or between phakic and pseudophakic groups. Mean FLIO lifetimes in PPS eyes were significantly longer than in controls (P: 905 ps v.s 500 ps, p<0.019), PP: 557 ps vs. 371 ps, p<0.044). Lifetimes also varied significantly with maculopathy severity (p<0.05). In 11 eyes, parafoveal foci with very short lifetimes (273-394 ps) were detected and differed significantly from central values (p=0.030). CONCLUSION: FLIO detected significant differences in fluorescence lifetimes between PPS maculopathy patients and controls. Discrete very short lifetime foci may represent early or distinct disease features. FLIO is a promising non-invasive tool for diagnosing and monitoring PPS maculopathy.

  PublisherDOI

• Dry eye disease treatment improves subjective quality-of-life responses in patients with AMD, independent of disease stage

  PLoS ONE · 2025-02-06 · 4 citations

  articleOpen accessCorresponding

  PURPOSE: To determine the impact of severity of age-related macular degeneration (AMD) on subjective treatment response in patients treated for dry eye disease. METHODS: A total of 203 eyes diagnosed with evaporative dry eye disease (DED) due to meibomian gland dysfunction were treated using the LipiFlow or MiBoFlo systems. From this cohort, 40 eyes with stable dry AMD (early, intermediate, or late stages) were included. Each participant completed the Ocular Surface Disease Index (OSDI) and Standard Patient Evaluation of Eye Dryness Questionnaire (SPEED) before treatment and at a 6-month follow-up. Changes in questionnaire scores were analyzed using one-way analysis of variance (ANOVA) to assess differences between AMD severity groups. RESULTS: Improvement in SPEED and OSDI scores, including vision related OSDI scores were observed across all AMD stages, with no significant differences between groups (p<0.05). CONCLUSION: Managing DED improved quality of life (QOL) in patients with AMD, regardless of retinal disease severity. This highlights the importance of treating coexisting ocular surface conditions to enhance patient outcomes, even in the presence of significant maculopathy.

  PublisherDOI

• Combination Intravitreal Steroid and Anti-VEGF Therapy for Double-Monotherapy–Resistant Chronic Diabetic Macular Edema

  Ophthalmic surgery, lasers & imaging retina · 2025-09-25 · 1 citations

  article

  Background and Objective: The purpose of this retrospective chart review was to evaluate the efficacy and safety of combined intravitreal long-acting corticosteroids and anti-vascular endothelial growth factor (anti-VEGF) therapeutic agents for resistant cystoid macular edema associated with diabetic retinopathy, or diabetic macular edema (DME). Patients and Methods: This study included 16 eyes of 15 patients. Patients with unresponsive DME, despite aggressive monthly monotherapy with steroid or anti-VEGF (decanted triamcinolone or 4 mg aflibercept) were included. Treatment consisted of simultaneous administration of anti-VEGF and decanted triamcinolone. Optical coherence tomography (OCT) and comprehensive ophthalmic exams were performed prior to treatment, as well as every 4 weeks (for 4 months), post-treatment. Main outcome measures were central retinal thickness (CRT) measured by OCT in microns and best-corrected visual acuity (BCVA). Results: The mean treated disease duration (months) was 38.93 ± SD 8.73 [16–108]. There was a CRT reduction at post-combination 1 month (159.18 µm, P = 0.0003), 2 month (169.58 µm, P = 0.0038), 3 month visit (167.88 µm, P = 00.0200), and 4 month (212.88 µm, P = 0.0276). BCVA improved by 8.5 letters ( P = 0.0216) at month 1. Improvement persisted with a mean of nine letters ( P = 0.039) at month 2, 10 letters ( P = 0.035) at month 3, and three letters ( P = 0.638) at month 4. No ocular or systemic safety issues were noted after the combination therapy. Conclusion: Combination therapy of steroid and anti-VEGF results in clear anatomic and function improvement in eyes with DME that were resistant to monotherapy with anti-VEGF or monotherapy with steroids. The use of this combination has shown a therapeutic effect in treatment-resistant DME.

  PublisherDOI

• Universal Vessel Segmentation for Multi-Modality Retinal Images

  IEEE Transactions on Image Processing · 2025-01-01

  articleOpen access

  We identify two major limitations in the existing studies on retinal vessel segmentation: 1) Most existing works are restricted to one modality, i.e., the Color Fundus (CF). However, multi-modality retinal images are used every day in the study of the retina and diagnosis of retinal diseases, and the study of vessel segmentation on other modalities is scarce; 2) Even though a few works extended their experiments to new modalities such as the Multi-Color Scanning Laser Ophthalmoscopy (MC), these works still require fine-tuning a separate model for the new modality. The fine-tuning will require extra training data, which is difficult to acquire. In this work, we present a novel universal vessel segmentation model (URVSM) for multi-modality retinal images. In addition to performing the study on a much wider range of image modalities, we also propose a universal model to segment the vessels in all these commonly used modalities. While being much more versatile compared with existing methods, our universal model also demonstrates comparable performance to the state-of-the-art fine-tuned methods. To the best of our knowledge, this is the first work that achieves modality-agnostic retinal vessel segmentation and the first to study retinal vessel segmentation in several novel modalities (Code, model and 3 new retinal vessel segmentation datasets are available at https://github.com/JRC-VPLab/URVSM).

  PublisherOA PDFDOI

• Impact of 85 kHz versus 125 kHz SHIFT OCTA scan speeds on image quality in retinal diseases and diagnostic reliability of choroidal neovascular membranes

  Scientific Reports · 2025-12-18

  articleOpen access

  The purpose of the study was to compare Optical Coherence Tomography Angiography (OCTA) scan speeds of 85 kHz and 125 kHz with respect to image quality, diagnostic reliability, and scan time in patients with retinal diseases. In this prospective cohort study, OCTA images were obtained at both scan speeds in 70 eyes from 40 patients with retinal diseases. Masked expert graders evaluated qualitative parameters including clinical utility, artifacts, and overall image quality. Quantitative parameters including scan time, Heidelberg Q-score, and OCTA-Q score were recorded. In 46 eyes with visible choroidal neovascular membrane in the avascular layer of OCTA, AngioTool (Image J) was used to assess vessel percentage area, vessel junction density, average vessel length, and E-Lacunarity. Acquisition speed of 125 kHz OCTA was significantly faster than that of 85 kHz. There were no statistically significant differences in AngioTool parameters between the two protocols. 125 kHz was significantly better than 85 kHz for image quality with fewer noise artefacts and vessel projection artefacts. In conclusion, 125 kHz SHIFT OCTA offers comparable to better image quality to the 85 kHz OCTA with significantly faster acquisition, potentially improving clinical workflow without compromising diagnostic reliability.

  PublisherOA PDFDOI

• Tunable Wavelet Unit based Convolutional Neural Network in Optical Coherence Tomography Analysis Enhancement for Classifying Type of Epiretinal Membrane Surgery

  ArXiv.org · 2025-07-01

  preprintOpen access

  In this study, we developed deep learning-based method to classify the type of surgery performed for epiretinal membrane (ERM) removal, either internal limiting membrane (ILM) removal or ERM-alone removal. Our model, based on the ResNet18 convolutional neural network (CNN) architecture, utilizes postoperative optical coherence tomography (OCT) center scans as inputs. We evaluated the model using both original scans and scans preprocessed with energy crop and wavelet denoising, achieving 72% accuracy on preprocessed inputs, outperforming the 66% accuracy achieved on original scans. To further improve accuracy, we integrated tunable wavelet units with two key adaptations: Orthogonal Lattice-based Wavelet Units (OrthLatt-UwU) and Perfect Reconstruction Relaxation-based Wavelet Units (PR-Relax-UwU). These units allowed the model to automatically adjust filter coefficients during training and were incorporated into downsampling, stride-two convolution, and pooling layers, enhancing its ability to distinguish between ERM-ILM removal and ERM-alone removal, with OrthLattUwU boosting accuracy to 76% and PR-Relax-UwU increasing performance to 78%. Performance comparisons showed that our AI model outperformed a trained human grader, who achieved only 50% accuracy in classifying the removal surgery types from postoperative OCT scans. These findings highlight the potential of CNN based models to improve clinical decision-making by providing more accurate and reliable classifications. To the best of our knowledge, this is the first work to employ tunable wavelets for classifying different types of ERM removal surgery.

  PublisherOA PDFDOI

• Universal Vessel Segmentation for Multi-Modality Retinal Images

  arXiv (Cornell University) · 2025-02-10

  preprintOpen access

  We identify two major limitations in the existing studies on retinal vessel segmentation: (1) Most existing works are restricted to one modality, i.e., the Color Fundus (CF). However, multi-modality retinal images are used every day in the study of the retina and diagnosis of retinal diseases, and the study of vessel segmentation on other modalities is scarce; (2) Even though a few works extended their experiments to new modalities such as the Multi-Color Scanning Laser Ophthalmoscopy (MC), these works still require fine-tuning a separate model for the new modality. The fine-tuning will require extra training data, which is difficult to acquire. In this work, we present a novel universal vessel segmentation model (URVSM) for multi-modality retinal images. In addition to performing the study on a much wider range of image modalities, we also propose a universal model to segment the vessels in all these commonly used modalities. While being much more versatile compared with existing methods, our universal model also demonstrates comparable performance to the state-of-the-art fine-tuned methods. To the best of our knowledge, this is the first work that achieves modality-agnostic retinal vessel segmentation and the first to study retinal vessel segmentation in several novel modalities.

  PublisherOA PDFDOI

• COMPARISON OF ULTRAHIGH-SPEED 250 kHz OCTA WITH HIGH-SPEED 125 kHz MODE USING HEIDELBERG SPECTRALIS SHIFT

  Retina · 2025-12-26 · 1 citations

  article

  PURPOSE: Optical coherence tomography angiography is widely used to detect choroidal neovascularization in retinal diseases, but acquiring high-density raster scans can prolong acquisition time and increase motion artifacts due to patient fatigue. This study evaluates whether the 250 kHz ultrahigh-speed mode (3× faster than baseline) reduces scan time and motion artifacts while maintaining image quality compared with the FDA-approved 125 kHz mode. METHODS: Fifty-eight eyes underwent 10°×10° fovea-centered optical coherence tomography angiography imaging using both modes. In 38 choroidal neovascularization eyes, AngioTool was used to analyze vascular metrics. In 20 nonchoroidal neovascularization eyes, superficial foveal avascular zone area was measured using ImageJ. Two masked retina specialists graded paired images for image quality, motion artifacts, and clinical diagnosis. Acquisition time was recorded. RESULTS: The 250 kHz mode reduced acquisition time by 43% (17.5 ± 9.0 seconds vs. 33.6 ± 20.7 seconds, P < 0.00001). Image quality and motion artifacts favored the 250 kHz mode. Clinical diagnosis was rated similarly across modes. Quantitative metrics, including vessel area, junctions, lacunarity, and foveal avascular zone area, showed no significant differences (all P > 0.1). CONCLUSION: This is the first intraindividual comparison of Heidelberg's 250 kHz and 125 kHz optical coherence tomography angiography modes. The findings demonstrate that ultrahigh-speed scanning can enhance efficiency and image quality without compromising diagnostic or quantitative utility, highlighting its potential clinical impact pending Food and Drug Administration approval.

  PublisherDOI

Recent grants

• Mechanistic-Based Non-Invasive Assessment of Retinal Damage in HAART Era

  NIH · $4.6M · 2006–2022

• NIH Grant R01EY013304

  NIH · $1.1M · 2005

• Mechanistic-Based Non-Invasive Assessment of Retinal Damage in HAART Era

  NIH · $388k · 2006–2021

• Vision Biostatistics

  NIH · $17.7M · 2023–2028

Frequent coauthors

• William R. Freeman

  Jacobs (United States)

  94 shared

• Lingyun Cheng

  45 shared

• Giulio Barteselli

  29 shared

• Robert N. Weinreb

  University of California, San Diego

  24 shared

• Jay Chhablani

  19 shared

• Maria Laura Gomez

  Spanish National Cancer Research Centre

  17 shared

• Joseph T Nezgoda

  Retina Macula Institute

  17 shared

• Natalia Camacho

  17 shared