About

William R. Freeman is a Professor of Ophthalmology at UCSD, with a focus on research related to retinal imaging, optical coherence tomography angiography (OCTA), and artificial intelligence applications in ophthalmology. His work involves the development and evaluation of advanced imaging techniques for diagnosing and understanding retinal diseases, including age-related macular degeneration, diabetic retinopathy, and choroidal neovascularization. Freeman's research emphasizes the integration of AI and OCTA to improve disease activity prediction, image quality, and structural analysis of retinal conditions. Throughout his career, Freeman has contributed to the advancement of multimodal retinal image analysis, the application of high-speed OCTA, and the use of machine learning for disease staging and prognosis. His research has been published extensively, highlighting innovations in retinal structural changes, disease activity prediction, and treatment response assessment, with a particular interest in leveraging artificial intelligence to enhance diagnostic accuracy and clinical decision-making in ophthalmology.

Research topics

• Medicine
• Ophthalmology
• Surgery
• Computer Science
• Artificial Intelligence
• Internal medicine
• Biology
• Pathology
• Computer vision
• Chemistry
• Pediatrics
• Endocrinology
• Pharmacology

Selected publications

• OPTICAL COHERENCE TOMOGRAPHY ANGIOGRAPHY ANALYSIS OF COHORT WITH PERIPAPILLARY CHOROIDAL NEOVASCULARIZATION

  Retina · 2025-07-25

  articleSenior authorCorresponding

  PURPOSE: Although many studies have analyzed macular choroidal neovascularization (CNV) using optical coherence tomography angiography, few have focused on peripapillary CNV, which comprises approximately 10% of CNV cases. This study examines optical coherence tomography angiography vessel changes in patients with treated and untreated peripapillary CNV to better understand disease progression. METHODS: Nineteen eyes with peripapillary CNV secondary to neovascular age-related macular degeneration were retrospectively analyzed. Treated eyes (n = 13) received anti-VEGF injections after a modified pro re nata regimen, whereas untreated eyes (n = 6) without macular involvement were closely observed. Median follow-up was 11.9 months (treated) and 8 months (untreated). High-quality optical coherence tomography angiography scans centered on the CNV lesions were selected and analyzed using validated software (ImageJ and AngioTool). RESULTS: Treated eyes showed significant reductions in CNV lesion size, vessel density, and lacunarity after anti-VEGF therapy (ImageJ: P < 0.001; AngioTool: P = 0.0004). Untreated eyes exhibited significant lesion enlargement (ImageJ: P = 0.0356; AngioTool: P = 0.0013). None of the patients in the untreated group developed macular exudation during the study. CONCLUSION: Peripapillary CNV without macular involvement may be considered for short-term observation in selected cases, as these lesions appeared indolent and stable over the limited follow-up period in our cohort. The vessels in optical coherence tomography angiography in the treated group showed regression in size after anti-VEGF therapy. These findings will potentially help clinicians better understand this entity.

  PublisherDOI

• 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

• Retina Spending in Medicare: The State of Ophthalmic Pharmaceutical and Surgical Costs

  Retina · 2025-07-01

  articleSenior author
  PublisherDOI

• Class-Conditioned Image Synthesis with Diffusion for Imbalanced Diabetic Retinopathy Grading

  Lecture notes in computer science · 2025-09-18 · 1 citations

  book-chapter
  PublisherDOI

• Pointwise Structure–Function Analysis of the Ellipsoid Zone in Retinitis Pigmentosa Using an Artificial Intelligence-Assisted OCT and Microperimetry Overlay

  Ophthalmology Science · 2025-07-22

  articleOpen access

  Objective: To perform a pointwise structure-function analysis of the ellipsoid zone (EZ) in retinitis pigmentosa (RP) using an artificial intelligence-based overlay to understand EZ structure-function relationships. Design: A single-center retrospective study. Subjects: Patients with clinically confirmed RP. Methods: Same-day spectral-domain OCT (SD-OCT) and microperimetry near-infrared images were overlaid in patients with confirmed RP. Overlay used a coarse alignment artificial intelligence model. Each locus, on a 68-point microperimetry grid spanning the central 20° of the macula, was identified on individual SD-OCT B-scans. Ellipsoid zone structure was graded at each locus on a 3-point scale: grade 0 = EZ not visible; grade 1 = EZ attenuated; grade 2 = EZ normal. Ellipsoid zone grades were correlated with microperimetry sensitivity scores recorded in decibels (dB). Main Outcome Measures: Correlation of EZ integrity on SD-OCT with microperimetric retinal sensitivity. Results: < 0.001). Correlation between EZ grade and sensitivity was 0.65 (0.64-0.67), whereas correlation of sensitivity with distance from the fovea was -0.41 (-0.43 to -0.39). Focusing on grade 0 loci, 57.5% had sensitivity scores >0 dB, and 4% had scores ≥20 dB, suggesting that these points had function despite no observable EZ on SD-OCT. Conclusions: We identified local EZ structure-function incongruencies in RP using a pointwise analysis of structure-function overlay. These loci of interest may be overlooked in analyses that average across the visual field. Preserved photoreceptor function, in the absence of visibly intact EZ, warrants further investigation. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

  PublisherOA PDFDOI

• 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

• Universal Wavelet Units in 3D Retinal Layer Segmentation

  ArXiv.org · 2025-07-22

  preprintOpen access

  This paper presents the first study to apply tunable wavelet units (UwUs) for 3D retinal layer segmentation from Optical Coherence Tomography (OCT) volumes. To overcome the limitations of conventional max-pooling, we integrate three wavelet-based downsampling modules, OrthLattUwU, BiorthLattUwU, and LS-BiorthLattUwU, into a motion-corrected MGU-Net architecture. These modules use learnable lattice filter banks to preserve both low- and high-frequency features, enhancing spatial detail and structural consistency. Evaluated on the Jacobs Retina Center (JRC) OCT dataset, our framework shows significant improvement in accuracy and Dice score, particularly with LS-BiorthLattUwU, highlighting the benefits of tunable wavelet filters in volumetric medical image segmentation.

  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

• Response to Letter to the Editor: Visual Outcomes after Hyperbaric Oxygen Therapy

  Retina · 2025-03-05

  articleSenior author
  PublisherDOI

• ARTIFICIAL INTELLIGENCE–ENHANCED OPTICAL COHERENCE TOMOGRAPHY ANALYSIS FOR DETECTING INTERNAL LIMITING MEMBRANE REMOVAL IN EPIRETINAL MEMBRANE SURGERY

  Retina · 2025-07-10 · 4 citations

  articleSenior author

  PURPOSE: This study evaluates how accurately humans and artificial intelligence can identify the type of surgery performed for epiretinal membrane (ERM) removal by analyzing postoperative optical coherence tomography (OCT) scans. METHODS: A retrospective analysis at the University of California San Diego included 250 eyes from 239 patients who underwent vitrectomy for idiopathic ERM between January 2013 and October 2024. Eyes were categorized into two groups: one with both the internal limiting membrane and ERM removed using indocyanine green staining, and another with only ERM removal, guided by triamcinolone. Postoperative OCT scans were labeled as either ERM-only or internal limiting membrane + ERM peel based on surgical notes. Both the human grader and artificial intelligence model were trained on 200 labeled OCT scans and tested on 50 masked OCT scans to classify the surgery type. RESULTS: Accuracy of the human grader in identifying the surgical technique was 50%, while the artificial intelligence models demonstrated significantly higher accuracy. The ResNet18 model achieved 61 ± 3%, while UwU-OrthLatt with DB4 initialization and UwU-PR-Relax with Symlet4 initialization reached 70 ± 5% and 69 ± 3%, respectively. CONCLUSION: Artificial intelligence outperformed human grading in detecting internal limiting membrane removal from OCT scans, demonstrating artificial intelligence's potential in improving ophthalmic imaging for clinical use.

  PublisherDOI

Recent grants

• NIH Grant R01EY007366

  NIH · $9.8M · 2015

• NIH Grant R01EY018589

  NIH · $2.0M · 2013

• Vision Biostatistics

  NIH · $17.7M · 2023–2028

Frequent coauthors

• Lingyun Cheng

  306 shared

• Dirk‐Uwe Bartsch

  Jacobs (United States)

  251 shared

• Igor Kozak

  134 shared

• Lingyun Cheng

  114 shared

• F. Mojana

  102 shared

• Dirk-Uwe Bartsch

  University of California, San Diego

  94 shared

• Giulio Barteselli

  94 shared

• Jay Chhablani

  74 shared

Education

• Ph.D., Ophthalmology

  University of California, San Diego

  1990

• M.D., Medicine

  University of California, San Diego

  1986

• B.S., Biology

  University of California, San Diego

  1982