Advertisement

PentaVision

Publications

Retinal Physician

Ophthalmology Management

Ophthalmology Management

Retinal Physician

Retinal Physician

Glaucoma Physician

Glaucoma Physician

Ophthalmic Professional

Ophthalmic Professional

The Ophthalmic ASC

The Ophthalmic ASC

Presbyopia Physician

Presbyopia Physician

Corneal Physician

Corneal Physician

New Retinal Physician

New Retinal Physician

Article

Refining a Deep Learning Model’s Ability to Identify GA Area

The GEODE study evaluated the use of 3D volumetric OCT data to accurately segment the area.

By: Simrat K. Sodhi, MSc, MB BChir (Cantab), Hasenin Al-Khersan, MD, David S. Boyer, MD, Charles C. Wykoff, MD, PhD, Netan Choudhry, MD, FRCSC

Retinal Physician March 1, 2025 Vol 22, Issue March 2025 Page(s): 19-21

Objective:

To evaluate the effectiveness of a deep learning model in accurately segmenting the area of geographic atrophy (GA), a significant condition in age-related macular degeneration (AMD), using 3D volumetric OCT data.

Key Findings:

Average R2 for Spectralis OCT scans was 0.906, indicating strong predictive capability.
Average DSC was 0.826, with a benchmark of 0.86 from previous studies, suggesting room for improvement.
The addition of near infrared (nIR) images did not significantly improve DSC (DSC=0.829), indicating that nIR may not be necessary for optimal performance.
Specific cases with smaller islands of atrophy showed lower DSC due to oversegmentation, highlighting challenges in segmentation accuracy.

Interpretation:

The deep learning model demonstrated high accuracy in segmenting GA areas, particularly in cases with clear boundaries, but faced challenges with smaller atrophic islands, indicating areas for further refinement to enhance clinical applicability.

Limitations:

The study's dataset may not encompass all variations of GA, potentially limiting generalizability to broader patient populations.
The model's performance may vary across different OCT devices, necessitating further validation to ensure reliability in diverse clinical settings.

Conclusion:

The GEODE study highlights the potential of a direct segmentation approach using deep learning for GA identification, with ongoing efforts needed to enhance precision in challenging cases, ultimately aiming to improve patient outcomes.

Hasenin Al-Khersan

Specialties: Ophthalmology

Bio: Hasenin Al-Khersan, MD, is a physician author in retina whose published work in Retinal Physician addresses geographic atrophy in age-related macular degeneration. He coauthored an article on the GEODE study evaluating a deep learning approach using 3D volumetric OCT data to segment areas of geographic atrophy in real-world clinical imaging.

Charles C. Wykoff, MD, PhD

Specialties: Ophthalmology

Bio: Charles C. Wykoff, MD, PhD, is a physician-scientist author in retinal disease. He coauthored a Retinal Physician article on the GEODE study, which evaluated a deep learning approach using 3D volumetric OCT data to segment geographic atrophy area in age-related macular degeneration.

Ashley Brissette

Ophthalmologist

Affiliations: Kelly Vision

Specialties: Ophthalmology

Bio: Ashley Brissette, MD, MSc, FRCSC, is an ophthalmologist in private practice at Kelly Vision in New York City. After a decade in academic medicine, she transitioned into private practice with a focus on patient-centered communication and refractive decision-making. Dr. Brissette writes about the use of social media as an extension of patient education, emphasizing clear, accurate, platform-specific digital content to help patients understand refractive surgery options and support practice growth.

Iqbal K. Ahmed

Specialties: Ophthalmology

Bio: Iqbal K. Ahmed, MD, FRCSC, authored clinical content for Glaucoma Physician on at-home intraocular pressure monitoring and the role of home rebound tonometry in assessing diurnal IOP fluctuations in glaucoma care.

Samuel N. Markowitz

Specialties: Ophthalmology

Bio: Samuel N. Markowitz, MD, FRCSC, is an ophthalmologist who has written on microperimetry and its expanding role in clinical visual function assessment. His work discusses the use of microperimeters to assess retinal sensitivity, scotoma characteristics, fixation stability, and preferred retinal loci in patients with central vision loss.

Netan Choudhry

Specialties: Ophthalmology

Bio: Netan Choudhry, MD, FRCSC, is an ophthalmologist and retina specialist who coauthored a 2025 Retinal Physician article on the GEODE study, which evaluated a deep learning approach using 3D volumetric OCT data to segment geographic atrophy area. His published work in this source focuses on medical retina, including geographic atrophy and age-related macular degeneration, with particular emphasis on imaging and artificial intelligence applications in retinal disease assessment.

Ashley Brissette

Ophthalmologist

Affiliations: Kelly Vision

Specialties: Surgery

Bio: Ashley Brissette is an ophthalmologist at Kelly Vision in New York City. In her Ophthalmology Management article, she describes using social media as an extension of patient education to reinforce refractive surgery decision-making and improve patient understanding. She notes a transition from a decade in academic medicine into private practice, and her current work focuses on creating clear, patient-centered digital educational content across social platforms.