Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Domain-specific augmentations with resolution agnostic self-attention mechanism improves choroid segmentation in optical coherence tomography images

Jamie Burke, Justin Engelmann, Charlene Hamid, Diana Moukaddem, Dan Pugh, Neeraj Dhaun, Amos Storkey, Niall Strang, Stuart King, Tom MacGillivray, Miguel O. Bernabeu, Ian J. C. MacCormick

TL;DR

The paper addresses the need for a robust, open-source, domain-general tool for choroid segmentation in OCT by proposing REACHNet. It combines domain-specific data augmentation, multi-resolution training, and a resolution-agnostic self-attention mechanism within a lightweight UNet-like architecture to improve segmentation of the choroidal region, vessels, and fovea across devices, while delivering faster inference. The study demonstrates superior Dice performance, stronger correlations and lower MAEs for clinically relevant measurements, and favorable qualitative assessments, including strong external test generalization and efficiency gains. These contributions support scalable, real-time choroid analysis across diverse datasets, facilitating broader biomarker discovery and cross-cohort research in ophthalmology and systemic health.

Abstract

The choroid is a key vascular layer of the eye, supplying oxygen to the retinal photoreceptors. Non-invasive enhanced depth imaging optical coherence tomography (EDI-OCT) has recently improved access and visualisation of the choroid, making it an exciting frontier for discovering novel vascular biomarkers in ophthalmology and wider systemic health. However, current methods to measure the choroid often require use of multiple, independent semi-automatic and deep learning-based algorithms which are not made open-source. Previously, Choroidalyzer -- an open-source, fully automatic deep learning method trained on 5,600 OCT B-scans from 385 eyes -- was developed to fully segment and quantify the choroid in EDI-OCT images, thus addressing these issues. Using the same dataset, we propose a Robust, Resolution-agnostic and Efficient Attention-based network for CHoroid segmentation (REACH). REACHNet leverages multi-resolution training with domain-specific data augmentation to promote generalisation, and uses a lightweight architecture with resolution-agnostic self-attention which is not only faster than Choroidalyzer's previous network (4 images/s vs. 2.75 images/s on a standard laptop CPU), but has greater performance for segmenting the choroid region, vessels and fovea (Dice coefficient for region 0.9769 vs. 0.9749, vessels 0.8612 vs. 0.8192 and fovea 0.8243 vs. 0.3783) due to its improved hyperparameter configuration and model training pipeline. REACHNet can be used with Choroidalyzer as a drop-in replacement for the original model and will be made available upon publication.

Domain-specific augmentations with resolution agnostic self-attention mechanism improves choroid segmentation in optical coherence tomography images

TL;DR

The paper addresses the need for a robust, open-source, domain-general tool for choroid segmentation in OCT by proposing REACHNet. It combines domain-specific data augmentation, multi-resolution training, and a resolution-agnostic self-attention mechanism within a lightweight UNet-like architecture to improve segmentation of the choroidal region, vessels, and fovea across devices, while delivering faster inference. The study demonstrates superior Dice performance, stronger correlations and lower MAEs for clinically relevant measurements, and favorable qualitative assessments, including strong external test generalization and efficiency gains. These contributions support scalable, real-time choroid analysis across diverse datasets, facilitating broader biomarker discovery and cross-cohort research in ophthalmology and systemic health.

Abstract

The choroid is a key vascular layer of the eye, supplying oxygen to the retinal photoreceptors. Non-invasive enhanced depth imaging optical coherence tomography (EDI-OCT) has recently improved access and visualisation of the choroid, making it an exciting frontier for discovering novel vascular biomarkers in ophthalmology and wider systemic health. However, current methods to measure the choroid often require use of multiple, independent semi-automatic and deep learning-based algorithms which are not made open-source. Previously, Choroidalyzer -- an open-source, fully automatic deep learning method trained on 5,600 OCT B-scans from 385 eyes -- was developed to fully segment and quantify the choroid in EDI-OCT images, thus addressing these issues. Using the same dataset, we propose a Robust, Resolution-agnostic and Efficient Attention-based network for CHoroid segmentation (REACH). REACHNet leverages multi-resolution training with domain-specific data augmentation to promote generalisation, and uses a lightweight architecture with resolution-agnostic self-attention which is not only faster than Choroidalyzer's previous network (4 images/s vs. 2.75 images/s on a standard laptop CPU), but has greater performance for segmenting the choroid region, vessels and fovea (Dice coefficient for region 0.9769 vs. 0.9749, vessels 0.8612 vs. 0.8192 and fovea 0.8243 vs. 0.3783) due to its improved hyperparameter configuration and model training pipeline. REACHNet can be used with Choroidalyzer as a drop-in replacement for the original model and will be made available upon publication.
Paper Structure (14 sections, 2 figures, 8 tables)

This paper contains 14 sections, 2 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Data
  4. Domain-specific data augmentation
  5. Multi-resolution training
  6. Model architecture with resolution-agnostic attention mechanism
  7. Model training and hyperparameter configuration
  8. Evaluation and statistical analysis
  9. Results
  10. Discussion
  11. Supplementary materials
  12. Population tables
  13. Mean absolute error analysis of choroid measurements
  14. Qualitative preference stratified by device and manufacturer

Figures (2)

  • Figure 1: Summary of the improvements we propose to build REACHNet (a), including domain-specific data augmentation (i), multi-resolution training (ii) and a resolution-agnostic self-attention mechanism at the deepest point in REACHNet's architecture (iii). Ground-truth label segmentations, and the output expected by REACHNet (b).
  • Figure 2: Domain-specific augmentations improve region and vessel segmentation performance of REACHNet over Choroidalyzer's underlying model.