Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Optimal Transport-based Graph Matching for 3D retinal OCT image registration

Xin Tian, Nantheera Anantrasirichai, Lindsay Nicholson, Alin Achim

TL;DR

This paper introduces an automated framework for longitudinal registration of 3D mouse retinal OCT images using Optimal Transport-based Graph Matching (OT-GM). It combines Adaptive Weighted Vessel Graph Descriptors (AWVGD) and 3D Cube Descriptors (CD) to robustly match vessel graphs extracted from projection images in the x–y plane, while HR and ILM anatomy guide z-direction alignment. The method constructs dense vessel graphs, solves a MILP OT problem with ghost nodes to handle outliers, and selects the best geometric transformation via a Gain Coefficient criterion. Experimental results on a challenging mouse dataset show that AWVGD combined with CD achieves superior accuracy and GC compared with several baselines, with a reasonable runtime. The framework promises robust cross-day registration in poor-quality mouse OCT data and suggests future multimodal extensions across imaging modalities.

Abstract

Registration of longitudinal optical coherence tomography (OCT) images assists disease monitoring and is essential in image fusion applications. Mouse retinal OCT images are often collected for longitudinal study of eye disease models such as uveitis, but their quality is often poor compared with human imaging. This paper presents a novel but efficient framework involving an optimal transport based graph matching (OT-GM) method for 3D mouse OCT image registration. We first perform registration of fundus-like images obtained by projecting all b-scans of a volume on a plane orthogonal to them, hereafter referred to as the x-y plane. We introduce Adaptive Weighted Vessel Graph Descriptors (AWVGD) and 3D Cube Descriptors (CD) to identify the correspondence between nodes of graphs extracted from segmented vessels within the OCT projection images. The AWVGD comprises scaling, translation and rotation, which are computationally efficient, whereas CD exploits 3D spatial and frequency domain information. The OT-GM method subsequently performs the correct alignment in the x-y plane. Finally, registration along the direction orthogonal to the x-y plane (the z-direction) is guided by the segmentation of two important anatomical features peculiar to mouse b-scans, the Internal Limiting Membrane (ILM) and the hyaloid remnant (HR). Both subjective and objective evaluation results demonstrate that our framework outperforms other well-established methods on mouse OCT images within a reasonable execution time.

Optimal Transport-based Graph Matching for 3D retinal OCT image registration

TL;DR

This paper introduces an automated framework for longitudinal registration of 3D mouse retinal OCT images using Optimal Transport-based Graph Matching (OT-GM). It combines Adaptive Weighted Vessel Graph Descriptors (AWVGD) and 3D Cube Descriptors (CD) to robustly match vessel graphs extracted from projection images in the x–y plane, while HR and ILM anatomy guide z-direction alignment. The method constructs dense vessel graphs, solves a MILP OT problem with ghost nodes to handle outliers, and selects the best geometric transformation via a Gain Coefficient criterion. Experimental results on a challenging mouse dataset show that AWVGD combined with CD achieves superior accuracy and GC compared with several baselines, with a reasonable runtime. The framework promises robust cross-day registration in poor-quality mouse OCT data and suggests future multimodal extensions across imaging modalities.

Abstract

Registration of longitudinal optical coherence tomography (OCT) images assists disease monitoring and is essential in image fusion applications. Mouse retinal OCT images are often collected for longitudinal study of eye disease models such as uveitis, but their quality is often poor compared with human imaging. This paper presents a novel but efficient framework involving an optimal transport based graph matching (OT-GM) method for 3D mouse OCT image registration. We first perform registration of fundus-like images obtained by projecting all b-scans of a volume on a plane orthogonal to them, hereafter referred to as the x-y plane. We introduce Adaptive Weighted Vessel Graph Descriptors (AWVGD) and 3D Cube Descriptors (CD) to identify the correspondence between nodes of graphs extracted from segmented vessels within the OCT projection images. The AWVGD comprises scaling, translation and rotation, which are computationally efficient, whereas CD exploits 3D spatial and frequency domain information. The OT-GM method subsequently performs the correct alignment in the x-y plane. Finally, registration along the direction orthogonal to the x-y plane (the z-direction) is guided by the segmentation of two important anatomical features peculiar to mouse b-scans, the Internal Limiting Membrane (ILM) and the hyaloid remnant (HR). Both subjective and objective evaluation results demonstrate that our framework outperforms other well-established methods on mouse OCT images within a reasonable execution time.
Paper Structure (11 sections, 6 equations, 3 figures, 1 table)

This paper contains 11 sections, 6 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. PROPOSED REGISTRATION FRAMEWORK
  3. Blood vessel extraction
  4. X-Y Plane Registration for Projection Images
  5. Graph Construction
  6. Optimal Transport based Graph Matching
  7. Feature Descriptors from Vessel Graph and 3D OCT Volume
  8. Transformation Model Selection
  9. Z Direction Registration
  10. EXPERIMENTAL RESULT AND ANALYSIS
  11. CONCLUSION AND FUTURE WORK

Figures (3)

  • Figure 1: The flowchart of the proposed 3D mouse OCT registration.
  • Figure 2: (a) Graph construction results before (top) and after (bottom) nodes insert and merge. (b) The 3D Volume feature extraction based on the graph nodes and ILM positions.
  • Figure 3: Comparison results of different registration methods with the checkerboard of the reference and moving image, and its vessel segmentation, B-scans at ONH and two scans 60 slices away from left and right. (a) Before registration (b) our AWVDG+CD (c) CD (d) VOTUS+Z (e) Harris-PIIFD+Z (d) SRWCR.