Related Experiment Video
Updated: Nov 24, 2025
Author Spotlight: An Efficient and Robust Software for Automated Fusion of Multiple Preclinical Imaging Modalities
Published on: October 27, 2023
Retinal image registration as a tool for supporting clinical applications.
Carlos Hernandez-Matas1, Xenophon Zabulis2, Antonis A Argyros1
1Institute of Computer Science, Foundation for Research and Technology-Hellas, Heraklion, 70013 Greece; Computer Science Department, University of Crete, Heraklion, 70013 Greece.
This study introduces the Retinal Image Registration through Eye Modelling and Pose Estimation (REMPE) framework for improved retinal image analysis. REMPE enhances diagnostic capabilities by accurately estimating eye shape and improving retinal image registration for various applications.
Area of Science:
- Ophthalmology
- Medical Imaging
- Computer Vision
Background:
- Small vessel analysis in the retina aids in diagnosing vasculopathy.
- Retinal Image Registration (RIR) applications like mosaicing and Super Resolution (SR) facilitate this analysis.
- Challenges in RIR include retinal changes, diverse devices, and retinal curvature.
Purpose of the Study:
- To develop and evaluate the Retinal Image Registration through Eye Modelling and Pose Estimation (REMPE) framework.
- To assess REMPE's effectiveness in RIR applications and eye shape estimation.
- To explore REMPE's utility in longitudinal studies and 3D eye modeling.
Main Methods:
- The REMPE framework was employed to simultaneously estimate camera poses and eye shape/orientation.
- Quantitative assessments were performed for SR and eye shape estimation.
- Qualitative evaluations were conducted for longitudinal studies, mosaicing, and multiple image registration.
Main Results:
- REMPE demonstrated quantitative suitability for SR and eye shape estimation.
- Qualitative results showed REMPE's usefulness in longitudinal studies, mosaicing, and multiple image registration.
- Key novelty includes eye shape estimation and 3D mesh generation, enabling non-distorted measurements.
Conclusions:
- Retinal Image Registration (RIR) effectively supports applications like SR, eye shape estimation, and longitudinal studies.
- The REMPE framework offers improved registration accuracy over state-of-the-art methods.
- Enhanced accuracy directly improves performance in supporting various retinal image analysis applications.
