EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A comparative study of deep learning-based retinal image registration methods

Thenuka Dharmaseelan, Neelabh Sinha, Samyyia Ashraf, Kimia Daneshvar, Amit John, Periklis Giannakis, Yik Ting Chan, Yiu Wai Chan, Nikolas Pontikos

Exploration of Digital Health Technologies · 24.05.2026 · 4

DOI: 10.37349/edht.2026.101194 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Exploration of Digital Health Technologies, online erschienen am 24.05.2026. Verfasst von Thenuka Dharmaseelan und 8 weiteren Autor:innen. Die Arbeit nennt 25 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Aim

To benchmark three deep learning-based retinal image registration methods RetinaRegNet, EyeLiner, and GeoFormer on the Fundus Image Registration (FIRE) dataset to compare registration accuracy and computational efficiency using mean landmark error (MLE) as the primary outcome measure.

Methods

The three image registration approaches were evaluated using the FIRE dataset under consistent conditions across varying image overlap conditions (Classes S, A, and P). These included: (a) RetinaRegNet, which incorporates diffusion features, dual keypoint sampling through Scale-Invariant Feature Transform (SIFT) and random, two-stage outlier removal, and a multilevel registration hierarchy progressing from homography to polynomial transforms; (b) EyeLiner, which integrates anatomical segmentation with SuperPoint feature extraction, LightGlue matching, and thin-plate spline warping; (c) GeoFormer, which builds on Local Feature Transformers (LoFTR) through cross-attention mechanisms and Random Sampling Consensus (RANSAC)-based refinement. Registration performance was quantified using MLE.

Results

Across all 134 FIRE image pairs, RetinaRegNet achieved the lowest overall MLE (3.12 pixels), outperforming EyeLiner (3.81 pixels) and GeoFormer (6.06 pixels). Class-specific analysis showed that RetinaRegNet delivered the highest accuracy in Class S images (1.70 pixels), competitive performance in Class A (5.24 pixels), and the strongest results in the most challenging Class P cases (4.57 pixels). GeoFormer demonstrated the shortest processing time at 0.32 seconds per image pair, compared with 4.92 seconds for EyeLiner and 31.23 seconds for RetinaRegNet. In Class P, RetinaRegNet achieved a 59.2% improvement in accuracy relative to GeoFormer (4.57 vs 11.20 pixels). The code is available at: https://github.com/ThenukaDharmaseelan/image_Registration.

Conclusions

Overall, the evaluation reveals a clear trade-off between registration precision and computational speed. RetinaRegNet achieves the lowest MLE for complex clinical cases despite higher computational cost. EyeLiner balances precision and speed for routine use, while GeoFormer prioritizes rapid throughput where processing speed is critical.

Zitieren

APA

Dharmaseelan, T., Sinha, N., Ashraf, S., Daneshvar, K., John, A., Giannakis, P., Chan, Y. T., Chan, Y. W., & Pontikos, N. (2026). A comparative study of deep learning-based retinal image registration methods. Exploration of Digital Health Technologies, 4. https://doi.org/10.37349/edht.2026.101194

Vancouver

Dharmaseelan T, Sinha N, Ashraf S, Daneshvar K, John A, Giannakis P, et al. A comparative study of deep learning-based retinal image registration methods. Exploration of Digital Health Technologies. 2026;4. doi:10.37349/edht.2026.101194

BibTeX
@article{Dharmaseelan2026,
  author = {Dharmaseelan, Thenuka and Sinha, Neelabh and Ashraf, Samyyia and Daneshvar, Kimia and John, Amit and Giannakis, Periklis and Chan, Yik Ting and Chan, Yiu Wai and Pontikos, Nikolas},
  title = {A comparative study of deep learning-based retinal image registration methods},
  journal = {Exploration of Digital Health Technologies},
  year = {2026},
  volume = {4},
  doi = {10.37349/edht.2026.101194}
}
Kostenlos starten →