An Investigative study of methods for Retinal Image Registration

AimThis study aims to compare three deep learning-based retinal image registration methods RetinaRegNet, EyeLiner, and GeoFormer on the FIRE dataset to determine which approach provides optimal registration accuracy and computational efficiency across varying image overlap conditions (Classes S, A, and P) using mean landmark error as the primary outcome measure. MethodsThe three pipelines were evaluated under consistent conditions. RetinaRegNet incorporates diffusion features, dual keypoint sampling (SIFT and random), two stage outlier removal, and a multilevel registration hierarchy progressing from homography to polynomial transforms. EyeLiner integrates anatomical segmentation with SuperPoint feature extraction, LightGlue matching, and thin-plate spline warping. GeoFormer builds on LoFTR through cross-attention mechanisms and RANSAC-based refinement. Registration performance was quantified using mean landmark error (MLE). ResultsAcross all 134 FIRE image pairs, RetinaRegNet achieved the lowest overall MLE (3.12 pixels), outperforming EyeLiner (3.66 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). DiscussionOverall, 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.