A coarse-to-fine registration method for multimodal retinal images

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Manual image registration for central serous chorioretinopathy (CSCR) is time-consuming and inconsistent.
• Existing methods like rigid and non-rigid registration have limitations in aligning multimodal retinal images effectively.

Purpose of the Study:

• To develop and evaluate a coarse-to-fine registration method for multimodal retinal images in CSCR.
• To improve the precision and reproducibility of preoperative image alignment for CSCR.

Main Methods:

• A three-step coarse registration using YOLOv8-pose for keypoint matching and affine transformation.
• A dual-component fine registration involving disentanglement learning and a deformable network for local refinement.
• Evaluation on a clinical dataset of color fundus (CF) and fundus fluorescence angiography (FFA) images.

Main Results:

• The proposed method achieves comparable performance to existing approaches.
• Quantitative metrics (Dice and Dice_s scores of 0.6759 and 0.4977) indicate successful alignment.
• Demonstrates potential for accurate local refinement in multimodal retinal image registration.

Conclusions:

• The developed coarse-to-fine registration method offers a promising solution for CSCR preoperative planning.
• It addresses limitations of traditional registration techniques by combining global and local alignment strategies.
• The method has potential clinical application value in enhancing surgical precision for CSCR treatment.