Precise Multiscale Registration and Anti-Shaking Algorithm for Stable Nailfold Vascular Area Visualization

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Ophthalmology

Background:

• Nailfold capillary imaging is crucial for assessing microvascular health.
• Image blurring from tremors and complex capillary distributions hinder accurate analysis.
• Existing methods struggle with real-time focus tracking and image registration.

Purpose of the Study:

• To develop a robust system for real-time focus tracking and image registration in nailfold capillary imaging.
• To overcome limitations of image blurring and enhance the quality of vascular data.
• To improve the accuracy of quantitative measurements from capillary images.

Main Methods:

• Combined YOLOv8 object detection with a Laplacian clarity evaluation function for real-time focus assessment.
• Employed normalized cross-correlation template matching for optimal frame alignment and registration.
• Validated the system using zooming and fixed-focus experiments on nailfold capillary images.

Main Results:

• Achieved significant improvements in image quality metrics: Mean Squared Error (MSE) reduced by up to 64.5%, Peak Signal-to-Noise Ratio (PSNR) increased by up to 3.88 dB, and Structural Similarity Index (SSIM) improved by up to 4.7%.
• Demonstrated superior performance compared to the original video and state-of-the-art registration algorithms across all evaluated metrics.
• Successfully obtained optimal frame offsets for accurate image registration, preserving vascular information.

Conclusions:

• The proposed integrated approach effectively addresses challenges in nailfold capillary imaging, providing enhanced clarity and accuracy.
• This method offers a significant advancement for non-invasive microvascular assessment.
• The system demonstrates potential for wider application in clinical diagnostics and research requiring high-quality capillary imaging.