Multi-frame fusion enhances analytical and diagnostic efficiency in corneal confocal microscopy

Area of Science:

• Ophthalmology
• Medical Imaging
• Biomedical Engineering

Background:

• Corneal confocal microscopy (CCM) is crucial for diagnosing corneal diseases.
• Current CCM imaging can suffer from noise and low clarity, hindering accurate analysis.
• Existing methods often require specialized hardware or complex workflows.

Purpose of the Study:

• To develop a low-cost, multi-frame fusion strategy for enhancing CCM image quality.
• To improve the accuracy of corneal nerve feature extraction and disease classification.
• To validate the clinical utility of the proposed image enhancement technique.

Main Methods:

• A novel image enhancement strategy based on aligning and integrating consecutive CCM frames.
• Systematic evaluation of image alignment accuracy, noise reduction, and morphological feature extraction.
• Assessment of classification performance using both traditional metrics and deep learning models.

Main Results:

• Significant improvements in structural clarity and measurement reliability of corneal nerve features.
• Substantial enhancement in corneal nerve fiber length (CNFL), density (CNFD), and branch density (CNBD), particularly in diabetic patients.
• Consistent improvement in disease discrimination and classification accuracy across various deep learning architectures.

Conclusions:

• The multi-frame fusion strategy effectively enhances CCM images without additional hardware or workflow changes.
• The method demonstrates significant clinical value by improving diagnostic accuracy and measurement reliability.
• This technique is highly suitable for real-world clinical applications, offering a practical solution for better corneal imaging.