Automated Image Registration Method for In Vivo Confocal Microscopy of the Corneal Sub-basal Nerve Plexus

Area of Science:

• Ophthalmology
• Biomedical Engineering
• Medical Imaging

Background:

• In vivo confocal microscopy (IVCM) is crucial for evaluating corneal innervation.
• Manual quantification of sub-basal nerve plexus from single scans limits diagnostic accuracy for limbal stem cell deficiency (LSCD).
• Current methods struggle with accurate 3D reconstruction and analysis of sequential IVCM scans.

Purpose of the Study:

• To develop and validate an automated 3D reconstruction method for IVCM image volumes.
• To enhance the quantification of sub-basal nerve density for improved LSCD diagnosis.
• To improve the alignment and analysis of sequential IVCM scans for reproducible biomarker evaluation.

Main Methods:

• Developed an automated image registration algorithm combining phase correlation and homography transformation.
• Applied the method to 99 IVCM stacks from LSCD eyes and 23 from normal eyes.
• Validated the algorithm on an independent dataset of 325 volume scans from severe LSCD patients.

Main Results:

• The automated registration significantly outperformed manual registration (correlation: 0.69 vs 0.60; MI: 0.60 vs 0.43; p<0.001).
• Validation demonstrated high performance with correlation of 0.75 and MI of 0.76.
• The method achieved accurate pairwise image correlation and mutual information.

Conclusions:

• The automated 3D reconstruction method significantly improves IVCM image alignment.
• This technique enables more accurate and reproducible 3D evaluation of corneal nerve biomarkers.
• The developed method holds promise for enhancing the diagnosis and monitoring of LSCD.