NerveStitcher

NerveStitcher performs neural-network-based automatic stitching of IVCM images to generate wide-field montages of the corneal nerve plexus and enable computation of corneal nerve morphological parameters.

Key Features:

• Automatic Image Stitching: Neural-network-based fully automatic stitching of in vivo confocal microscopy (IVCM) images to create wide-field montages of the corneal nerve plexus.
• Self-Supervised Point Detection: Extends a self-supervised point detector to identify feature points within IVCM images for reliable correspondences between image pairs.
• Attention Mechanism for Feature Correspondence: Employs an attention mechanism to establish flexible, partial point correspondences and account for local offsets during alignment.
• Improved Matching Accuracy and Efficiency: Improves matching accuracy and reduces processing time compared to conventional methods.
• Morphological Parameter Calculation: Calculates morphological parameters of corneal nerves from single images and stitched montages for quantitative evaluation.

Scientific Applications:

• Early detection of neurological disorders: Supports clinical research into early detection of neurological disorders via detailed corneal nerve morphology assessment.
• Longitudinal studies and large-scale screenings: Enables processing of large IVCM datasets for longitudinal monitoring and large-scale screening of corneal nerve changes.
• Corneal nerve health evaluation and monitoring: Provides quantitative measures to evaluate corneal nerve health and monitor disease progression or response to treatment.

Methodology:

Feature point detection using a self-supervised point detector; correspondence assignment via an attention mechanism allowing partial matches; and image integration and fusion based on local offsets to produce stitched wide-field montages.