Large-area 3D reconstruction of corneal tissues from oscillating focus confocal microscopy.

In vivo confocal microscopy (IVCM) is being used extensively to examine the tissues of the human cornea at the cellular level. IVCM facilitates the examination of alterations of cellular features in a large variety of ocular and systemic pathologies that affect the cornea. A research focus has been on the corneal subbasal nerve plexus (SNP), but recent literature indicates a growing interest in IVCM examinations of immune cells in the cornea. Such applications are currently limited to the assessment of either single IVCM images, wide-field 2D representations of a laterally extended area (primarily of the SNP), or volume reconstructions from a single focus scan, but with a lateral extent of only a single field of view (for example for thickness measurements of corneal layers). The present contribution describes for the first time a workflow to generate large-area volume representations of corneal tissues from laterally extended 3D-IVCM datasets acquired with oscillating focus scans. Three novel techniques for volume reconstruction with different degrees of computational complexity are described, that decompose the dataset into separate focus stacks and create single-stack volume representations that are finally merged. The proposed methods differ in their capability to compensate axial movement occurring during the imaging process. The three novel methods are evaluated on 15 datasets of the SNP and the adjacent tissue layers. All three methods work well by visual inspection for the examined use case, but quantitative image quality assessment shows statistically significant improvement of the merged volume quality with axial motion correction techniques.