Evaluation strategy for image acquisition protocols in confocal microscopy mosaic imaging of the cornea

The experimental data contains datasets from four specimens:

SiliconeTiO2: a transparent silicone volume with differently-sized TiO2 particles embedded,

PorcineEye1: a dead porcine eye,

PorcineEye2: a second dead porcine eye,

PorcineCornea2: the cornea trephined from PorcineEye2.

All four specimens were fixated in specific, custom-made holding fixtures, which were mounted on a motorized x-y-translation stage (8MTL20XY-S, Standa, Vilnius, Lithuania) for image acquisition. An HRTII (Heidelberg Engineering, Heidelberg, Germany) with an RCM2.0 (research prototype, Rostock University Medical Center) (see https://doi.org/10.1364/BOE.9.002511) was used for CLSM imaging. This setup has a FOV of 350 x 350 µm². According to the imaging protocol for human eyes, a protective single-use cap covered the RCM2.0 objective lens. A self-developed C# software program controlled the image acquisition process and the movement pattern of the translation stage.

The experiments included three different movement patterns, one for the ground truth data acquisition and two for the evaluation datasets. The ground truth data of each specimen encompasses an area of approximately 2.38 x 2.38 mm² that was imaged by a rectilinearly arranged grid of 8 x 8 single CLSM images, spaced 290 µm so that adjacent images overlap by 60 µm. The movement between grid positions happened in between the recording of the single images, while no movement was exerted during the actual acquisition of the images.

Two predefined movement patterns were used to record the experimental data: An extending spiral pattern and a simulated fixational eye movements pattern. The spiral pattern simulates our in vivo imaging technique, which uses a moving fixation target on a display device in front of the contralateral eye of the examined person to guide their eye movements. The spiraling movement starts at a central position and then expands outwards until the spiral radius reaches 1 mm. The simulated fixational eye movements pattern with a length of 30 s was generated using the open-source FixEM simulation software (see https://doi.org/10.1515/cdbme-2024-2007). The movement patterns were used on the four specimens with different numbers of repetitions.