WIDE-RANGE MOSAIC OF CORNEAL SUB-BASAL NERVES IMAGES BY AUTOMATIC MONTAGING

Abstract

 

Purpose

 

To provide a wide-range mosaic of the sub-basal nerve layer of central cornea, built from several images acquired in-vivo with confocal microscopy The montage is performed by a fully automatic computerized system, which allows clinicians to examine a single large, high quality image.

 

Methods

 

A laser scanning confocal microscope (IVCM; Heidelberg Retinal Tomograph 3 with Rostock Corneal Module; Heidelberg Engineering, Germany) was used in-vivo to acquire the images. Images represented an en face view of a 400 × 400 μm corneal area. 50 images were acquired from each eye of 2 subjects. The mosaicking process is composed of 3 steps. At first, all possible pairs of images are considered and evaluated in terms of their similarity and attributed a score. A pair of images that share a large overlap area will have a high score of similarity, while a low score will correspond to pairs of images with little or no overlap, hence impossible to register. In the second step the actual registration is carried out: images in the N pairs with the highest score are registered between themselves, applying translation, rotation, and affinity transformations. After the registration of these N pairs (each now a single image), the previous steps are repeated until a single large image is obtained. In the third and last step, a custom blending procedure, based on pixel intensity weighting, provides the final high quality image with homogeneous luminosity and contrast.

 

Results

 

A visual inspection of the resulting images confirmed the capability of the proposed system to provide high quality wide-range images. Fig. 2 shows one of the final mosaic image, with a reconstructed area of ~2000×1600 μm, from 45 400×400 μm single images. The total processing time was 6 minutes.

 

Conclusions

 

The proposed system allowed in this dataset the successful montaging of the sub-basal corneal nerve images. The resulting mosaic provides a larger high quality image than the single original ones, which should significantly aid clinicians in evaluating and assessing in a more reliable way the pathologic signs of interest.

 

 

Fig. 1. Example of the fully automatic image mosaic with no blending (steps 1 and 2).

 

 

Fig. 2. Example of the fully automatic image mosaic after our custom blending algorithm (step 3).

 

Keywords: 484 cornea: stroma and keratocytes • 550 imaging/image analysis: clinical • 596 microscopy: confocal/tunneling