Three dimensionnal microscopic imaging and morphology of the human corneal endothelium – CorImMo 3D

The strength of this project is to bring together researchers and practitioners from different fields (modeling and applied mathematics, biology and ophthalmology ) with engineers from an IT company, whose purpose is the dissemination and commercialization .
Design and transfer of the complete processing chain of the endothelium in conventional optical microscopy are achieved by a post- doctoral student and a biomedical engineer. They combine 2D and 3D surface approaches from previous research and provide practical technology transfer on a 3D microscopy ( macroscope ). Tribvn will then integrate the first prototype into its own products and start the commercialization for European banks corneas.
Regarding morphological cellular models , the approach is to originally combine computational geometry, mathematical morphology and stochastic models. The mathematical developments lead to the realization of computer code. The results produced are then faced with the reality on the application issue ( ophthalmology and human corneal endothelium ) .
Finally, part of the project aims to improve the in vivo observation of human corneal endothelium by developing a wide field microscope, using structured light , to produce images with up to 20 times more cells than classical specular microscopy images.

Concerning the automatized corneal graft measurement chain, two functional software prototypes have been realized. The first prototype is based on algorithm developed by the I&M research team, while the second use free algorithm from the scientific community. These two software are in a phase of test and optimization. First results seem to show that 2D endothelial cells counts in eye bank lead to an overestimation of 5% of the actual endothelial cell count.

Original mathematical models have been proposed to reproduce the cells mosaics. Generalized Voronoi diagrams, with the germs as well as the distance being the result of random fields, define the simulation of cellular patterns. They have been selected by morphological and geometrical criteria. The choice of the germs is the key point of the process, and will nowadays study this choice with regard to real data.

Last, the work on in vivo endothelium observation with structured illumination raise the inherent difficulty of the cornea, e.g. the stroma thickness and the presence of noise, du to the several cell layers.

The technology transfer into the IT company Tribvn has started. The commercialization of a complete acquisition and analysis solution for cornea grafts is expected within the 3 next years. Long term perspectives will associate the mathematical models to this analysis, and also mix several imaging modalities in the analysis process.

Mathematical models reproducing mosaics by their geometry and morphology, as well as their evolution, will be used to enhance the diagnosis and predict long term impacts of surgeries or pathologies.

International communications:
1. A BERNARD, G HOR, T TIFFET, K RANNOU, Y GAVET, JC PINOLI, S ACQUART, P GAIN, G THURET. 3D reconstruction and segmentation methods for endothelial cell count of stored corneas. Acta Ophthalmologica 2014;92 (s253). European association for Vision and Eye Research EVER 2014. Prix du meilleur poster section cornée/surface oculaire
2. ABERRA GUEBROU S, HE Z, GUILHEM M, GAVET Y, PINOLI JC, GAIN P, THURET G. Innovative, non-contact wide field imaging of corneal endothelium. European association for Vision and Eye Research EVER 2014.
3. K RANNOU, P GUERRERO, Y GAVET, JC PINOLI, G THURET, P GAIN. Spatial statistical analysis of the human corneal endothelial mosaic. Acta Ophthalmologica 2014;92 (s253). European association for Vision and Eye Research EVER 2014.
4. M TRONE, R JULIENNE, K RANNOU, A BERNARD, Z HE, Y GAVET, G THURET, P GAIN. Mathematical model of corneal reendothelialization after inadvertent descemetorhexis. Acta Ophthalmologica 2014;92 (s253). European association for Vision and Eye Research EVER 2014.

Papers with reviewing committee
1. CAMPOLMI N, ACQUART S, HE Z, GAVET Y, JULLIENNE R, NAIGEON N, BERNARD A, FOREST F, PEOC'H M, THURET G, GAIN P. Inherent errors of the fixed-frame counting method for corneal endothelial cell density in eye banks. Cell Tissue Bank 2014 Sep;15:451-9
2. CAMPOLMI N, HE Z, ACQUART S, TRONE MC, BERNARD A, GAUTHIER AS, GARRAUD O, FOREST F, PEOC'H M, GAIN P, THURET G. Comparison of endothelial cell density of organ cultured corneas with the Cornea Donor Study. Cornea 2014;33:597-603
3. Y GAVET, JC PINOLI. Comparison and Supervised Learning of Segmentation Methods Dedicated to Specular Microscope Images of Corneal Endothelium. International Journal of Biomedical Imaging, vol. 2014, Article ID 704791, 13 pages, 2014. doi:10.1155/2014/704791

Needs in corneal endothelial imaging concerns 2 aspects of ophthalmology: evaluation of corneal graft by eye banks and diagnosis of corneal diseases. Endothelial cells (EC) are essential to preserve corneal transparency in any patient and in grafted patient. EC do not replicate; The long term corneal function depends on their cell pool (EC density (ECD) cells/mm2) and their morphometry. In eye banks, ECD is the main criterion to validate the use of the cornea. In patients, EC analysis helps in diagnosing common endothelial diseases (guttata), in verifying that ECD is sufficient in pre and post ocular surgery (refractive and intro ocular lens) and in corneal graft follow-up.

The project is structured through 3 tasks:

Task 1: To develop for eye banks a complete quantitative 3D image acquisition and analysis of the endothelium
Eye banks use specular microscopy (SM) and optic microscopy (OM) for EC analysis. SM are used in the USA (100000 gratf/year) for short-term hypothermic storage, because EC remain visible on graft cornea that have neither oedema nor endothelial folds. OM is used in Europe (100000 graft/year) for grafts stored 3 to 5 weeks in organoculture, that triggers temporary oedema and endothelial folds. EC are made visible by osmotic treatment before counting. The main counting difficulty relies on an uncorrected visualization of EC on folds (3D). The BiiGC lab (Corneal graft biology, engineering and imaging) demonstrated that manual counting (yet largely used at present) is not reliable and must be replaced by image analysis. The I&M team (Imaging and morphology) developed algorithms that improve the recognition of CE in 2D and 3D.

Task 2: Morphological modelling of the endothelial mosaic
The EC mosaic is a joint monolayer without significant regenerative capacity. The BiiGC lab has however highlighted a centripetal slow EC migration from clusters located in extreme periphery. This task will be focused on the morphological modelling and numerical simulation of the endothelial mosaic. This will lead to define concepts, models and methods relevant to represent image cells patterns issued from different acquisition modalities. More precisely, Voronoi diagrams will be studied and applied in 2D and 3D. Originally, they will be extended to gray tone images. This novel approach will enable to simulate gray tone images of human corneal endotheliums both in space and time (for both normal and pathological endotheliums).

Task 3: To develop an innovative microscope based on structured light
At present, patient’s EC evaluation is made by non-contact SM based on principles established in 1968. These SM provide small images (approx 100 EC) without a global mosaic analysis. The I&M team obtained the proof of concept (POC) of a broad endothelial field analysis by non-contact structured-light illumination on patients and on grafts.

This ANR is in continuity with previous joint works (2 PhD , 1 post-Doc) from the EA2521 BiiGC (Jean Monnet University) and the I&M team (FRE ENS des Mines / CNRS 3312), under the aegis of Federative Institute of Research in Health Engineering Science. These works have been valorized by numerous publications, a software protection (Program Protection Agency) and the industrialization of an eye bank endothelial analyzer (Cornea, Tribvn company).