Posters

Testing Ann: a neural network to guide keratoconus treatment with ICRS

Poster Details

First Author: L.Bataille SPAIN

Co Author(s):    J. Alio   E. Chorro   F. Versaci   S. Faini   A. Vega-Estrada        

Abstract Details

Purpose:

To describe and validate a neural network aiming to optimized the treatment of keratoconus with ICRS

Setting:

CSO, Florence, Italy. Vissum, Alicante, Spain. Division of Ophthalmology, Miguel Hern�Ã�¡ndez University, Alicante, Spain.

Methods:

ANN is a neural network created to optimize the treatment of keratoconus with ICRS. To achieve that, ANN uses known preoperative and postoperative data to predict the unknown postoperative results from the new preoperative data. These data include visual, refractive parameters, corneal topography and pachymetry.In previous works the simulation of postoperative results from empirical company nomograms (group 1) was compared with the predicted results from the decision proposed by ANN (group 2). Now new data collected from the IBERIA data base let us compare the results obtained with the neural network to the real postoperative clinical data.

Results:

In group 1 we observed a reduction of flattest keratometry (K1) (49.24 �Â�± 4.66D to 46.26 �Â�± 3.41D), steepest keratometry (K2) (52.98 �Â�± 4.81D to 50.18 �Â�± 4.78D) and mean keratometry (KM) (51.11 �Â�± 4.66D to 48.22 �Â�± 4.02D). In group 2 we observed a reduction of K1 (49.24 �Â�± 4.66D to 48.14 �Â�± 4.34D) of K2 (52.98 �Â�± 4.81D to 50.76 �Â�± 4.30D) and KM (51.11 �Â�± 4.66D to 49.45 �Â�± 4.25D). No significant differences were found in K2 (p=0.06) and KM (p=0.18). K1 (p=0.04) showed significant differences.The study of postoperative validation with ANN is in process.

Conclusions:

Neural network analysis correctly proposed an adequate ICRS selection for the treatment of keratoconus. Increasing the input data into a neural network may lead to a more accurate and optimized results.

Financial Disclosure:

NONE