Abstract
Purpose: :
To predict the nerve fiber layer thickness in association with retrobulbar hemodynamis in open angle glaucoma using artificial neural networks.
Methods: :
Simulated data set of 1000 entries was generated on 60 subjects’ data with open angle glaucoma case. Color Doppler imaging was used to asses peak systolic velocity (PSV), end-diastolic velocity (EDV), resistance index (RI) in the ophthalmic (OA), central retinal (CRA) and short posterior ciliary arteries (SPCA). Standard nerve fiber layer parameters were assessed using scanning laser polarimetry. As follows glaucoma mathematical model was created. The probability of the correct prognosis of the nerve fiber layer thickness in respect to retrobulbar hemodynamics was estimated using the radial basis function classifier (RBF), multiple layer perceptron (MLP) and novel glaucoma neural network (NNT).
Results: :
After controlling for age, gender, blood pressure and intraocular pressure every 10 microns decrease in nerve fiber layer thickness was associated with an increase in RI by 0.45 in CRA (p<0.01) and decrease in RI by 0.48 in SCPA (p<0.001).There was a significant increase in EDV by 5cm/s in CRA for every 10 microns decrease in nerve fiber layer thickness (p<0.01). Significant increase in PSV by 20cm/s and 10cm/s in OA and CRA, respectively for every 10 microns decrease in nerve fiber layer thickness was determined (p<0.01). The probability of the correct prognosis of the nerve fiber layer thickness in respect to retrobulbar hemodynamics in open angle glaucoma was estimated to be 0.84 (95% Confidence interval 0.78,0.89), 0.95 (95% Confidence interval 0.91,0.98), and 0.99 (95% Confidence interval 0.998,0.999) for RBF, MLP and NNT, respectively.
Conclusions: :
The diminished nerve fiber layer thickness is closely associated with significant changes in retrobulbar hemodynamics in open angle glaucoma. Artificial neural networks are a potential tool to predict nerve fiber layer thickness in relation to retrobulbar hemodynamics alterations.
Keywords: blood supply • nerve fiber layer • computational modeling