A Neural Network Approach to Monitor Intraocular Pressure for Glaucoma Diagnosis

A nanoarray-enhanced, Fabry-Perot intraocular pressure (IOP) sensor has been recently fabricated for implantation in the eye for glaucoma diagnosis [1]. This work involves the development of an algorithm to process reflectivity data from the sensor collected via remote optical readout under pressure conditions observed in glaucoma patients. The method involves pressure extraction using a neural network approach based on prominent optical spectra characteristics. With the sensor in a pressure chamber, a correlation coefficient of 0.9994 was obtained between the measured and neural network extracted pressure with a run time of less than 30 s, demonstrating the accuracy and efficiency of the algorithm. BACKGROUND A nanoarray-enhanced, Fabry-Perot sensor (Fig. 1) has been developed [1] whose membrane deformation depends on IOP. This deformation is found by measuring the optical spectrum of the sensor. However, as no analytical method exists to determine the IOP from the optical spectrum, a numerical method (Fig. 2) is developed here to extract the pressure from arbitrary optical data. Optical data was collected from the sensor when tested in a pressurized water chamber, ex-vivo rabbit eyes, and in-vivo rabbit eyes. Neural networks have been successful for optical signal processing [3], making them an attractive tool for extracting IOP from collected optical data.