Predicting intraocular pressure utilizing highly nonlinear solitary waves as inputs to a CNN

Highly nonlinear solitary waves (HNSWs) are traditionally used in the field of nondestructive evaluation to inspect a material's property without causing damage. The research in this paper proposes a new application for HNSWs: predicting changes in intraocular pressure (IOP) to ensure optimum treatment and prevent the progression of Glaucoma in an eye. The HNSWs used for assessment were collected from a Polydimethylsiloxane (PDMS) eye model and are initiated and stored with a solitary wave transducer. To collect a full range of HNSWs that represent the biological range of IOPs in humans, the PDMS eye model is pressurized from 12mmHg to 26mmHg with 1mmHg increments and waves are collected at each pressure point. Once a HNSW is collected, it is wirelessly transmitted to a server where it is fed into a convolutional neural network to predict the IOP. This is done by extracting relevant features from the HNSW with a Fast Fourier Transform and constructing a spectrogram which can be fed into the algorithm pixel by pixel. This methodology works due to the association of frequency content in the HNSW and changes of the stiffness in the material. In the case of high IOP, the increased pressure pushes against the artificial PDMS cornea and causes it to become stiffer with a higher Young's modulus. We evaluated the ability of the algorithm to predict IOP based on the spectrogram.