Towards estimation of optical and structural ophthalmic properties based on optical coherence tomography

Abstract

Early diagnosis of retinal diseases such as glaucoma will benefit from unbiased an precise estimation of both optical and structural properties of the RNFL as they provide a better understanding of the tissue characteristics. The main objective of this thesis was first to improve the estimation of the attenuation coefficients of layered samples and, second, to estimate the structural properties of RNFL. Unbiased estimation of optical tissue properties such as the attenuation coefficient require a model of the recorded OCT signal. To study the characteristics of the OCT signal, in Chapter 2, two simulation methods were presented for homogeneous samples. In both methods single-scattering of the OCT light was assumed and the effect of the shape of OCT beam was taken into account. The more complex simulation also takes into account the interference of the electrical fields in the sample and reference arms and several post processing steps. Later in this thesis the simpler model was used to model the OCT signal since both simulation methods generated similar results. In Chapter 3, we improved an existing depth-resolved method to estimate the attenuation coefficients. The existing method does not handle noise at the larger depths, where the OCT light is fully attenuated, which results in a variation of the estimated attenuation coefficient values. We introduced a technique to detect and exclude the noise regions from the OCT scans to improve the accuracy and reduce the Aline-by-Aline variation of the estimated attenuation coefficients. The results show a better accuracy of the estimated attenuation coefficient, especially in sub-RPE regions and a better quality of the attenuation coefficient images. In Chapter 4, a method was presented to estimate the attenuation coefficients of a homogeneous medium accounting for the shape of the focused light beam. For this, the model presented in Chapter 2 was fitted to the measured OCT signal of a homogeneous sample to estimate the model parameters. The presented method was first implemented for the semi-infinite samples and was tested for different concentrations of TiO2 in silicone for different locations of focus. In addition, a statistical and numerical analysis was performed to evaluate the presented method under various experimental conditions. The estimation result shows a reasonable correlation between the TiO2 weight-concentration and the estimated attenuation coefficient. While the method could estimate the attenuation coefficients of a uniform samples, most biological tissues such as the retina are layered, hence the method was extended in Chapter 5 to estimate the attenuation coefficients of the multi-layer samples. This method was tested on the simulation and measurements of a multi-layer phantom with different concentration of TiO2 in silicone with two systems: one with a small (40 μm) and one with a larger (300 μm) Rayleigh length. The numerical results show an acceptable estimation of the attenuation coefficients for the Rayleigh lengths less than 0.5 mm in air and acceptable for clinical application using clinical OCT systems. For both single- and multi-layer samples, a linear relation between the estimated attenuation coefficients and the particle concentration of the perspective layer was observed while using single and multiple B-scans. In Chapter 6, an automatic technique was developed to estimate the orientation of RNFBs from volumetric OCT scans. The RNFB orientations of six macular scans from three subjects were used to evaluate the results. We observed a good correlation between the manual tracing and the estimated orientations of the RNFLs. RNFBs orientation in combination with other techniques such as VF and SAP can assist the ophthalmologists to have a more reliable measurement for an early diagnosis of retinal diseases such as glaucoma.