Study Of Methods For Image Quality Improvement In Optical Coherence Tomography

Optical Coherence Tomography (OCT) is a new kind of biomedical optical imaging technology similar to ultra-sound imaging, in which instead of sound wave, the light wave is used and the direct measurement is replaced by interferometric measurement. Because of its advantages of high sensitivity, non-destructive, and high resolution, OCT has the capability of generating cross-sectional images of live tissues in real time, gradually developing into a new clinical diagnostic tool. In this thesis, the imaging principles, system design, and methods for image quality improvement in both the time domain optical coherence tomography (TDOCT) and spectral domain optical coherence tomography (SDOCT) were investigated, including the following main contexts:1. Empirical Mode Decomposition (EMD)-based method was proposed to suppress the scanning noises in TDOCT. Because of its time-varying properties, the time-frequency analysis has better filtering effects than the conventional direct filtering method and the wavelet filtering method. Experimental results showed that EMD-based method can effectively remove the scanning noise as well as the speckle noise without detail dropping.2. A fiber-optics-based SDOCT system was designed and was set up, based-on which the parameters that describe the performance of the SDOCT system and factors that influence the system properties were analyzed in details. Experimental data shows that the depth resolution of the system is about9μm and the system sensitivity is105dB. The cross-sectional images of the tissue simulating liquid and human skin were obtained, demonstrated the tissue imaging capability of the system.3. A novel real-time automatic spectral calibration method for SDOCT was proposed, in the facts were explored that the signal back-scattered from tissue surface is generally much stronger than that from positions in the interior tissue and the surface of tissue is not a plane and phase difference exists between the light reflected from two different points on the surface. The method has the advantage of performing the spectral calibration without needing external measurement steps and external calibration instruments. The two-dimensional (2D) cross-sectional images of human finger skin demonstrated the effectiveness of the method.4. A windowed Fourier transform (WFT) based method was proposed for extracting and compensating depth-resolved phase error in SDOCT system. First, the2D depth-frequency distribution of the interference spectrum of an A-scan of the sample was obtained by using a WFT. Based on the depth-frequency distribution, the phase errors corresponding to different depths were separated and then compensated. Simulation tests and processing of the cross-sectional images of the four-layer cover glass demonstrated effectiveness of the method, improving the depth resolution of the images a lot. The cross-sectional images of a volunteer’s fingernail further demonstrated the power of the method for depth-dependent dispersion compensation in complex tissues.5. The digital image processing methods for further improving the resolution of the OCT images were discussed. A method was then proposed in which the point spread function (PSF) was first obtained directly from the OCT signal itself and the image deconvolution restoration technique was used to reconstruct the high resolution images. Various deconvolution restoration algorithms were compared. Experimental results showed that each deconvolution restoration algorithm can relatively improve the image resolution, among which the CLEAN algorithm and the maximum entropy method are most effectively, and the longitudinal width of the OCT signal can be compressed by them about4and3times respectively.