Hyperspectral Digital Holography for Bio-tissue Segmentation

Abstract

Hyperspectral digital holography is a relatively new imaging method. This work discusses and demonstrates its advantages compared to traditional imaging methods. The main problem with current imaging systems is that they do not capture transparent objects or their shapes very well. This is especially important in cells and tissues, which consist mostly of water and are thus transparent. Holography has the advantage of utilizing the full complex amplitude, instead of just its intensity. It can be utilized in different detection and segmentation problems, but here the focus is in biotechnology and tissue segmentation. The hyperspectral imaging process in this work consists of capturing diffracted images of the sample at different wavelengths, backpropagating the diffracted images to correct focus and then segmenting the acquired image stack. A lensless capturing setup is used, which has the advantage of having no chromatic aberration. The imaging process in an adaptation of the Fourier transform spectrometer, which requires only one scan to get the full hyperspectral holographic information. The phase delay for the spectrometer is created by adjusting the optical distance of one of the wavefronts in the system. This is done by a mirror controlled by a piezotransducer. To be able to capture hyperspectral information, a supercontinuum laser source is used. The backpropagation is performed with the angular spectrum method, which is computationally efficient. Segmentation is done using k-means clustering in MATLAB. The results show that hyperspectral digital holography is better for capturing transparent objects due to the phase information and wide wavelength spectra. Some of the possible error sources and complications in the process are also discussed.

Chapter 1 looks at some of the most recent studies regarding the subject. Chapter 2 describes the theory of wave optics and Fourier transforms necessary for this work. Chapter 3 illustrates the practical process for capturing, reconstructing and segmenting the images. In chapter 4 the resulting segmented images are discussed and chapter 5 concludes the work