Development and application of phase sensitive algorithms for improving three dimensional imaging through turbid media
Ανάπτυξη και εφαρμογή αλγορίθμων ανάκτησης φάσης για βελτίωση της τρισδιάστατης απεικόνισης σε σκεδαστικά υλικά
Great progress has been made in biological and biomedical imaging through the last centuries and especially in latest years a variety of optical technologies have been developed and advanced in order to provide diagnostic tools at the services of medicine and biology. However, the penetration depth of conventional optical imaging techniques is limited because of the light scattering. Therefore, scientists require new optical techniques which can image deeper in tissue non-invasively by combining the already existing methods with computational techniques. Moreover, three dimensional imaging can improve significantly imaging methods and give a more accurate depiction of the sample. When light propagates through a diffusing layer to illuminate a sample, the phase information of the recorded image is scrambled seeming information-less and resembles a diffuse halo. However, due to the angular memory effect the diffuse image includes speckle correlations that can help to recover the scrambled phase and reconstruct the hidden object. It can be proven mathematically that the autocorrelation function of the recorded image is equal to the autocorrelation of the hidden object. In fact, only an inversion of the autocorrelation is required in order to achieve a reconstruction of the object. Phase retrieval algorithms are iterative algorithms that can recover the Fourier phase of the image by exploiting simple Fourier theory. These algorithms can reconstruct perfectly the object of interest, but they are not able to retrieve its position. We developed a phase sensitive algorithm to reconstruct an object hidden behind an opaque layer experimentally from images acquired from our lab setup, by exploiting the correlations in the speckle pattern. The angular memory effect provides information for the field of view and the magnification of this imaging technique. We achieved to reconstruct objects of size from 80μm to 160μm that are hidden behind a scattering medium at a distance of 0.6cm to 5cm. In addition, we achieved to detect the movement of the hidden object and quantitatively predict the direction of its shift. The displacement is estimated by the speckles’ shift divided by the magnification of our imaging system. The estimated shift is in agreement with the real displacement of the object from our experiments for horizontal and vertical movements. Therefore, here we present a method to image an object hidden behind a scattering medium and estimate if it has changed position at a specific span of time. The goal of this study was to develop a three dimensional non-invasive imaging technique through opaque media. Optical Projection Tomography is a well-established technique for three dimensional imaging, ideal for semi-transparent samples when scattering is not intense. The specimen is rotated 360 degrees to acquire the projections. The main drawback of this method is that the rotation axis should be always centered on the image camera plane. However, mechanical instabilities usually lead to misaligned projections resulting to an inaccurate reconstruction. Here we developed a new technique to register the projections by exploiting phase retrieval algorithms, and more specifically the feature of the autocorrelation to be always centered, insensitive to the object’s position. The autocorrelation of each projection is always at the center, therefore, the back-projected autocorrelation can be used as an input at a phase retrieval
algorithm to reconstruct the object. This technique improves the imaging capability of the Optical Projection Tomography by correcting for the misaligned projections. We used the phase retrieval algorithms combined with OPT to successfully reconstruct a two dimensional Shepp-Logan phantom from a perturbed stack (sinogram) of projections. In order to image an object hidden behind a scattering layer, the combination of optical projection tomography with a phase sensitive algorithm seems ideal, under the conditions of the angular memory effect. In this project we combined Optical Projection Tomography with phase retrieval algorithms in order to achieve a three dimensional reconstruction of an object embedded between two scattering mediums. The method we developed seems promising for visual representation of complex biological samples in-vivo. Light scattering limits the penetration depth of the current optical techniques, but phase retrieval algorithms can recover the missing information in order to achieve a perfect reconstruction. New three dimensional non-invasive optical imaging implementations are required to improve biomedical imaging, and especially to find applications to studies for the human evolution and diseases.