Iterative Optical Diffraction Tomography for Reconstruction of Multiply-scattering Objects

As a label-free, non-destructive, high-resolution, and quantitative imaging technique, optical diffraction tomography (ODT) has been widely used to image biological samples and microstructures, such as cells, tissues, and optical fibers. The refractive-index (RI) distribution of an object is reconstructed from multi-view measurements of diffracted fields emerging from the object. Typical ODT setups include the object rotating configuration (ORC) and the illumination scanning configuration (ISC). One major limitation of ODT is that it is only applicable to weakly-scattering objects. In this dissertation, novel methods have been developed to overcome the reconstruction difficulty caused by multiple scattering, so as to extend ODT applications. First, an iterative ODT (iODT) algorithm has been developed by iteratively reducing the differences between the forward and backward propagation fields through the scattering area. The perturbative correction to the reconstructed object is computed from the field discrepancies by use of the Rytov-based inversion. iODT as originally designed for ORC, can also be applied to ISC with an appropriate coherent transfer function normalization. Both simulation and experimental results demonstrate that iODT provides accurate and efficient reconstructions of multiply-scattering phase objects with high RI contrasts, large optical path-length differences (OPDs), and/or complicated structures. Furthermore, with the same prior knowledge, iODT also outperforms ODT in resolving the missing-angle problem, in terms of convergence and reconstruction quality. For imaging multiply-scattering objects with complex RI distributions, iODT is prone to crosstalk-like artifacts between the real and imaginary parts. An error subtraction (ES) method has been developed, serving as an add-on module to iODT, to simultaneously reconstruct the RI and the absorption/gain distributions of multiply-scattering objects, even from noisy measurements with signal-to-noise ratio of 20 dB. Finally, we have explored the complementarity of iODT and optimization-based ODT in terms of their advantages and disadvantages, and proposed a combined strategy — iODT initialization for optimization-based ODT. Because the perturbative correction in iODT relies on the Rytov-based inversion instead of a generic gradient method, iODT has a physics-based component that alleviates the trapping in local minima. Numerical results demonstrate that reconstruction only under this combined strategy can accurately converge to the global minimum, especially for multiply-scattering objects with large OPDs