Pixel super-resolution using compressive sensing for interferometric quantitative phase imaging

Abstract

Interferometric quantitative phase microscopy (iQPM) with high resolution holds great potential for efficient and label-free investigation of biological systems. However, due to the limited spatial bandwidth product, there is a trade-off between the field of view (FOV) and spatial resolution, which hinders the applications of iQPM for large-scale in situ cellular phenotyping. To address this issue, we developed CS-iQPM, a method applying compressive sensing to capture sub-pixel signals in an iQPM image. In CS-iQPM, we first propose an efficient and easy-to-implement frequency ordering sampling scheme to reorder the Hadamard basis, according to which the mask was selected. This scheme outperforms random sampling as evaluated by three key metrices: peak signal-to-noise ratio, root mean square error, and structural similarity. We second integrated the iterative hard thresholding algorithm with pseudo-inverse matrix to improve the efficiency and structural similarity of reconstruction. Applying these two techniques to our iQPM, the CS-iQPM enhanced the pixel resolution from about 4 pixel/μm² using 10x objective lens to 64 pixel/μm² within 2 s. Experimental results demonstrated that CS-iQPM significantly enhanced the resolution of fine grid patterns and cells in quantitative phase images with a range larger than 500 μm. We envision that CS-iQPM could be applied to many high-throughput in situ cellular phenotyping applications, such as measuring cellular morphologies or cellular mechanical properties at sub-single-cell level in a monolayer during cell migrating.