Computational methods for microfluidic microscopy and phase-space imaging

Abstract

Modern optical devices are made by assembling separate components such as lenses, objectives, and cameras. Traditionally, each part is optimized separately, even though the trade-offs typically limit the performance of the system overall. This component-based approach is particularly unfit to solve the new challenges brought by modern biology: 3D imaging, in vivo environments, and high sample throughput.

In the first part of this thesis, we introduce a general method to design integrated optical systems. The laws of wave propagation, the performance of available technology, as well as other design parameters are combined as constraints into a single optimization problem. The solution provides qualitative design rules to improve optical systems as well as quantitative task-specific methods to minimize loss of information. Our results have applications in optical data storage, holography, and microscopy.

The second part of this dissertation presents a direct application. We propose a more efficient design for wide-field microscopy with coherent light, based on double transmission through the sample. Historically, speckle noise and aberrations caused by undesired interferences have made coherent illumination unpopular for imaging. We were able to dramatically reduce speckle noise and unwanted interferences using optimized holographic wavefront reconstruction. The resulting microscope not only yields clear coherent images with low aberration -- even in thick samples -- but also increases contrast and enables optical filtering and in-depth sectioning.

In the third part, we develop new imaging techniques that better respond to the needs of modern biology research through implementing optical design optimization. Using a 4D phase-space distribution, we first represent the state and propagation of incoherent light. We then introduce an additional degree of freedom by putting samples in motion in a microfluidic channel, increasing image diversity. From there, we develop a design that is minimally invasive yet optimizes the transfer of information from sample to detector. This optimization best responds to the desired imaging application.

We present three microfluidic devices which can all be implemented as a compact add-on device for commercial microscopes. The first is a flow-scanning structured illumination microfluidic microscopy device demonstrating enhanced resolution in 2D. The second is a method for 3D deconvolution microscopy with a tilted channel to acquire and deconvolve gradually defocused images. Finally, we demonstrate optical projection microscopic tomography with simultaneous phase and intensity imaging capabilities in 3D by combining flow-scanning and optical acquisition in phase space. Experimental results utilize yeast cells as well as live C.elegans.

In the fourth part, we show that optical system optimization also has non-imaging applications such as solar cell engineering. Instead of looking for an optical setup that maximizes the transfer of information, we implement inexpensive surface wrinkles and folds in the layered structure of organic solar cells and optimize their surface density. This strategy enhances light trapping and further improves the electric conversion of solar energy.