Microscopy is a major research tool for scientific inquiry, providing both morphological and molecular imaging capabilities. Furthermore, it is the only imaging modality that routinely enables non-ionizing and radiation-free functional imaging at sub-micron spatial resolutions. Unfortunately, traditional microscopy systems often decouple data acquisition and image reconstruction protocols, which limit their imaging capabilities with respect to imaging resolution, depth, contrast, throughput, and phototoxicity. In turn, this limits the scope of real-world phenomena observable with traditional microscopy.
In this talk, I present computational microscopy techniques that leverage recent developments in computing power and useable data, to enable unprecedented optical imaging capabilities in the biological sciences. These new techniques utilize data-driven image-reconstruction pipelines that tightly integrate together system hardware with computational reconstruction frameworks. I will discuss particular applications that focus on large-scale phase and fluorescent imaging, multimodal super-resolution microscopy, and 3D imaging through optical scatter. I will also discuss the computational frameworks that underpin these applications, which are often based on large-scale nonlinear and nonconvex optimization. Lastly, I will discuss future research directions that can capitalize and extend these developments to dramatically extend the utility and scope of optical imaging in scientific research.