Improved generative models for colloidal specimens in digital holographic microscopy

Digital holographic microscopy (DHM) is an important tool for characterizing colloidal specimens. Traditionally, the inverse problem of determining the size, shape, orientation and composition of an object from a hologram obtained by DHM is approached by numerical reconstruction of the incident and scattered electromagnetic fields. Recently, we have made advances by comparing holograms to predictions of a forward model for hologram formation. For systems colloidal spheres and cylinders in which multiple scattering is negligible, the scattering problem is solvable analytically and/or numerically. These calculations provide the basis for a model of image formation in DHM. We present progress highlighting improvements to a generative model for DHM of systems of colloidal particles. In particular, we show that explicit modelling of the scattering and diffraction effects in a microscope’s optical train increase the model’s predictive power. We also investigate the challenges for extending our technique of characterization through fitting of a forward model to biological systems such as bacteria and and animal cells.