Microscopic specimen delineation using auto-phase correlation index

Rapid and high-quality digital acquisition of microscopic specimens is critically dependent upon an initial low-resolution delineation of the specimen and the construction of a focus map to guide the subsequent high-resolution scan. In this paper, we propose a novel field-of-view (FOV) based algorithm that not only accurately delineates the specimen, but also highlights superior FOVs from which to build the focus map. Specifically, we evaluate two FOV evaluation metrics: one “baseline” method based on a threshold of intensity values in each FOV and the other based on the proposed auto-phase correlation index (APCI). While phase correlation is a well known measure of phase diversity, we show that APCI can be efficiently utilised as a metric to both delineate the specimen and highlight superior FOV focus candidates. The experimental results, on real-world cytology specimens, demonstrate that APCI is both effective and robust at specimen delineation, particularly in the presence of artifacts such as glue and dust. In addition, APCI is shown to be effective at selecting a subset of FOVs that can be correctly focused.