CIF:Small:Developing Theory of Spatiotemporal-Resolution and Spatiotemporal-Localization Algorithms for Single-Molecule Localization Microscopy

Cuny City College, New York NY

Investigators

Abstract

Single-molecule localization microscopy (SMLM) overcame the longstanding light-diffraction barrier to provide super-resolution optical imaging, significantly impacting biological research. Nevertheless, two important problems remain before SMLM can be advanced further. First, still needed is a theory of spatiotemporal resolution which speaks to the inherent power of an SMLM system in resolving a molecule in space and time, thereby quantifying how each part of an SMLM system affects the information-theoretical limit of spatiotemporal resolution. Thus, a spatiotemporal resolution theory would lay an information-theoretic foundation and provide guidance in the development of electro-optical hardware, fluorescence molecules, and localization algorithms to advance SMLM. Second, the majority of localization algorithms in literature exploit the information of only a single data frame. Advanced spatiotemporal localization algorithms need to be developed to fully exploit the information of the multiple frames of a data movie in order to approach the information-theoretic limit of spatiotemporal resolution. Advanced spatiotemporal localization algorithms can significantly enhance spatiotemporal resolution and fulfill the needs of both super spatial and temporal resolutions in biomedical research. The spatiotemporal resolution theory and spatiotemporal location algorithms to be developed in this project will broadly impact interdisciplinary research of SMLM in both theory as well as in applications to biomedical research. In this project, first, a conceptually novel theory of spatiotemporal resolution with respect to one-dimensional (1D), 2D, and 3D spatial resolutions will be developed based on the Fisher information of a universal model of a data movie. The effect of system parameters on the spatiotemporal resolution and the tradeoff between spatial and temporal resolutions will be analytically and numerically investigated. The unbiased Gaussian information-achieving estimator that achieves the Fisher information of a data movie will be developed and simulated. Second, two types of advanced spatiotemporal localization algorithms will be developed to fully exploit the spatial and temporal information in a data movie. One is the maximum movie-likelihood (UGIA-M) algorithm that maximizes the likelihood of an entire data movie. The other is the temporal correlation-enhancement algorithms that exploit the temporal correlation embedded in the frame-by-frame localized SMLM images. The two types of algorithms will be evaluated via simulation in terms of spatiotemporal resolution using a universal partition-based metric driven by root-mean-square minimum distance and root-mean-square error with respect to a UGIA-M benchmark as well as by comparison to existing high-performance localization algorithms from the literature. The two types of algorithms will be applied to the analysis of real datasets of biological specimens, and source code for the developed theory and algorithms will be posted in public repositories for open access by the community. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.