Method To Identify Biological Signals In Fluorescence Microscopy Imaging

Fluorescence microscopy offers many advantages:noninvasive imaging、specific labeling and being suitable for living cell imaging. It has become a basic tool in biological and medicine research. Recently, localization-based super resolution imaging has broken the diffraction limited resolution. Its spatial resolution can achieve 20-50 nm, which has a widely application in different fields, especially for the imaging on micro structure of living cell. Two photons activation-based Ca2+ imaging can simultaneously record Ca2+ signals from several tens of neurons with single-cell spatial resolution and milliseconds temporal resolution, which make it possible to reconstruct neuronal microcircuits on the time scale of milliseconds.Identification of biological signals plays a key tool in these two kind of fluorescence microscopy. In localization-based supper resolution, identification of biological signals, i.e. localization for fluorescence molecules, is an important part in the process of imaging. The results of the identification can seriously influence on the spatial and temporal resolution of the imaging. In Ca+ imaging, Identification of biological signal, i.e. reconstruction of neuronal spiking, provide indispensable tool to reconstruct neuronal microcircuits, which can significantly boost the value of Ca2+ imaging. Localization for fluorescence molecules and reconstruction of neuronal spiking come from two different fields. Intrinsically, they can attribute to the problem that in the pulsed input-output system, the pulsed input is inferred from the output. Even if the mechanism of pulsed input-output system is clear, this is still a NP problem that computation load increases exponentially with the computing dimensions. In the identification of biological signal, it is avoidable the cases including low signal noise ratio experimental signal, several gigabytes level datasets, and lack of sufficient kwnowleges of the pulsed input-output system These cases make the identifaiction keep challenges. Aimed to thses challenges, a series of methods have been developed.(1) According to the features of signals in supper resolution and Ca2+ imaging, the identifications of these two biological signals is attributed to the same model. For the different conditions in this model, namely, identification of spares signal, linear superimposed signal and nonlinear superimposed signal, the corresponding methods has been presented.(2) Identification of sparse signal. For sufficient sparseness of the identifying signals, their locations are roughly estimated by the features of the observed signals; and then, using the estimatied locations, the corresponding durations of signals are extracted from the observed signals; finally, combining these extracted signals and the input-ouput system, the loacations of these identifying signals are precisely estimated. We use this method to analyze the images in super resolution imaging. The results indicate that the the acquired images can be online analyzed even fast EMCCD camera works at full frame rate, which provides a strongle tool for monitoring the process of the imaging.(3)Identifacation of linear superimposed signal. Firstly, the candidates of the identified signal are ascertained based on the features of the observed signals. Secondly, the optimal signal is selected from the candidates using the input-output system. The efficiency of this method is verified by two examples including localization of high-density molecules and reconstruction of neuronal burst fiing. In supper resolution imaging, compared to the typical method, this method can give a six times improvement in the molecules density, which extends the applications of super resolution imaging and depicts a prospect of the promotions of 10 times in temporal super resolution imaging. In Ca2+imaging, neuronal burst firings are successfully reconstructed from the Ca2+ signal with low signal noise ratio, which is beneficial for observing the complex patters of neuronal microcircuits.(4) Identifacation of nonlinear superimposed signal. The identification of this signal is converted to solving the reconstructed optimization problem. In the optimization problem the impulse model of the input are fully considered and a new mathematic tool is introduced to solve the optimization problem. This method keeps a markedly high level of immunity to noise and to nonlinear response of input-output system and can successfully reconstruct neuronal burst firing from the Ca2+ signal with extremely low signal noise ratioIn this dissertation, based on a combination of the profound analysis of the features of the signal and appropriate mathematic tools, we present a series of methods to identify the biological signals in fluorescence microscopy. The results indicate that the proposed methods are effective and valueable.