Research On Neural Metabolic Mechanism Of Negative Signals Based On MEA And OI

Functional imaging of brain activity consists of two types, one is positive response signals, and the other is negative response signals. The origin of the former one is well known for researchers. However, the current explanations for the negative signals remain controversial because of lack of direct evidence and the conflict experimental results. The mechanism of negative signals and its couple relationship with neuronal activity, could impove the positioning accuracy and information mining capabilities in functional magnetic resonance imaging, but also help to the study of"attention"and"functional connectivity".This dissertation, based on the micro electrode array (MEA) and intrinsic optical imaging (OI), concentrates on the neural metabolic mechanism of negative signals. With the model of somatosensory cortex of rat, the dissertation investigates the neural spike activity, measures the positive and negative hemodynamics reponse (PHR and NHR), and analyzes the spatio-temporal characteristics of PHR and NHR. The above results of neural and hemodynamic analysis provide direct evidence that the deactivation of spike activity has close relationship with NHR signals. The main contributions can be exhibited as follows.1. The feature extraction algorithm of neural spike data sets, based on the theory of locality preserving projections (LPP), is proposed. The activity state of each neuron could be obtained after spike classification, and the feature extraction is indispensable basis of classification. In this dissertation, based on LPP theory, an unsupervised feature extraction algorithm is introduced. The automatic neighbor parameter selection is presented, and the distribution separated standard based on the original data sets is used. The application in both simulation data and real experimental data sets show that the proposed method outperforms the traditional principle component analysis (PCA) on extracting and separating features.2. The spike classification algorithm, based on theories of finite mixture model (FMM) and expectation maximum (EM) algorithm, is proposed. The spike sorting and classification is important prerequisite for the analysis of neural activity. Inspired by the theory of FMM and EM, scaled gradient EM (SGEM) algorithm is presented for spike classification. The SGEM algorithm utilizes scale step for the search procedure of likelihood function in spike feature space. Maintaining the same results as the original one, SGEM algorithm has more faster convergent rate. As the estimation of scale is complicate, the empirical range is provided by simulation.3. The noise reduction algorithm, based on theories of Gaussian white noise and canonical correlation analysis (CCA), is proposed. Because OI data sets have low signal to noise ratio (SNR), the negative signals in OI data sets is difficult to be detected or explained. The application in both simulation data and real optical imaging data showed that, this method could significantly minish the influence of structure noise, which including the breath and heartbeat signals. After the noise reduction, the detection and space allocation of negative signals is analyzed.4. The direct evidence of neural deactivation mechanism of negative signals is presented. The results of OI data sets show that, while stimulated by current impulse in hindlimb paw, the expected PHR signal is observed in contralateral primary somatosensory hindlimb cortex (SIHL), and that the NHR signals is observed in both contralateral and ipsilateral primary somatosensory forelimb (SIFL), primary and secondary motor (M1M2) and primary and secondary visual (V1V2) cortex areas. The NHR signals in different cortical areas had similar time courses but were in the opposite direction of the PHR signals, and the amplitude of NHR signals was significantly smaller than that of PHR signals. Meanwhile, the MEA recordings in visual cortex area show that spike activities decrease significantly during external stimulation. The above results suggest that the neural activity reduction has a strong relationship with the NHR signals. The OI and MEA results of this dissertation provide direct evidence for the neural deactivation origin of negative signals, but also highlight the importance of a negative response in a hemodynamics-based interpretation of neuroimaging signals.