State-space Models And Transmission Mechanism In The Analysis Of Acupuncture Neural Data

One of the critical scientific problems in research of acupuncture mechanism isthat of establishing the quantitative relationship between acupuncture stimuli and thespike responses of neural system, i.e. neural encoding and decoding of acupuncture. Afull understanding of the neural code will mean that we can predict the responses ofneurons to arbitrary novel stimuli, and can decode the information contained in spiketrains to reveal the stimuli that gave rise to them. Models of the neural code, as thetache between neurophysiological experiments and computational neuroscience,provide mathematical carrier for understanding of acupuncture coding mechanism. Inthis thesis, we pursue the construction of model for modeling the neural responsesevoked by acupuncture. We focus specifcally on studying encoding form ofacupuncture stimulation and analyzing conduction and effect of the acupunctureneural electrical signals.This thesis provides descriptions of the characteristics of manual acupuncturemanipulation and the stimulation attributes of acupuncture for sensory neural system.According to the feedforward pathway of acupuncture neural electrical signal,experiments are designed that manual acupuncture (MA) manipulations with diferenttypes and frequencies are taken at’Zusanli’ points of experiment rats. Neuralelectrical signals evoked by acupuncture are recorded and reprocessed. This workstudies mechanism of acupuncture according to the experimental data.In the frst portion of the thesis, the acupuncture neural spike trains aretransformed into point process trains which are depicted by conditional intensityfunction. Then we view the acupuncture stimulus as implicit stimulus and the pointprocess as observation equation, and construct a state-space model with latent processfor describing neural encoding process of neural spiking activity induced byacupuncture stimulus. A Bayesian decoding algorithm based on state-spacerepresentations of point processes is used to reconstruct stimulus oscillograms ofacupuncture. Meanwhile, good-of-fit of the encoding model and precision of thedecoding algorithm is evaluated based on the time-rescaling theorem.In the second portion of the thesis, the acupuncture neural spike trains aretransformed into inter-spike interval trains. The inter-spike interval statisticalhistogram is fitted by Gamma distribution, whose shape parameter is the measure of spiking irregularities. Analyses suggest that different acupuncture manipulations arerelated to different spiking irregularity measures. Furthermore, we estimate fringirregularity and the fring rate simultaneously for neural spike sequence evoked byacupuncture based on state-space model, and we convert the firing characteristics intoinput parameters using a transformation formula. Differences exist among differentacupuncture manipulations according to the characteristics.In the last portion of the thesis, the transmission pathway of sensory neuralelectrical signal can be abstract to feedforward neuronal network, and we constructthe acupuncture neural electrical signal transmission model. The critical parameters offeedforward neuronal network are estimated based on the adaptive synchronizationcontrol scheme. Moreover, the acupuncture stimulation is equivalent to low-frequencyinput signal. The high-frequency driving and noise can induce information code,signal transmission and resonance in feedforward neuronal network, which may bethe underlying mechanism of acupuncture.This work combines the neurophysiological experiments and computationalneuroscience to investigate acupuncture, which provides new perspective foracupuncture research and has important implications for improving the clinicaltreatment of acupuncture.