A Research On Analysis Method Of Functional Near-infrared Spectroscopy Based On Attribute Partial-ordered Theory

The brain is an important organ in the human central nervous system, giving human intelligence and emotion, and many unique behavior capabilities. However, human knows a little about the operation mechanism of the brain and brain network, at the same time, a variety of brain function disorders such as Parkinson, epilepsy, etc. are threatening the health of many people. Recent years, promoted by non-invasive brain imaging(monitoring) techniques, in the field of neuroscience, medicine and psychology the researches for brain function achieved a remarkable progress. Compared with the functional magnetic resonance imaging and other common brain function monitoring technologies, functional near-infrared spectroscopy not only achieves higher temporal and spatial resolutions, but also ensures the practical requirements as non-invasive, equipment portable and economical etc. Through spontaneous measurement of brain hemodynamic changes, functional near-infrared spectroscopy signal directly reflects brain especially the metabolic activity of cortex in particular oxygen, realizing the monitoring of different brain functional states. As a new type of brain function monitoring technology, functional near-infrared spectroscopy technology is commonly used in brain researches. Nevertheless, the signal or image processing analysis methods of functional near-infrared spectroscopy lack unified standards, in the actual brain function researches there are a lot of big data analysis and processing problems. Therefore, this paper focuses on the analytical methods for brain hemodynamic response signal of functional near-infrared spectroscopy, mainly as listed in the following.First, the basic optics principles used by functional optical near-infrared spectroscopy are introduced through explanation of the human body’s main components absorption and scattering of near-infrared light. The basic hemodynamic measurement and extraction method of functional near-infrared spectroscopy technology are explained on the basis of relevant principles of Beer-Lambert law correction and differential path length. In the functional near-infrared spectroscopy application, the collected original functional signals will be disturbed by all kinds of noises, especially the physiological noise and motion artifacts. Thus, common motion artifacts denoise methods of functional near-infrared spectroscopy signals are briefly discussed and compared.Second, signal characteristics and statistical feature extraction method of functional near-infrared spectroscopy technology have been briefly reviewed. Then combined with multiple diagram theory, radar map of the functional near-infrared spectroscopy and variable scale figure feature extraction methods are proposed. With actual functional near-infrared spectroscopy data, the proposed method is performed comparative experiments with common feature extraction and pattern classification methods. In order to test the feature extraction method on the basis of multivariate diagram theory, the differences of pattern classification performance between traditional methods and newly proposed method have been evaluated.Thirdly, in order to expand a variable scale feature extraction method to variable scale pattern classification, an analysis method of functional near-infrared spectroscopy has been developed based on partial ordered theory. Layer-by-layer generation algorithm of structural partial-ordered attribute diagram is proposed for achieving full scale analysis of functional near-infrared spectroscopy based on visualization method, and obtains structural patterns of functional near-infrared spectroscopy data. Through the basic theories and algorithm of supervised dictionary study sparse representation, the order representation dictionary by attribute partial order is proposed for sparse representation of spectrum models, furthermore, corresponding pattern classifier tool are built up to solve pattern matching problem between test samples and attribute partial ordered structure diagram. By analysis of real functional near-infrared spectroscopy, the visualization pattern classification method has been evaluated.Furthermore, n-back prefrontal working memory load functional near-infrared spectroscopy signal is chosen for the quantitative research of prefrontal working memory. Comparative study in the n-back test between the functional near-infrared spectroscopy signal hemodynamic mean change and radar charts characteristics shows by radar charts the functional near-infrared spectroscopy signal can achieve better signal factors extraction and visualization. Quantitative research of prefrontal working memory by attribute partial order method proposed in this article reached higher classification accuracy than similar researches abroad.Finally, on the basis of above studies and clinical research for monitoring the depth of anesthesia based functional near-infrared spectroscopy, this paper presents a signal data synthetic method of functional near-infrared spectroscopy for simulation signal of anesthesia state. Such synthetic signals of anesthesia have been used in pattern classification research based on the method proposed in this study. This research provides a foundation of monitoring the depth of anesthesia based on functional near-infrared spectroscopy and the method proposed in this study.