FNIRS Analysis Method And System Design For Infants

With the development of Functional Near-Infrared Spectroscopy(fNIRS)brain imaging technology and the continuous upgrading of commercial imaging equipment,especially for neonatal brain development problems,fNIRS data analysis methods still need to be continuously enriched and developed.Compared with the time-series amplitude analysis method based on oxyhemoglobin or deoxyhemoglobin,the time-delay estimation and phase synchronization analysis of the fNIRS signal is currently a novel research direction.Existing fNIRS brain imaging data analysis toolkits,such as NIRS-SPM,Homer,fOSA,NINPY,NAP,and FC-NIRS,although they have analysis functions such as basic signal spectrum analysis,noise removal,or brain functional connectivity,they are mostly based on time series.Amplitude and simple linear correlation analysis are the basis of brain functional connectivity analysis and cannot fully characterize cerebral oxygen metabolism and assess neonatal cerebral oxygen metabolism.Therefore,this paper studies the fNIRS analysis method from the perspective of fNIRS data preprocessing,time delay estimation and phase analysis,and analyzes the fNIRS data of newborns and adults.Finally,a portable fNIRS brain functional imaging system was designed as follows:First,the more important motion noise filtering,frequency band extraction,time delay estimation and phase synchronization problems in newborn fNIRS monitoring are analyzed.For motion noise filtering,a kurtosis-based wavelet filter(Kurtosis-based wavelet filter,kbWF)was used,and the performance of this method and other three motion noise filtering algorithms was compared.A simulation signal was generated based on the typical hemodynamic response function.The effect of the motion noise filtering algorithm was evaluated from the three indicators of mean square error,signal to noise ratio and correlation coefficient.The results show that the kbWF algorithm has the best comprehensive effect.For frequency band extraction,this paper proposes an infinite impulse response filter and zero-phase digital filter to calculate the convolution scheme,and compares the finite impulse response filter and harmonic wavelet filter based scheme.The results show that the infinite impulse response filter and the zero-phase digital filter calculation of the convolution scheme have the best effect.For the time delay problem,this paper compares the performance of two time delay estimation methods.The results show that the cross-power spectral density method has a better estimation effect.For the synchronization analysis method,the average phase difference algorithm is proposed.Secondly,this article aims at the specific newborn and adult fNIRS data,uses the kbWF algorithm to filter out the motion noise,and then extracts data from the 0.01-0.05 Hz,0.05-0.1 Hz,and 0.01-0.1 Hz frequency bands.For the 0.05-0.1 Hz band signal,the delay parameters and the vector average phase difference index of the premature infant group,the full-term infant group and the adult group were analyzed.The results showed that the vector average phase difference index pVAPD can distinguish the cerebral oxygen metabolism status of adults and newborns.Statistical analysis using the Watson-Williams test showed a significant difference in mean mean phase difference between the three age groups(F[2,27] = 11.513,p<0.001).Delayed deoxyhemoglobin parameters have the potential to distinguish preterm and full-term infant groups.Finally,a portable wireless fNIRS imaging system was designed to meet the needs of clinical neonatal data collection,and supporting high-reliability data acquisition software was developed.This device was compared with a commercial NIRScout device and the results showed that the performance was close to that of the commercial device.