Chinese Journal of Biomedical Engineering

Home

Editorial Board

Download

Quick Search

Office Online

Current Issue

Download Articles

Download

Links

2020 Vol. 39, No. 6 Published: 2020-12-20

Reviews Communications Pegular Papers

Pegular Papers

	641	Edge Detection Method Based on the Long- and Short-Term Synaptic Complementary Networks
		Yu Xiang, Fan Yingle, Fang Tao, Wu Wei
		DOI: 10.3969/j.issn.0258-8021.2020.06.001
		The accuracy of edge detection is of great significance for improving the performance of artificial visual perception systems. In this work, a neural network with long and short-term synaptic complementarity was constructed. First, the dominant color antagonistic characteristics of cone cells were introduced, and the color antagonistic channels of the image to be tested were weighted to obtain the primary edge perception of the image to be tested. the synchronous firing characteristics of the neuron group were simulated, the synapse dynamically connected neuron action window was defined, and the group discharge time coding of the primary edge perception was realized; then a long- and short-term synaptic complementary module was built based on the synchronous firing of the neuron group in the short-term characteristics and the temporal and spatial dependence of neuron firing activity in the long-term, to achieve long- and short-term synaptic plasticity coding and complementary fusion. At last, the edge response by encoding the temporal information stream was obtained. The 20 pairs of colony images collected by the laboratory according to the needs of routine microbiological experiments were used as experimental materials, and the reconstructed similarity MSSIM, edge confidence BIdx, and comprehensive index EIdx were used as evaluation indicators. Results showed that, compared with the three mainstream methods of VSC, NIS and MSP, the detection results of the algorithm of this study had accurate edges and a low missed detection rate, which was consistent with the results of artificial subjective observations; meanwhile, the mean and standard deviation of the three indicators of MSSIM, BIdx and EIdx was 0.909 6±0.037 7, 0.671 5±0.105 7, and 0.804 8±0.052 1 respectively, and the overall performance was better than the above three mainstream methods. The method provided a new idea for realizing the construction of visual perception computing model and its application in image processing by simulating the long- and short-term synaptic complementarity of the neuron population.
		2020 Vol. 39 (6): 641-651 [Abstract] ( 350 ) HTML (1 KB) PDF (20133 KB) ( 424 )

	652	Semantic Segmentation of Subcortical Brain Structures Based on DenseMedic Network
		Yang Binbin, Liu Linwen, Zhang Weiwei
		DOI: 10.3969/j.issn.0258-8021.2020.06.002
		Subcortical segmentation is the basis for computer-aided diagnosis and treatment of central nervous system diseases. By segmenting and analyzing the brain structures in MRI image, early diagnosis and treatment of diseases such as autism spectrum disorder, stroke, and brain tumors can be performed. In order to solve the problem of accurate subcortical segmentation, based on the basic theory of deep learning, an algorithm named DenseMedic for subcortical segmentation on MRI image is proposed. First, the OreoDown method increases the growth rate of the characteristic receptive field by increasing the stride of convolutions in early layers, and uses convolutions with constant input and output sizes to restore the network depth in a sandwich-like manner, so that the increase in growth rate brings an effective receptive field increase. Second, DenseMedic uses the idea of DenseNet to instantiate the OreoDown framework. Multi-scale context information is obtained through densely connected feature extracting operations. Finally, hybrid dilated convolution is utilized in each layer to further expand the receptive field and solve the problem of rough feature extraction. Four metrics namely Dice similarity coefficient (DSC), Intersection over Union (IoU), 95% Hausdorff surface distance (HSD95) and the average surface distance (ASD) were used to evaluate the segmenting performance of the neural networks. Experiments perform on the public IBSR dataset (18 subjects of images), in which DenseMedic reached 89.2%, 80.7%, 1.982 and 0.882 respectively in 4 metrics; experiments perform on the public MRBrainS18 dataset (7 subjects of images), in which DenseMedic reached 88.7%, 79.8%, 1.249 and 0.570 respectively in 4 metrics. The experimental results show that the segmented subcortical structures and corresponding ground truths have more overlaps in regions and more similarities in outlines, which indicates that DenseMedic can effectively accomplish the segmentation of major subcortical structures. In clinical applications, the presented DenseMedic will help to accurately measure the key indicators for the central nervous system related diseases and provide rapid computer-aided diagnosis and treatment.
		2020 Vol. 39 (6): 652-666 [Abstract] ( 400 ) HTML (1 KB) PDF (14862 KB) ( 123 )

	667	Study on the Correlation between Imaging Features and Gene Expression in Non-small Cell Lung Cancer
		Wang Ting, Gong Jing, Duan Huihong, Wang Lijia, Nie Shengdong
		DOI: 10.3969/j.issn.0258-8021.2020.06.003
		Radiogenomics combines the complementary advantages of radiomics and genomics by mining the association of them to guide the development of individualized treatment regimens, prognosis evaluation and efficacy detection for different patients. This paper established the mapping between quantitative characteristics of CT images and gene expression for non-small cell lung cancer (NSCLC). Firstly, the tumor regions in the corresponding CT images were segmented and features were extracted. We selected 66 kinds of three-dimensional quantitative features as the imaging feature set of the tumor area. Secondly, the first principal component was obtained as the representative of the clustering results with similar expression profiles after preprocessing and clustering the original genetic data by using genomics data analysis process. Finally, the algorithm about significance analysis of microarray was used to explore the correlation between imaging features and gene expression. We also carried out the enrichment analysis of gene sets and established the prediction models. The 26 cases of NSCLC image data from this study were selected from the Cancer Imaging Archive (TCIA) and the corresponding genetic data were derived from the Gene Expression Omnibus (GEO). Analysis of these data revealed a significant association of 126 pairs (q<0.05). Prediction models were established for 29 sets of genes in the obtained results. In addition, the updated 211 sets of data from TCIA were used to verify the prediction model with the predicted metagenomic significance in 10 of the 29 groups, whose prediction accuracy was more than 70%. In addition, 10 predictive models with prediction accuracy of more than 70% and biological significance were verified by 211 groups of updated data in TCIA. The final prediction accuracy was 35.48～80.85% and the accuracy of six of the 10 prediction models was above 70%. These experimental results showed that the specific image features or their combination could be used as image markers of gene expression.
		2020 Vol. 39 (6): 667-675 [Abstract] ( 411 ) HTML (1 KB) PDF (1438 KB) ( 414 )

	676	The Specific Index Model of Resting-State fMRI Functional Connectivity in the Application to the Evaluation of Cognitive Score of Healthy Elderly
		Guo Zhitong, Ge Manling, Zhang Fuyi, Song Zibo, Xie Chong, Yang Zekun
		DOI: 10.3969/j.issn.0258-8021.2020.06.004
		The resting-state functional magnetic resonance imaging (rfMRI) has more advantages than the traditional scale test and task-state fMRI, especially in the cognitive detection on the elderly. However, functional bio-markers of healthy brain aging remain not totally clear. Herein, a specific index model derived from functional connectivity was proposed, aiming to study the possibility to identify the excellent or poor cognitive scores of the healthy elderly by the index model, and to seek the potential functional bio-markers to evaluate the cognitive scores by rfMRI instead of the traditional scale testing. A total of 98 healthy old people and a total of 90 healthy young people were volunteers. The former came from a cohort study of cognitive function of healthy elderly people in Portugal. According to the cognitive scores estimated by the scale tests before the rfMRI scan, 55 subjects with the excellent scores and 43 subjects with the poor scores were involved in the experiment group; the latter data came from GSP opened by the Harvard Hospital, the subjects aged between 18-35 years old with a medium level of cognition evaluation tested before rfMRI, were involved in the control group. After pre-processing the rfMRI data, the functional connectivity (FC) was computed on the whole brain one by one, then a FC-based specific index model was built up to estimate the FC deviation degree of old people relative to that of youth at a single brain area. Furthermore, the specific index values of the marked brain areas sensitive to the excellent scores and the poor scores could be estimated by statistics in a comparison study, by which the eigenvectors matrix were formed and input the machine learning model thereafter. Finally, the model of probability neural network (PNN) was utilized to classify the scores in the experimental groups and then the sorting rate was defined by N-fold validation. The specific index model could localize the functional bio-markers brain regions sensitive to the cognitive scores of healthy aging. There were 5 brain regions in the frontal lobe, temporal lobe and parietal lobe. By considering the indexes of 5 brain regions as inputs to the machine learning, the cognitive scores of healthy aging could be effectively classified with a sorting rate of 81.7%. This work was expected to provide an effective index and a new method for rfMRI to test the cognitive scores of the healthy elderly by combining the specific index modeling with a machine learning model.
		2020 Vol. 39 (6): 676-684 [Abstract] ( 375 ) HTML (1 KB) PDF (3481 KB) ( 311 )

	685	Analysis of MI Brain Causal Networks Under Different Visual Stimuli Guidance
		Bian Yan, Zhao Li, Fu Xing, Qi Hongzhi
		DOI: 10.3969/j.issn.0258-8021.2020.06.005
		Brain-computer interface (BCI) based on motor imagery (MI) is believed to be a potential approach for motor rehabilitation. However, classical MI-BCI leaves the questions of large individual differences and low recognition accuracy. By making use of visual stimuli guidance could enhance MI features and improve BCI recognition accuracy. Nevertheless, rare attention has been paid to the features of brain causal networks during MI under different visual stimuli paradigms and their impacts on motor recovery. We designed four different types of visual stimuli guidance in this paper, including dynamic/non-dynamic stimuli and simple/complex MI tasks. The beta one - tailed one - sample t test (P<0.01) causal significant connectivity networks of seven regions of interest located in motor-sensory cortex were built based on isolated effective coherence (iCoh) during MI, and parameters of degree distribution, clustering coefficient, global efficiency and betweenness centrality were further analyzed. The outcomes showed that compared with non-dynamic and simple MI task experimental paradigm, the average degree distribution of dynamic visual paradigms combined with complex MI task was varied from 2.143 to 2.429; the clustering coefficient was varied from 0.643 to 0.767; the global efficienciewas varied from 0.393 to 0.417. Under dynamic with complex task paradigms, the significant connectivity exist between SMA and SPL, IPL, thus SMA becomes the key node in the brain causal networks.
		2020 Vol. 39 (6): 685-692 [Abstract] ( 399 ) HTML (1 KB) PDF (7174 KB) ( 119 )

	693	An Asynchronous Control Strategy Based on Hybrid BCI
		Xu Wei, Zhao Yawei, Qi Hongzhi
		DOI: 10.3969/j.issn.0258-8021.2020.06.006
		Asynchronous control problems of automatically identifying whether users are controlling the system is a focal point of research in BCI. Traditional paradigms induce EEG characteristics by distinguishing different instructions, causing the limitation of the information, so it is difficult to obtain ideal results of asynchronous identification. Aiming to resolve the problem, we developed a hybrid BCI system containingevent-related potential and steady-state visual evoked potential signals to promote the effect. In the study, we collected 19 channel EEG signals from 10 subjects under the control condition when the subjects used the BCI system normally and two idle conditions when subjects moved their eyes out of the screen and when they opened their eyes and rested. In order to identify the different conditions, we extracted the amplitude of the ERP and the correlation coefficient of SSVEP, then we estimated the posterior probability of the control state by ERP and SSVEP features, using a Bayesian method. Using 10 subjects' EEG features, we found out that the hybrid paradigm performed better than a single paradigm, because the differences under the two paradigms coexisted. We achieved the accuracy of 92.1% and an AUC of 0.98 in identifying the control condition and the idle condition. This study demonstrated that hybrid had the ability to improve asynchronous identifying and it was worthy of a further study and development.
		2020 Vol. 39 (6): 693-699 [Abstract] ( 320 ) HTML (1 KB) PDF (3651 KB) ( 279 )

	700	Epileptic States Recognition using Transfer Learning and Dilated CNN
		Shen Lei, Geng Xinyi, Wang Shouyan
		DOI: 10.3969/j.issn.0258-8021.2020.06.007
		The automatic detection and seizure diagnosis of EEG signals in patients with epilepsy is of great significance for clinical treatment of epilepsy. Aiming at solving the problem in the conventional method that the labeled training data volume is insufficient and the classification accuracy of seizure is low due to the inconsistent distribution of training and test data, a joint knowledge transfer method between domains was proposed in this work. Firstly, the EEG signal was decomposed by four-layer wavelet packet, and the wavelet packet decomposition coefficients of 16 frequency bands were extracted as features. The marginal and joint distribution iterative adaptation were used to complete the knowledge transfer between the source and target domain. The dilated convolutional neural network was trained to complete the target domain recognition. In this study, the algorithms were estimated on two public EEG datasets including CHB-MIT dataset (22 patients, 790 hours' recording) and Bonn dataset (5 groups, one hundred 23.6 s episodes in each group). The experimental results showed that the average recognition accuracy, sensitivity and specificity of the proposed method for different epilepsy states was 96.8%,96.1%,96.4% on CHB-MIT dataset respectively. The average recognition accuracy was 96.9% on the Boon dataset, which effectively improved the comprehensive performance of seizure detection and achieve reliable detection results.
		2020 Vol. 39 (6): 700-710 [Abstract] ( 444 ) HTML (1 KB) PDF (3335 KB) ( 528 )

	711	Research on Brain Dynamics of Precision Grip Force Control
		Zhang Na, Li Ke, Hou Ying, Zhang Dongmei, Wei Na
		DOI: 10.3969/j.issn.0258-8021.2020.06.008
		Precision grip is the basis of many delicate and complex operations, it contains complex neural control mechanism. However, less is known about the dynamics of the brain during precision grip force regulation. This study aimed to explore the finger force behaviors and electroencephalogram (EEG) characteristics at different force amplitude during precision grip. Twelve healthy subjects were asked to grasp precisely at 10%, 20% and 30% of maximum voluntary contraction (MVC), during which the force signals, the center of pressure (COP) of thumb and index digits and the EEG signals were recorded and then analyzed by coefficient of variation (CV), the COP velocities and COP areas, recurrence quantification analysis (RQA).Results showed that there was a linear positive correlation between CV and force level (thumb: r=0.624, P<0.001; index: r=0.721, P<0.001). The COP areas of thumb and index fingers at 30% of MVC were (1.94±1.21) and (2.02±1.45) mm², which were significantly higher than that at 10% ((1.01±0.81), (0.89±1.02) mm²) and 20% of MVC ((1.20±0.62), (1.16±0.63) mm²,P<0.05). For the thumb fingers, the COP velocity values of x-axis and y-axis under 10%, 20% and 30% of MVC were (4.23±1.11), (2.11±0.50), (1.70±0.40) mm/s and (6.22±1.45), (3.39±0.70), (2.90±0.69) mm/s respectively, showing a decreased tendency with the force level increased (P<0.01). For the index fingers, the COP velocity values of x-axis and y-axis at 10% of MVC ((4.95±1.34) and (7.04±1.75) mm/s) were significantly higher than that at 20% ((2.78±0.53) and (3.79±0.63) mm/s) and 30% MVC ((2.95±0.94) and (3.54±0.82) mm/s, P<0.05). The RQA parameters of α-band EEG signals decreased significantly with the force level increased (P < 0.05). These results suggested that the variability of force, the reduction of the digits’ adjustment speed and control instability, the complexity of EEG signals was increased with the force level augmented. Furthermore, α-band of EEG was closely related to the motor control of precision grip. This study revealed the close coupling between finger force control and dynamic behavior of the central nervous system, which provides a new path for the in-depth study of the central peripheral cooperative mechanism and the quantitative evaluation of the neuromuscular system function.
		2020 Vol. 39 (6): 711-718 [Abstract] ( 377 ) HTML (1 KB) PDF (5762 KB) ( 290 )

	719	Arrhythmia Beat Classification Model Based on CNN and BiLSTM
		Yang Hao, Huang Maolin, Cai Zhipeng, Yao Yingjia, Li Jianqing, Liu Chengyu
		DOI: 10.3969/j.issn.0258-8021.2020.06.009
		In order to extract the abnormal beats more accurately from the dynamic electrocardiograph (ECG), a deep learning model combining convolutional neural network (CNN) and bi-directional long short-term memory network (BiLSTM) was proposed in this study. Firstly, ECG signals were segmented into two types of time window lengths: a small-scale length of 0.75 s and a large-scale length of 4 s. Then, features were extracted from the small- and large-scale length ECG segments using an 11-layer CNN network and a 3-layer BiLSTM network, respectively. Finally, the extracted features were combined and were then reduced using a 3-layer fully connected network. In addition, two data enhancement methods by adding random motion noise and baseline drift were used to attenuate the influence of over-fitting due to the unbalanced data distribution. The proposed model was tested on the MIT arrhythmia database using a patient-based 5-fold cross-validation method, and its accuracy for classifying the 4 types (normal, atrial premature, ventricular premature and unclassified) on 116,000 heartbeats was 90.42%, which was 13.97% and 7.14% higher than the CNN model (76.45%) and BiLSTM model (83.28%), respectively. This study validated that the proposed model with combining CNN and BiLSTM reports higher accuracy than only using CNN or BiLSTM model when performing the abnormal beat classification task.
		2020 Vol. 39 (6): 719-726 [Abstract] ( 440 ) HTML (1 KB) PDF (2718 KB) ( 360 )

Reviews

	727	Electric Field Analysis Progress of Transcranial Magnetic Stimulation Device
		Xia Siping, Xu Yajie, Yu Yingcong, Gu Weiguo, Ma Changyu, Yang Xiaodong
		DOI: 10.3969/j.issn.0258-8021.2020.06.010
		Transcranial magnetic stimulation (TMS) is a cortical regulation technique that utilizes the induced electric field in brain resulting from the stimulation current. The technique is now widely applied in the treatment of neurology, rehabilitation science, etc. Analysis of electric field induced by TMS has been a hot spot and has play an important role in TMS related safety issues and stimulation effect. It’s also necessary in stimulation scheme optimization and coil design. In this paper, we firstly introduced regular clinical side effects of TMS, and summarized the conventional electric field analysis methods in the current TMS research. Numerical methods and analytical methods that have been applied in TMS electric analysis were included along with their application scenarios, followed by the discussion of physiological modeling methods in TMS. In addition, due to the close relationship between the magnetic stimulation coil and the electric field distribution in brain tissue, this paper introduced several critical structures of TMS coil. Electric field distribution characteristics of several typical designs of the magnetic stimulation coil were simulated and analyzed with the spherical model based on finite element analysis software. In the end, the future development tendency of TMS electric field analysis was prospected.
		2020 Vol. 39 (6): 727-735 [Abstract] ( 384 ) HTML (1 KB) PDF (2439 KB) ( 438 )

	736	Large-Scale Brain Networks Interactions Support Internal and External Directed Cognition
		Xin Fei, Xie Chao, Wang Lijun, Lei Xu
		DOI: 10.3969/j.issn.0258-8021.2020.06.011
		Converging evidence has indicated that the high-level cognitive functions are carried out through the dynamic interactions among large-scale brain networks instead of stand-alone brain regions. Among them, the frontoparietal control network plays a pivotal gate-keeping role in goal-directed cognition, modulating the dynamic balance between the dorsal attention network and the default network. The present article reviewed the advances of this field from several aspects, including the neuroanatomy of the default network, the dorsal attention network and the frontoparietal control network, as well as their respective functional roles and dynamic interactions in the internal- and external- directed attention tasks. Future research needs to further explore the functional roles of the subsystems within each network and uses the effective connectivity method to examine the direction and dynamics of the information transmission within- and between- networks.
		2020 Vol. 39 (6): 736-746 [Abstract] ( 372 ) HTML (1 KB) PDF (2631 KB) ( 596 )

	747	Research Development of Electroencephalogram Acquisition Devices for Brain Computer Interface
		He Qing, Hao Sicong, Si Juanning, Wu Yingnian, Cheng Jie
		DOI: 10.3969/j.issn.0258-8021.2020.06.012
		Brain computer interface (BCI) builds a novel and direct interactive mode between human brain and the outside, so that it has a very broad application prospects. Electroencephalogram (EEG) acquisition device, as an important means and approach of signals acquisition for BCI, is the critical point and foundation of the BCI technology and has received intensive attention. Owing to the explosive growth of the BCI research recently, various techniques of EEG acquisition spring up constantly. In the future, EEG acquisition devices will have huge application potential in the fields of science, medicine, military and human life. In order to clarify the development status and development direction of EEG acquisition devices, this paper discussed basic structure, performance optimization circuit and existing EEG acquisition products. Firstly, four main components of EEG acquisition equipment were analyzed. The performance optimization method of EEG acquisition equipment was classified and summarized, the key indicators of the existing mainstream EEG products were compared, their functional characteristics were expounded. Finally, the shortcomings of the existing EEG acquisition equipment were analyzed and its future development tendency was prospected.
		2020 Vol. 39 (6): 747-758 [Abstract] ( 967 ) HTML (1 KB) PDF (5437 KB) ( 1867 )

Communications

	759	Attention State Detection Based on Feature Encoding and Convolutional Neural Networks
		Wu Ruoyou, Wang Dexing, Yuan Hongchun, Gong Peng, Qin Enqian
		DOI: 10.3969/j.issn.0258-8021.2020.06.013

		2020 Vol. 39 (6): 759-763 [Abstract] ( 310 ) HTML (1 KB) PDF (3054 KB) ( 305 )

	764	Influence of Low Frequency Rhythm Source Direction on Dynamical Parameters of Field Potentials on Surface of A Quasi-real Head Model
		Ge Manling, Yang Zekun, Cui Jiajun, Guo Zhitong, Yang Minghao, Zhang Fuyi
		DOI: 10.3969/j.issn.0258-8021.2020.06.014

		2020 Vol. 39 (6): 764-768 [Abstract] ( 235 ) HTML (1 KB) PDF (6532 KB) ( 205 )

Copyright © Editorial Board of Chinese Journal of Biomedical Engineering
Supported by:Beijing Magtech