Chinese Journal of Biomedical Engineering

Home

Editorial Board

Download

Quick Search

Office Online

Current Issue

Download Articles

Download

Links

2021 Vol. 40, No. 6 Published: 2021-12-20

Reviews Communications Regular Papers Indexes

Regular Papers

	641	Feature Fusion Based Deep Residual Networks Using Deep and Shallow Learning for EEG-Based Emotion Recognition
		Zhou Rushuang, Zhao Huilin, Lin Weiyue, Hu Wanrou, Zhang Li, Huang Gan, Li Linling, Zhang Zhiguo, Liang Zhen
		DOI: 10.3969/j.issn.0258-8021.2021.06.01
		Electroencephalography (EEG) has advantages of portability, high temporal resolution and real-time operation, therefore has been used to recognize, monitor, and track human's emotion in the fields of healthcare, entertainment, education and so on. However, due to the non-stationarity and individual differences in EEG signals, it is difficult to effectively and efficiently extract informative and useful emotion related characteristics using traditional methods. To obtain representative features in an efficient manner and improve emotion classification accuracy, we proposed a feature fusion based deep residual networks using deep and shallow learning for EEG-based emotion recognition. The proposed model consisted of three modules: a shallow feature extraction module, a deep feature extraction module, and a classification module. First, the shallow feature extraction module was designed with multiple convolution layers to extract shallow tempo-spatial features. Second, the deep feature extraction module employed a Bi-GRU layer and the attention mechanism to extract deep tempo-spatial features from the extracted shallow features. Third, the classification module was designed with a fully connected layer for binary classification. The proposed model used a subject-based leave-one-out cross-validation on the DEAP database with 76 800 samples, and achieved a good performance in emotion recognition with the binary classification accuracy of 96.95% for valence and 97.22% for arousal. Comparing to the existing methods, our proposed model increased the accuracies by 3.53% and 4.25% for valence and arousal, respectively. Further, the good performance of the proposed model in binary emotion classification was also validated in MAHNOB-HCI and SEED databases as well.
		2021 Vol. 40 (6): 641-652 [Abstract] ( 531 ) HTML (1 KB) PDF (2347 KB) ( 950 )

	653	Study on the Differences of Resting-State EEG Microstate in Children with Autism Spectrum Disorder
		Zhang Suoliang, Wan Lingyan, Zhang Zhiming, Kang Jiannan, Li Xiaoli, Pang Jiao
		DOI: 10.3969/j.issn.0258-8021.2021.06.02
		This study aimed to use the microstate analysis method to investigate the differences in brain mechanism between children with autism spectrum disorder (ASD) and healthy children on the electroencephalography (EEG) scale in the resting state. According to the guidelines in Cartool and the degree of interpretation of the participants' EEG data by the number of different microstates, the number of microstates was determined to be 4. Using atomize & agglomerate hierarchical clustering algorithms, the microstates at the individual level and group level were segmented and labeled as microstate Class A, microstate Class B, microstate Class C, and microstate Class D. Next, fit the data back to the EEG data according to the topographic maps of the four classes of microstates and the level of the GEV correlation of the EEG data at each time point, and finally obtained the microstate time series, and extracted the characteristics in the time domain to compare the differences between the ASD group and the TD group. The time parameters selected in this study included average duration, frequency of occurrence, time coverage and transition probability. And the method of calculating the Markov model explored the independence of the microstates sequence. The microstate time parameters that showed there were differences (P<0.05) in the ASD group vs. the TD group for the duration (A: 0.110±0.013 vs 0.180±0.048, C: 0.140±0.024 vs 0.220±0.067, D: 0.130±0.050 vs 0.190±0.037, unit: s), time coverage (A: 22.0±5.4 vs 27.0±7.2, B: 27.0±4.7 vs 18.0±5.5, unit: %), occurrence(A: 1.93±0.52 vs 1.55±0.22,B: 2.08±0.46 vs 1.39±0.32,C: 2.10±0.49 vs 1.47±0.30,D: 1.78±0.19 vs 1.27±0.27, unit: times/s). Moreover, chi-square test did not supported the hypothesis that the microstates was independent zero (P<0.01), suggesting the dependence and information sharing between the microstates. The results of this study provided objective indicators and scientific basis for the assessment of autism.
		2021 Vol. 40 (6): 653-661 [Abstract] ( 344 ) HTML (1 KB) PDF (4865 KB) ( 699 )

	662	Complex Networks Analysis of the Elderly People with Mild Cognitive Impairment by Nonlinear Interdependence of EEG
		Yan Yan, Xiao Shasha, Liu Meng, Li Yunxia, Li Yingjie
		DOI: 10.3969/j.issn.0258-8021.2021.06.03
		The aim of this study was to preliminarily investigate brain functional networks of the elderly with mild cognitive impairment (MCI) and the normal elderly during emotion regulation. Fourteen MCI patients and eighteen healthy subjects participated in this experiment. The scalp electroencephalography (EEG) was recorded whenthe subjects performed the cognitive reappraisal task, including neutral/negative viewing task and negative reappraisal task. We applied nonlinear interdependence in different EEG bands to measure the connectivity between brain regions. Then the nonlinear interdependence index was used to construct the brain functional networks of subjects in both groups. The global efficiency and average local efficiency were used to analyze the efficiency of information transmission among brain regions. The result showed that the nonlinear interdependence index of EEG theta activity (F=6.805,P=0.014,η²=0.185) in MCI patients were significantly lower than that in the control group, as well as of alpha band under negative viewing task (t=2.437,P=0.021, Cohen's d=0.865). Under some specific thresholds (threshold was 0.070 in low gamma band, threshold was 0.075 in theta band, threshold were 0.115 and 0.125 in alpha band), the network efficiency of MCI patients was significantly lower than that in the control group. In addition, the reappraisal effect (F=5.549,P=0.008, η²=0.246) was found in the control group (threshold was 0.115), showing that the global efficiency of EEG alpha activity(P=0.018) under the negative viewing condition(0.713±0.042) in controls was higher than that under the condition of negative reappraisal(0.699±0.045). However, similar reappraisal effect was not found in MCI patients. We therefore concluded that MCI patients had the deficit in brain network during emotion regulation, which was manifested in low efficiency of brain information transmission. In addition, no significant Pearson's correlation was found between nonlinear interdependence, network efficiency and behavioral scores, cognitive scores. More research of nonlinear correlation is necessary in order to find more stable brain network features with MCI in emotion regulation.
		2021 Vol. 40 (6): 662-673 [Abstract] ( 294 ) HTML (1 KB) PDF (4124 KB) ( 551 )

	674	Effect of Intranasal Mechanical Vibrational Stimulation on Resting-State EEG RelativePower and Effective Connectivity in Healthy Subjects
		Yu Xiaoru, Xu Wenlong, Xu Bingqiao, Ge Qiaoling
		DOI: 10.3969/j.issn.0258-8021.2021.06.04
		Intranasal mechanical vibrational stimulation (iMVS) is a new noninvasive nerve stimulation technique,and it gives rise to the adjustment of the intrinsic functional activity in the limbic system and restores the homeostasis in the autonomic nervous system. In this work, the neurophysiological mechanism of iMVS was explored by analyzing the effect of iMVS on the EEG relative power and EEG effective connectivity of healthy adults. The 22 healthy adults were recruited into the study and randomly divided into experimental group and control group. Each nasal cavity of 11 subjects in experimental group were performed with iMVS for 10 minutes, and the remained 11 subjects in the control group were performed with the sham stimulation. The resting-state EEG was recorded before and 30 minutes after iMVS. Welch transform was used to analyze the relative power; direct directed transfer function (dDTF) was used to obtain the effective connectivity; independent sample t-test, paired t-test and false discovery rate method were used forstatistical analysis. Results showed a significant improvement of alpha band relative power in experimental group (baseline: 51.57%±5.93%, 30 min: 57.33%±4.59%) and the relative power of C3, C4, T8, O1, O2 leads increased significantly in alpha band (P<0.05); beta band relative power was also enhanced significantly in experimental group (baseline: 7.28%±0.11%, 30 min: 8.36%±0.44%) and the relative power of C3, C4, T7, T8, O2 leads increased significantly in beta band (P<0.05). The dDTF value of alpha band was significantly enhanced in the experimental group (baseline: 0.052±0.0017, 30 min: 0.0592±0.0028), and the dDTF value of the directions, including F4 to F3, O2 to F3, C4 to F4, O2 to F4, F3 to C3, C4 to C3, F3 to T7, C4 to T8, O2 to O1, increased significantly in alpha band (P<0.05). There were no significant changes in the EEG relative power and EEG effective connectivity in the control group before and after stimulation. The intrinsic functional activity of limbic system was positively correlated with the EEG relative power of alpha and beta band as well as the EEG effective connectivity of alpha band. The results showed that iMVS enhanced the intrinsic functional activity of limbic system 30 minutes after the end of stimulation. The study explored the neurophysiological mechanism of iMVS, which improved the homeostasis of autonomic nervous system from the perspective of EEG analysis for the first time, and providedevidence for the use of EEG relative power and EEG effective connectivity as biomarkers of iMVS utility evaluation.
		2021 Vol. 40 (6): 674-680 [Abstract] ( 263 ) HTML (1 KB) PDF (2408 KB) ( 263 )

	681	Prediction Tumor Mutation Burden of Lung Adenocarcinoma Based on Deep Learning
		Sun Dewei, Wang Zhigang, Yang Xiaolin, Meng Xiangfu
		DOI: 10.3969/j.issn.0258-8021.2021.06.05
		A number of existing medical studies have found out that tumor mutation burden (TMB) is positively correlated with the efficacy of non-small cell lung cancer (NSCLC) immunotherapy, and in the recent related studies, tumor mutation load also has a certain predictive effect on the efficacy of targeted therapy and chemotherapy. Based on above situations, this paper proposed an inception deep learning model CAIM (combine attention and inception-block module) to identify the pathological sections of lung adenocarcinoma in non-small cell lung cancer from the Cancer Genome Atlas (TCGA) dataset. First, by segmenting the data samples, cutting them into small slices, and then sending them to the deep learning model, learning image features through convolution, and then combining with the attention mechanism to further strengthen the feature extraction. Finally, through the integration of the prediction information of the small slices, the TMB value of the pathological tiles of lung adenocarcinoma (LUAD) could be automatically determined. The data set consisted of 337 LUAD pathological tissue sections, including 271 data with high TMB value and 66 data experiments with low TMB value. Experimental results showed that the averaged area under the curve (AUC) of the proposed method was 0.82, which significantly better than the AUC value of 0.66 for the residual network of image classification method and was of great significance for the detection of tumor mutation burden and auxiliary diagnosis in clinical practice.
		2021 Vol. 40 (6): 681-690 [Abstract] ( 417 ) HTML (1 KB) PDF (3690 KB) ( 478 )

	691	Size-Adaptive Deep Neural Networks Based Pulmonary Nodule Detection in CT Scans
		Ai Qi, Wang Jun, Ren Fuquan, Weng Wencai, Yu Qiulei
		DOI: 10.3969/j.issn.0258-8021.2021.06.06
		Computed tomography (CT) screening for pulmonary nodules is an important method for early diagnosis of lung cancer. However, the automatic detection of pulmonary nodules, especially small nodules, is still challenging due to the large differences in shape, size and location of pulmonary nodules. To achieve highly sensitive detection of pulmonary nodules, in this paper, a new computer-aided detection system for pulmonary nodules detection was proposed. The system adopted two new strategies: size adaptive candidate test (SACD) and size adaptive false positive reduction (SAFPR). First, SACD combined deep and shallow convolution features to construct advanced features and detected CT images to obtain the location and size information of the region of interest. Then, the detection results were sent to three parallel sub-networks for screening different sizes of nodules, so as to refine the detection results of SACD and improved the accuracy and robustness of the computer-aided detection system. The results on the LIDC-IDRI dataset (1186 nodules) demonstrated that the proposed system achieved the high sensitivity of 96% at 1 FPs/scan, which was superior or comparable to the state-of-the-art systems, while in an independent dataset containing 430 nodules, the detection sensitivity of the system was 69.53% at 0.3 FPs/ scan for the nodules with the size of 3.89±2.34 mm, which was comparable to the human screening results of two experienced radiologists, indicating that the system has certain clinical application value.
		2021 Vol. 40 (6): 691-700 [Abstract] ( 287 ) HTML (1 KB) PDF (10477 KB) ( 123 )

	701	Segmentation of Thoracic Image Organs at Risk Based on Multi-Scale Feature-Aware
		Deng Shijun, Tang Hongzhong, Zeng Li, Zeng Shuying, Zhang Dongbo
		DOI: 10.3969/j.issn.0258-8021.2021.06.07
		Automatic segmentation of organs at risk (OARs) in medical images is an essentialconstituent of computer-aided diagnosis, and it plays a vital role in assisting doctors to completeradiotherapy with high quality and efficiency. There are some challenges in the accuratesegmentation of OARs for thoracic CT images, including low intensity contrast, different organswith interlaced and overlap regions, and different structure without clear boundaries. In this paper,a multi-scale feature-aware encoding-decoding network (FA-Unet) was proposed to segmentOARs in thoracic CT images.To address the problem of the size difference among fourkinds of organs in the thoraciccavity, an input-aware module was designed to extract multi-scale features in four types of organs. In order to bridge the semantic gap between theencoding and decoding layers, the modified inception module was introduced to long-range skipconnections between the encoding part and the decoding part in our architecture. Furthermore,we replaced the traditional serial convolution operation with the efficient spatial pyramid(ESP) andpyramid spatial pooling (PSP) modules to make our network more lightweightand avoide over-fitting caused by insufficient data effectively. We formulated a novel lossfunction by combining Dice coefficient and cross entropy to train our network to resolve theclass imbalance in thoracic CT images. Finally, we evaluated the effectiveness of our model onthe SegTHOR data set released by ISBI in 2019, and the dataset includes 7 390 thoracic CTimages of 40 patients with lung cancer or Hodgkin's lymphoma. Experimental results showedthat Dice coefficient of each organ in thoracic CT image was 0.793 2 of esophagus, 0.935 9of heart, 0.854 9 of trachea and 0.889 0 of aorta. Hausdorff distances was 1.420 7 of esophagus,0.212 4 of heart, 0.627 3 of trachea and 0.887 0 of aorta. Experimental results verified that ourproposed model outperformed other state-of-the-arts on the segmentation results of OARs andachieved very competitive performance on small target organs.
		2021 Vol. 40 (6): 701-711 [Abstract] ( 277 ) HTML (1 KB) PDF (4978 KB) ( 310 )

	712	The Formation of Coating Film on the Surface of Urinary Catheter for Sustained-Release of Local Anesthetic and its Analgesic Effect
		Lai Xinning, Xiao Bo, Huang Yuguang, Xu Haiyan, Xu Li, Liu Jian
		DOI: 10.3969/j.issn.0258-8021.2021.06.08
		Catheterization, especially the indwelling of urinary catheter may cause severe discomfort and pain in the patients, while the clinical approaches can only relieve the uncomfortable symptoms in a very short time. Therefore, this research has developed a technology that can simply and conveniently form a long-term release coating of local anesthetics on the surface of the catheter to achieve long-term anesthesia effects of indwelling catheters. In this study, polyvinyl alcohol, poloxamer 407, and lidocaine were blended as a film-forming liquid to form a film on the surface of the catheter (LID-P407-PVA), and the final sustained-release drug film was flat, with appropriate thickness and good wettability. Thein vitro release results showed that LID-P407-PVA can continuously release lidocaine in water for up to 48 h. The mouse model of pain caused by heat stimulation (n ≥ 5) was used to evaluate the analgesic effect of LID-P407-PVA in vivo. The results showed that the implantation of LID-P407-PVA next to the mouse sciatic nerve can produce significant sensory block for more than 20 h. Besides, it had a significant prolongation compared with the free drug group, showing a good sustained analgesic effect in vivo. The histopathological evaluation of the implantation site after the operation showed that the histocompatibility of LID-P407-PVA was good.
		2021 Vol. 40 (6): 712-718 [Abstract] ( 216 ) HTML (1 KB) PDF (4995 KB) ( 309 )

Reviews

	719	Motor Training and its Rehabilitation Application Based on the Neurofeedback Methods of Brain-Computer Interaction
		He Feng, He Beibei, Wang Zhongpeng, ChenLong, Ming Dong
		DOI: 10.3969/j.issn.0258-8021.2021.06.09
		Movement (or motor) is one of the most important forms of human survival, labour and communication with the outside world. However, some accidents or diseases can lead to the loss of partial or total motor functions of the human beings. Specifically, stroke has become the first major factor leading to the disability disease in China and the whole world. However, the conventional rehabilitation therapies are difficult to induce the synchronous coupling of corticomuscular function, especially the theoretical guidance of the mirror neuron system and neuroplasticity is lack, resulting in the limitation of the final rehabilitation effectiveness. Recently, the brain-computer interface (BCI) and/or other emerging human-machine technologies based on the neurofeedback training (NFT) method has emerged, which makes it quantitatively observable for the information of central nervous system and real-time perceptible for limb movement, thereby promoting functional reconstruction of the whole neural pathway and motor system. By summarizing the basic NFT principle of BCI training, combined with fusion and clinical application of current feedback training methods based on visual, auditory, tactile, and multi-sensory, it is expected that the future BCI training works in multi-sensory coordination and training mode becomes closed-loop controllable and adjustable based on the combination of initiative and external auxiliary function.
		2021 Vol. 40 (6): 719-730 [Abstract] ( 505 ) HTML (1 KB) PDF (6872 KB) ( 168 )

	731	Influence of Ultrasonic Detector Characteristics on Image Quality in Biological Photoacoustic Tomography and its Solution
		Sun Zheng, Sun Huifeng
		DOI: 10.3969/j.issn.0258-8021.2021.06.10
		Biomedical photoacoustic tomography (PAT) is an emerging hybrid functional imaging modality by multi-physics coupling for early detection and accurate diagnosis of tumors and cardiac vessel diseases. For simplicity, most PAT image reconstruction methods are based on an ideal assumption that the photoacoustically generated ultrasonic waves are collected by an ideal point-like detector with an omnidirectional response forming a continuous and complete measuring surface around the object. The influence of the spatial impulse response (SIR) and electrical impulse response (EIR) of the detector on the reconstruction quality is not considered. However, in practical applications, this assumption is usually infeasible, resulting in the reduction in the imaging resolution and the degradation of the image quality. This paper aimed to analyze the influence of the characteristics of the ultrasonic detector on PAT image reconstruction including limited aperture effect, SIR and EIR, directivity, scanning radius, limited view-angle and frequency bandwidth, and positional uncertainty. Moreover, the solutions to above problems were summarized and their advantages, limitations, applications, and potential developments in the future were discussed as well.
		2021 Vol. 40 (6): 731-742 [Abstract] ( 290 ) HTML (1 KB) PDF (7883 KB) ( 159 )

	743	Influence Factors of Corticomuscular Coherence and Rehabilitation Application
		Shi Xianle, Xu Rui, Wang Yaoyao, Meng Lin, Liu Yuan, Ming Dong
		DOI: 10.3969/j.issn.0258-8021.2021.06.11
		Corticomuscular coherence (CMC) is a tool for understanding how cerebral cortex controls muscle movement and evaluating the functional coupling between the cerebral motor cortex and related muscles. It is of great significance in related research on rehabilitation evaluation. Starting from the general definition of CMC, this paper summarized the frequency distribution characteristics of CMC within α, β and γ bands, introduced the influence of force level, age and pathological state on CMC from the view of theoretical research, and emphasized the application of CMC in the rehabilitation of stroke and other motor disorders. In the last part, we discussed some limitations in the research of CMC, and proposed potential improvements and explorations for the follow-up development, aiming to provide a new perspective and means for related rehabilitation research.
		2021 Vol. 40 (6): 743-751 [Abstract] ( 541 ) HTML (1 KB) PDF (1946 KB) ( 481 )

	752	Advances in Automatic Classification and Prediction Study of Neuropsychiatric Diseases
		Chen Xiaoyi, Zhou Jing, Ke Pengfei, Kong Lingyin, Wu Fengchun, Wu Kai
		DOI: 10.3969/j.issn.0258-8021.2021.06.12
		There are still many unknown neuropathological mechanisms of neuropsychiatric diseases, and objective clinical diagnostic criteria are lacking, which brings great challenges to the diagnosis and prognosis of neuropsychiatric diseases. With the rapid development of neuroimaging technology, neuroimaging data have been widely used to explore the neuropathological mechanism and potential biomarkers of neuropsychiatric diseases. Compared with traditional univariate analysis methods, that can only perform population-level analyses, neuroimaging-data-driven machine learning models can realize individualized and automated prediction of neuropsychiatric diseases. In this paper, we reviewed recent research progress of automated classification and prediction of neuropsychiatric diseases based on machine learning technology, and summarized and analyzed the basic principles of machine learning technology and the latest research achievements of four typical neuropsychiatric diseases, including schizophrenia, depression, Alzheimer‘s disease and Parkinson's disease. It was shown that current studies still face the challenge of small sample size and low reproducibility. Nonetheless, the sample size can be increased through collaborative analysis of multi-site data in the future. Meanwhile, deep learning and cross-disease diagnosis and prediction are also important directions of future research.
		2021 Vol. 40 (6): 752-763 [Abstract] ( 335 ) HTML (1 KB) PDF (1855 KB) ( 863 )

Communications

	764	Mirror Motion Realization Method of Upper Limb Rehabilitation Training Robot
		Li Junqiang, Xu Shu, Yang Dong, Li Tiejun
		DOI: 10.3969/j.issn.0258-8021.2021.06.13

		2021 Vol. 40 (6): 764-768 [Abstract] ( 326 ) HTML (1 KB) PDF (3142 KB) ( 411 )

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