Chinese Journal of Biomedical Engineering

Home

Editorial Board

Download

Quick Search

Office Online

Current Issue

Download Articles

Download

Links

2020 Vol. 39, No. 2 Published: 2020-04-20

Reviews Communications Regular Pagers

Regular Pagers

	129	Pulmonary Nodule Detection Algorithm Based on Faster-RCNN
		Song Shangling, Yang Yang, Li Xia, Feng Hao
		DOI: 10.3969/j.issn.0258-8021.2020.02.01
		Aiming to overcome the problems of individual differences and the homograph in the detection of pulmonary nodules, this paper presented a method of automatic recognition of pulmonary nodules based on Faster-RCNN. By comparing the adaptability of the current in-depth learning model, a general strategy was proposed to continuously improve the detection rate of pulmonary nodules with the increase of the number of samples. First of all, the hardware and software environment for deep learning was built; next, the data interface was set to match with the network interface of Faster-RCNN. Secondly, the single-category classification network of Faster-RCNN was set up and the parameters were adjusted. Thirdly, the pulmonary nodules database containing 2 000 patients was utilized to train different feature extraction models (including ZF and VGG models), and the features of CT pictures in different networks were calculated. The test results, missed detection rate and detection accuracy were evaluated. Finally, the influence of different training numbers and data augmentation types on the final detection accuracy was analyzed. The accuracy rate of ZF model was 90.82%, the variance of accuracy rate was 13.30%; the detection accuracy of VGG model was 87.02%, the variance of accuracy rate was 37.10%. Taking into account the balance between the missed detection rate and detection accuracy rate, the ZF model showed small fluctuation variance, a slight low accuracy, and high detection precision. Therefore, the ZF model for pulmonary nodules was better than VGG model. Our proposed lung nodule detection technology has a good theoretical value and engineering application value.
		2020 Vol. 39 (2): 129-136 [Abstract] ( 547 ) HTML (1 KB) PDF (3315 KB) ( 500 )

	137	Research on Adaptive Antagonism Method of ERP-BCI Under Parallel Task Interference
		Huang Yihao, Chen Yuqian, Qi Hongzhi
		DOI: 10.3969/j.issn.0258-8021.2020.02.02
		ERP-BCI is a classic brain-computer interface paradigm, useing event-related potential (ERP) features to decode users thinking activities. Recent studies have found out that the decrease of ERP-BCI recognition rate induced by impacts of ERP features will happen if human brains control BCI simultaneously and perform other thinking activities. To explore solutions to this problem, this paper established an ERP-BCI interference antagonism method that applied dynamic stop criterion to adaptively adjust the stimulus repetition to maintain recognition performance. In the research, the working memory n-back task was used to construct the thinking interference task parallel to the ERP-BCI operation. The ERP data achieved from the task of no interference were used to establish the discriminant model and the dynamic stop algorithm which are employed in the operation of ERP-BCI online under different levels of the interference with dissimilar tasks. Ten subjects participated in the operation of ERP-BCI online. The experimental results showed that the proposed method of interference antagonism could obtain the character recognition rate without significant difference between interference and no interference (the average recognition rate reached 95%). This study provides a certain technical foundation for establishing highly robust ERP-BCI.
		2020 Vol. 39 (2): 137-144 [Abstract] ( 334 ) HTML (1 KB) PDF (2622 KB) ( 354 )

	145	Study on Brain Function Network of Emotional Conflict Response inCollege Students with Anxiety
		Ji Shumei, Su Xinle, Xun Xingmiao, Bu Xinxin, Xu Quansheng
		DOI: 10.3969/j.issn.0258-8021.2020.02.03
		The aim of this work is to analyze the topological structure and properties of brain network with complex network analysis method based on graph theory exploring the characteristics of brain functional network in emotional conflict response on anxious population of college students. Sixteen volunteers in the anxiety group and the normal control group were selected for the emotional word-face stroop conflict task, and 64 EEG were recorded simultaneously. The EEG data with beta(14 ～ 30 Hz) and high gamma(50 ～ 80 Hz) were analyzed by synchronous likelihood analysis, and the appropriate threshold value was selected to construct the brain network topology, and the node degree and clustering coefficient of the network were calculated. Results showed that in beta and high-gamma rhythms, there were abnormal connections in the brain areas including frontal lobe, temporal lobe, and parietal lobe in anxiety group, and the node degree of frontal lobe and parietal lobe was lower than that of normal group (P<0.05), while the node degree of temporal lobe was higher than that of the normal group (P<0.05)(in beta rhythm,the node degrees of the frontal FP1, parietal CP1 and temporal T7 electrodes were 5.21±0.62, 6.25±0.53, 7.91±0.71 respectively in anxiety group and 10.42±1.53, 7.94±0.55, 3.55±0.36 in the normal group, indicating that the function of frontal lobe and parietal lobe were decreased in the anxiety group, while the function of temporal lobe was increased. The clustering coefficient of the brain network in the anxiety group was lower than that in the normal group (P<0.05) (the clustering coefficients of beta and high gamma rhythm in anxiety group were 0.523 8±0.039 2 and 0.586 4±0.055 8, respectively, while those in normal group were0.603 2±0.071 1 and 0.664 7±0.060 1), which indicated that the degree of internal clustering of the brain network in the anxious group and the information transmission ability of the network were decreased. This study can provide a new perspective for the neural mechanism research of psychological and mental diseases such as anxiety and depression.
		2020 Vol. 39 (2): 145-151 [Abstract] ( 470 ) HTML (1 KB) PDF (6396 KB) ( 343 )

	152	A Method of Measuring Blood Pressure by Combining with Traditional Oscillography and Novel k-Means Algorithm
		Hua Bin, Chen Yuting, Lv Qingpu
		DOI: 10.3969/j.issn.0258-8021.2020.02.04
		The improvement of the accuracy of measuring blood pressure is an important issue to prevent hypertension. Oscillography uses statistical experience to determine a set of coefficients, which is inaccurate enough in the measurement of blood pressure. In this work, a method of measuring blood pressure by combining multiple techniques was proposed. Five characteristic parameters from the de-interference oscillometric waveform envelope were extracted to match the envelope to the best set of coefficients through the classification model and measure the blood pressure by oscillography. The classification model was trained offline. Specifically, the first thing was to uniquely cluster the envelopes using the novel k-means. Next, adaptively determined a set of optimal coefficients for each type of envelope according to the gradient descent, and there was a one-to-one correspondence between the centroid vectors and the calculated coefficients. Based on above mentioned, the classification model was finally constructed. There were 33 subjects in the experiment. Experimental results showed that the proposed blood pressure measurement was similar to that of mercury method, the correlation coefficient could reach 0.91, and met the AAMI-SP10 standards even the first-level requirements of ESH. Furthermore, the proposed measurement was superior to oscillography, with regard to the deviations between the estimated blood pressure and the target blood pressure, the number of deviations within 5 mmHg was increased by 10.5% on average, and the number of deviations within 10 mmHg was increased by 6.8% on average.
		2020 Vol. 39 (2): 152-159 [Abstract] ( 405 ) HTML (1 KB) PDF (2760 KB) ( 358 )

	160	EEG Signal Analysis of Fatigue Caused by Virtual Reality Immersive Visual Experience
		Wang Lei, Zhang Tianheng, Guo Miaomiao, Xu Guizhi
		DOI: 10.3969/j.issn.0258-8021.2020.02.05
		Virtual reality (VR) is a computer technology that generates realistic images, sounds and other sensations that simulate a user′s physical presence in a virtual environment. With the rapid development of virtual reality technique, brain fatigue caused by VR has raised concerns. In this work, 16 healthy subjects were recruited, and EEG signals were acquired synchronously while the subjects were watching videos in similar types presented by traditional displayer and virtual reality separately. Two questionnaires were conducted by all subjects to evaluate the state of fatigue before and after the experiment. We also compared the relative energy of alpha, beta, delta and theta band during two experiments. What′s more, the fatigue factor and center of gravity were analyzed. Results showed that in the VDT and SPFS questionnaires, the average score was 0.44±0.22 and 3.28±1.03 after watching virtual reality video, which was higher than 0.31±0.20 and 2.26±0.98 after watching traditional plane video, with a significant difference (P<0.05). And from the EEG signals’ results, significantly at the temporal lobe area, after watching the VR video, the alpha band energy decreased (before: 0.249±0.007, after: 0.234±0.005, P<0.05) and the value of gravity frequency decreased (before: 7.545±0.950 Hz, after: 3.717±0.398 Hz, P<0.05) while the delta band energy increased (before: 0.295±0.012, after: 0.314±0.007, P<0.05), and moreover, the variation trend of these parameters was significantly different from that after watching traditional plane video, showing an opposite variation trend. Furthermore, the alpha band energy at the temporal lobe and parietal lobe area decreased during watching VR video, while the beta band energy decreased following the initial increase. These results indicated that watching VR videos was more likely to induce subjective fatigue and the changes of rhythm activity in characteristic frequency band of EEG signals, providing theoretical basis for objective evaluation of brain fatigue caused by immersive experience of VR.
		2020 Vol. 39 (2): 160-169 [Abstract] ( 577 ) HTML (1 KB) PDF (4712 KB) ( 565 )

	170	Analysis Method of Current Density Based on Magnetocardiosignal
		Ai Haiming, Primin Mykhailo, Mi Wang, Wu Hongjin
		DOI: 10.3969/j.issn.0258-8021.2020.02.06
		The analysis of current density vector as diagnostic tool is a critical technique in clinical application of magnetocardiography, its reconstruction and classification performance will directly determine the accuracy of cardiac disease screening. A current density analysis method was studied in this paper, which includes inverse problem solutionfor current density reconstruction and new proposed algorithm of automatic classification. The current density analysis method was used to clinically validate 27 healthy volunteers and 75 patients with myocardial ischemia. First, the SQUID sensor of four-channel magnetocardiography was used to acquire magnetocardiosignal in 36 measurement sites, ECG signal base on standard lead ⅡECG was applied to synchronize magnetocardiosignal. Second, inverse problem solution algorithm based on Fourier transformation was applied to reconstruct current density distribution for each subject, automatic classification parameters of CDI, IPCD, E_RT were automatically calculated, the linear discriminant function of healthy group and myocardial ischemia group in automatic classification algorithm was determined by multiple linear regression model. Finally, automatic classification of reconstructed current density was performed by the new automatic classification algorithm based on KL entropy calculation and liner discriminant function. Results showed that significant difference of three classification parameters of CDI, IPCD, E_RT between healthy group and myocardial ischemia group was 2.0×10^-11, 2.7×10^-5and 3.0×10^-4, respectively; sensitivity and specificity in myocardial ischemia diagnosis for CDI, IPCD, E_RT was 73%, 81%, 60% and 100%, 65%, 73%, respectively; sensitivity, true positive rate, specificity, false positive rate and true negative rate of automatic classification results was 89.33%, 100%, 100%, 0% and 76.47%, respectively. Therefore, the proposed current density analysis method is reliable, accurate and available for clinical diagnosis of myocardial ischemia.
		2020 Vol. 39 (2): 170-179 [Abstract] ( 421 ) HTML (1 KB) PDF (7035 KB) ( 271 )

	180	The Establishment of a Heart Model: From CT Data to Finite Element Model with Fiber Orientation
		Qian Li, Wang Jianfei, Jin Lian, Song Biao, Huang Yanqi, Zhu Honglei, Yan Shengjie, Wu Xiaomei
		DOI: 10.3969/j.issn.0258-8021.2020.02.07
		Cardiac computer model is a powerful tool for studying cardiac physiological/pathological function and treating patients with arrhythmias. The finite element model of the heart is the basis for building various heart models. Establishing personalized cardiac computer models from patients' clinical image data provides great convenience for clinical diagnosis and treatment. This paper developed a method for establishing a full-heart finite element model from human chest CT imaging data. It included following two steps: 1, establishment of a cardiac anatomical surface model by MIMICS based on chest CT images; 2, repairing the surface model with HyperMesh to obtain a finite element model of the heart entity. As the direction of myocardial fiber is closely related to the electrical/mechanical activity of the heart, after the reconstruction of cardiac anatomy from image data, we specifically investigated the addition of the myocardial fiber orientation. First, using the rule-based approach to add the direction of ventricular muscle fibers. Afterwards, based on rule-based approach, using structural tensor analysis to do the smoothing filtering. By this way, the atrial muscle fiber orientation was obtained. In order to verify the correctness of adding myocardial fibers, the fiber orientation of the special part in the heart was investigated. The angle between the direction of the myocardial fibers and the OX axis of the Bachmann′s bundle, the posterior left atrium, the superior posterior left atrium, and the interatrial groove was 4.97°±4.84° (Mean±Standard deviation), 111.99°±3.72°, 178.89°±3.73°, and 86.48°±4.01° respectively, consistent with myocardial fiber observations reported in the literature. The method proposed could construct a finite element model of the heart from the cardiac image data, which included myocardial fiber.
		2020 Vol. 39 (2): 180-189 [Abstract] ( 354 ) HTML (1 KB) PDF (7405 KB) ( 419 )

	190	Biomechanical Study of Heel Pain During Push-off Period Based on Finite Element Method
		Zhang Haowei, Chen Liang, Yang Junyan, Liu Ying, Zheng Yongjun
		DOI: 10.3969/j.issn.0258-8021.2020.02.08
		The pathogenesis and mechanism of rehabilitation of heel pain was studied in this paper. Images from CT scan and MRI of patients with heel pain were collected for three-dimensional reconstruction. Geomagic was used to optimize the surface of the obtained model, and the model was preprocessed with finite element through Hypermesh. The obtained lower limbs finite element model was imported intoAbaqus for analysis and calculation. The validity of the model was verified by comparing with the test results of plantar pressure plate. According to the calculation, the influence of the changes of triceps surae force on the biomechanical behavior characteristics of foot and ankle gait during the push-off was analyzed. The result showed that the muscle strength of the triceps surae increases from 550 N to 1100 N, the peak pressure in the first phalangeal area increased by 32.8%, and the peak pressure in the metatarsal area increased by 14.3%; the stress of the first plantar fascia was up to 4.69 MPa; the stress peaks at the junction of tendon and calcaneal and the calcaneal tuberosity were 7.41 MPa and 6.79 MPa respectively. These indicated that contracture of triceps surae and the windlass effect in the push-off period will lead to excessive stretch of plantar fascia, which would lead to stress level improvement at calcaneal tuberosity and change in the biomechanical environment of the foot, thus inducing plantar fasciitis and heel pain. Relieving the contracture of the triceps surae, reducing the muscle force and avoiding the overstretch of the plantar fascia and reducing the stress level at the attachment point, thus restoring the normal biomechanical environment of the foot are the main rehabilitation mechanisms for the treatment of heel pain.
		2020 Vol. 39 (2): 190-196 [Abstract] ( 424 ) HTML (1 KB) PDF (4386 KB) ( 344 )

	197	Quantitative Analysis of Biological 3D Printed Artificial Skin Using Optical Coherence Tomography
		Hu Jie, Wang Ling, Xu Mingen, Wang Zhongkun
		DOI: 10.3969/j.issn.0258-8021.2020.02.09
		The non-destructive quantitative detection and analysis of the artificial skin three-dimensional features is a key problem to be solved in the research of skin printing and induced culture technology. This paper used spectral domain optical coherence tomography (SD-OCT) to perform non-destructive imaging and quantitative analysis on biological 3D printed artificial skin. The adaptive peak detection algorithm based on OCT intensity signal quantified skin three-dimensional thickness distribution and roughness variation. The overall thickness of the skin at the location and its fluctuation quantitatively visualized the spatially resolved structural features of the skin. The structural characteristics of OCT imaging artificial skin were consistent with the results of sliced hematoxylin and eosin (H&E) staining. The difference between the two measured skin thicknesses was only 3.59 μm, which verified the feasibility and accuracy of the method. Through the continuous detection of artificial skin in the culture period by SD-OCT, the two-dimensional artificial skin thickness distribution detection can show the thickness growth curve of the skin at different culture time, and the three-dimensional spatial resolution thickness distribution map and surface roughness map can be more visually show skin growth status. Quantitative statistical results show that during the gas-liquid culture, the overall average thickness of the artificial skin is constantly increasing and stabilizes. When the skin matures, the overall average thickness is 83.91 μm. The surface roughness of the skin first decreases and then increases with the change of keratinization. Therefore, the method based on OCT intensity signal quantitative analysis can truly and effectively reflect the structural parameter changes of biological 3D printed artificial skin, which provides a reliable monitoring method for quality assessment in artificial skin preparation.
		2020 Vol. 39 (2): 197-205 [Abstract] ( 316 ) HTML (1 KB) PDF (4921 KB) ( 306 )

	206	Coaxial 3D Bioprinting of Vascular Tissue Engineering Scaffolds forPromoting Endothelial Cell Growth
		Zhang Yifan, Xu Mingen, Wang Ling, Zhang He
		DOI: 10.3969/j.issn.0258-8021.2020.02.10
		In the field of tissue engineering, the construction of vascular networks in vitro is very important for the regeneration of thick tissue and organs. In this study, with alginate/silk fibroin as bioink, vascular tissue engineering scaffolds containing human umbilical vein endothelial cells (HUVECs) were rapidly prepared by coaxial 3D bioprinting. Firstly, the optimal concentration of materials for coaxial system were determined through compression modulus and printability test. Then, the effects of coaxial printing parameters on the shape of hollow fiber were studied by using optical coherence tomography, and the parameters were optimized. Combined with the simulated perfusion experiment, the vascular structure was characterized. Finally, the growth of HUVECs in the scaffolds was verified by cell live/dead staining and Alamar Blue method. Results showed that the optimized bioink and printing parameters could successfully prepare vascular tissue engineering scaffolds with complete internal connectivity. The HUVECs showed agglomeration growth in vitro, and the presence of vascularization channels was beneficial to maintain overall tissue activity, with a one-week survival rate of more than 97% and a higher proliferation rate than the control group. Our study proved that the vascular tissue engineering scaffolds built by coaxial 3D bioprinting technology promoted endothelial cell growth, providing new possibilities for thick tissue and organ regeneration.
		2020 Vol. 39 (2): 206-214 [Abstract] ( 474 ) HTML (1 KB) PDF (13562 KB) ( 232 )

Reviews

	215	Research Advancements of Deep Learning on EEG Decoding
		Liu Zheng, He Feng, Tang Jiabei, Wan Baikun, Ming Dong
		DOI: 10.3969/j.issn.0258-8021.2020.02.11
		Electroencephalography (EEG) has millisecond time resolution, which can acquire real-time neurophysiology information of brain activity. EEG has been widely used in cerebral imaging and has become an important tool in neuroscience and neuroengineering in 21st century. However, the original signal-to-noise ratio (SNR) and spatial resolution are poor and the decoding is seriously hindered due to the volume condution effect. With the great development of Deep Learning (DL) in this century, researchers have been trying to combine the two to explore the application of deep learning in EEG data processing, and leading to some phase achievements. Nevertheless, there are still challenges in applying the current DNNs to EEG data processing. Combining with recent studies on DNN-based EEG data processing, this paper introduced the implementation of deep EEG decoding and discussed existing problems and future directions.
		2020 Vol. 39 (2): 215-228 [Abstract] ( 576 ) HTML (1 KB) PDF (6408 KB) ( 1339 )

	229	Progress on Single-Molecule Localization Algorithms for Super-Resolution Imaging
		Lin Wanni, Jin Luhong, Xu Yingke
		DOI: 10.3969/j.issn.0258-8021.2020.02.12
		The development of super-resolution microscopy has provided an unprecedented opportunity for biomedical research. The single-molecule localization based super-resolution imaging has been widely used in biomedical field due to its relatively simple hardware structure. Single-molecule localization microscopy utilizes the characteristics of fluorescence molecules to randomly activate and image a small number of discrete distributed fluorophores, and then achieves high-precision spatial localization of single molecules by fitting analysis. In this process, the algorithms for single-molecule localization and the speed of image processing are particularly critical. This review briefly elaborated the single-molecule localization algorithms according to their basic principles, including model selection, estimator selection and reconstruction result evaluation methods. Based on that, we classified and compared the capabilities of more than ten developed algorithms, and also selected a few representative algorithms to discuss and introduce their functions. We hope this review would provide reference for future related research.
		2020 Vol. 39 (2): 229-237 [Abstract] ( 573 ) HTML (1 KB) PDF (5651 KB) ( 729 )

	238	New Progress of Left Myocardial Segmentation Based on Cardiac Cine MRI
		Wang Hui, Wang Lijia
		DOI: 10.3969/j.issn.0258-8021.2020.02.13
		The segmentation of left myocardium can be used not only to calculate the volume of left ventricle, but also to evaluate the quality of myocardium, track the movement of myocardium, reconstruct the heart, therefore is appliable for clinical evaluation of cardiac function, which is of great significance for the diagnosis and treatment of myocardial infarction, hypertrophy and other cardiac diseases. Cardiac cine magnetic resonance image is widely used for left myocardial segmentation and functional evaluation, which has high temporal and spatial resolution. However, left myocardium is adjacent to tissues of similar gray scale outside, and may be connected with trabecular and papillary muscles inside, which brings great difficulty to left myocardium segmentation. Great efforts have been made in the left myocardium segmentation algorithm. This paper reviewed the progress on the left myocardial segmentation based on CCMRI. We focused on the new methods, including optical flow method and deep learning following the introduction of traditional methods and its advantages and disadvantages that were summarized and compared. At last, evaluation criteria in common were briefly introduced. In conclusion, segmentation methods based on deep learning has higher accuracy and faster speed than the traditional ones, which is worth of further investigations in the future while problems such as massive data requirements and hyper parameter determination still need to be solved in deep learning.
		2020 Vol. 39 (2): 238-246 [Abstract] ( 404 ) HTML (1 KB) PDF (3304 KB) ( 484 )

Communications

	247	Inter-digit Force Coordination and Control During Multi-Finger Precision Grip Based on Perception of Object′s Moment
		Liu Mengjie, Wei Na, Li Ke
		DOI: 10.3969/j.issn.0258-8021.2020.02.14

		2020 Vol. 39 (2): 247-251 [Abstract] ( 370 ) HTML (1 KB) PDF (1011 KB) ( 402 )

	252	Experimental Study on the Mechanism of Cell and Protein Cleavage During Low Temperature Plasma Ablation
		Hu Xiaojing, Yu Liyang, Gu Yunru Ying, Nanjiao, Zhao Weijie, Hao Zhuzhu, Pan Song, Yang Yong
		DOI: 10.3969/j.issn.0258-8021.2020.02.15

		2020 Vol. 39 (2): 252-256 [Abstract] ( 387 ) HTML (1 KB) PDF (4706 KB) ( 385 )

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