Chinese Journal of Biomedical Engineering

Home

Editorial Board

Download

Quick Search

Office Online

Current Issue

Download Articles

Download

Links

2024 Vol. 43, No. 1 Published: 2024-02-20

Reviews Regular Papers

Regular Papers

	1	Sepsis Real-Time Risk Prediction Model for Intensive Care Unit Patients Based on Machine Learning
		Li Runfa, Yang Meicheng, Li Jianqing, Liu Chengyu
		DOI: 10.3969/j.issn.0258-8021.2024.01.001
		Sepsis is a syndrome of organ dysfunction caused by the body's dysfunctional response to infection, with high morbidity and mortality. The traditional scoring system has low specificity. Based on the LightGBM machine learning framework, this study proposed a model for early prediction and risk assessment of sepsis to provide timely intervention for patients with potential risk of sepsis. In order to realize the model, a time series feature construction method based on LASSO feature selection and sliding window path reintegration and a time series clustering sampling method based on dynamic time regularization algorithm were proposed. We selected clinical information from 29 239 patients in the MIMIC-III dataset and 40 336 patients in the PhysioNet/CinC 2019 challenge dataset to train and validate the model. The sensitivity, specificity, and area under the receiver operation characteristic curve (AUC) of the proposed model on the MIMIC-III and PhysioNet/CinC 2019 independent test sets were 0.737 7, 0.730 4, 0.814 7 and 0.802 6, 0.789 1, 0.873 0, respectively. Compared with the state-of-the-art method EASP, the improvement of AUC was 3.62% and 2.83% respectively. In conclusion, the established model could predict the risk of sepsis in real time, reveal the important factors affecting the occurrence of sepsis, and provide a basis for timely intervention of people at risk of sepsis.
		2024 Vol. 43 (1): 1-9 [Abstract] ( 470 ) HTML (1 KB) PDF (4471 KB) ( 361 )

	10	Effects of Isoflurane Anesthesia on Spontaneous and TUS/TMAS Induced Electromyography in Mice
		Wang Ruru, Zhou Xiaoqing, Zhao Yuhui, Liu Xu, Liu Zhipeng, Wang Xin, Yin Tao
		DOI: 10.3969/j.issn.0258-8021.2024.01.002
		Transcranial ultrasound stimulation (TUS) and transcranial magneto-acoustic stimulation (TMAS) are effective in modulating motor cortex. Previous studies have mostly been conducted under anesthesia confined to the difficulty of tethering conscious animals, and the analysis of the attenuating modulatory effects of anesthesia has primarily focused on the central nervous system. In this study, spontaneous electromyography (EMG) and TUS/TMAS-induced EMG were recorded during isoflurane anesthesia in 24 mice, the effects of anaesthetize on spontaneous and induced EMG were analyzed quantitatively in terms of firing rate, latency, duration and amplitude. Experimental results showed that the frequency of spontaneous EMG decreased by 50% per cycle approximately and the duration decreased, as the concentration of isoflurane increased from 0.4% to 0.75%, indicating an inhibitory state. The induced-EMG success rates of TUS/TMAS both decreased by 50% and 70% respectively, with an average increase of 0.1 s in latency and a decrease of 0.3 s and 0.5 s in duration respectively, suggesting that the modulatory effect of TUS/TMAS on the motor cortex attenuated as the depth of anesthesia increased. According to the observed correlation between the firing rate and duration of spontaneous and induced EMG, it is rational to infer that the suppression of spontaneous EMG in mice under anesthesia was one of the factors contributing to the attenuated modulatory effect.
		2024 Vol. 43 (1): 10-17 [Abstract] ( 222 ) HTML (1 KB) PDF (1315 KB) ( 155 )

	18	Heart Rate Variability Based Automatic Sleep Staging and its Validation with EEG
		Ying Shaofei, Qin Daiyou, Xie Jiaxin, Gao Dongrui, Qin Yun, Liu Tiejun
		DOI: 10.3969/j.issn.0258-8021. 2024. 01.03
		Sleep disorders seriously affect life quality, therefore, early sleep monitoring is important for the prevention and diagnosis of sleep diseases. In this paper, we proposed a portable polysomnography and completed in-home sleep data collection for 103 nights by this system, including EEG, EOG, EMG and ECG signals. Time-domain, frequency, and nonlinear features were extracted from the RR intervals of the synchronously acquired ECG data, and up to 426 heart rate variability (HRV) features were combined to construct models based on the Xgboost algorithm to predict wake, non-rapid eye movement I(N1), non-rapid eye movement II (N2), non-rapid eye movement III(N3), and rapid eye movement (REM) stages of sleep with five-classification (wake, N1, N2, N3, and REM), three-classification (wake+N1, REM, N2+N3), and two-classification (wake, N1+N2+N3+REM), and to validate them with the EEG sleep staging labels. Among these, the accuracy of the five-classification, three-classification and two-classification test results reached 84.0%, 89.1% and 95.2%, respectively, and the F1-score reached 83.2%, 88.9% and 94.9%, which was the best performance among other model studies of this kind. It indicated that HRV had good correlation with sleep stages, and the HRV-based algorithmic models constructed based on the data collected from portable devices identified the sleep states well.
		2024 Vol. 43 (1): 18-26 [Abstract] ( 250 ) HTML (1 KB) PDF (2516 KB) ( 187 )

	27	Analysis of Feature Fusion Strategies of Resting-State Brain Functional Connectivity Network in Patients with Negative Temporal Lobe Epilepsy
		Wang Kaiwei, Ge Manling, Wang Lina, Cheng Hao, Zhao Xiaohu, Chen Shenghua, Zhang Qirui
		DOI: 10.3969/j.issn.0258-8021.2024.01.004
		Resting-state fMRI (rfMRI) can provide abnormal functional indicators by the functional connectivity (FC) analysis, however, the features redundancy would affect the classification precise. To address this issue, a feature fusion strategy combining specificity index model with discriminant correlation analysis (DCA) was proposed in this study to improve the identifying accuracy for patients with MRI-negative temporal lobe epilepsy. Firstly, the rfMRI data of 20 patients and 20 healthy people were preprocessed. Taking the healthy group as a control, two specificity index models were constructed by the conventional FC of pearson correlation and the network FC of graph theory. Secondly, both minimum redundancy maximum relevance (mRMR) and independent sample t test were used to eliminate redundant features, and DCA method was used to fuse feature. Finally, three machine-learning classifiers such as k-nearest neighbor (KNN), support vector machine (SVM) and logistic regression (LR) were used to validate our feature fusion method, and the nested stratification cross validations, such as 10 times 10 fold and 10 times 5 fold were used to evaluate the performance of three classifiers. The fusion feature of DCA could achieve the recognition rate of 91.25%~92.5%, higher than non-fusion strategies. In conclusion, the feature fusion strategy proposed in this paper could effectively deal with the redundant information and enhance feature discrimination. This work may provide new thoughts for the identification for MRI-negative temporal lobe epilepsy.
		2024 Vol. 43 (1): 27-38 [Abstract] ( 165 ) HTML (1 KB) PDF (11405 KB) ( 102 )

	39	Augmented Region-Growing-Based Motion Tracking Using Bayesian Inference and Local Polynomial Fitting for Quasi-Static Ultrasound Elastography
		Wen Shuojie, Zhou Jingyu, Zhou Wenjun, Jiang Jinfeng, Peng Bo
		DOI: 10.3969/j.issn.0258-8021.2024.01.005
		Ultrasound elastography is a non-invasive imaging method for assessing tissue stiffness and has been used in clinics in the examination of breast, prostate, and abdominal organs. In ultrasound elastography, speckle tracking is a crucial step. The block matching method-based motion estimation and their variants (such as guided displacement tracking algorithm) are commonly used. However, it often introduces peak-hopping errors during the imaging process due to signal de-correlation caused by out-of-plane probe or unrelated physiological motion, resulting in poor quality of the estimated displacement and corresponding strain images generated by such methods. Based on the principle of tissue motion continuity, this study proposed a motion tracking algorithm (BRGMT-LPF) that incorporated Bayesian inference and local polynomial fitting (LPF) into a region-growing motion tracking (RGMT) framework. Firstly, the proposed approach replaced the traditional cross-correlation with the maximum posterior probability. Secondly, LPF was applied to remove and update the peak-hopping or wrong estimated displacement point. The proposed approach was compared with conventional RGMT algorithm, the RGMT with LPF, and the RGMT with Bayesian inference (BRGMT) on the computer-simulated and in vivo ultrasound data. Experimental results showed that on 10 pairs of ultrasound data simulated by finite element software and FIELD II, BRGMT-LPF achieved the lowest average absolute error (MAE) of 0.1699 (at least 0.25% reduction) and the highest contrast-to-noise ratio (CNR) of 1.1625 (at least 4% increase). On 16 pairs of vector data collected from patients with pathologically confirmed breast tumors, BRGMT-LPF obtained the highest CNR of 1.50 (at least 0.37% increase) and the highest motion compensation cross-correlation (MCCC) of 0.84 (at least 9.4% increase). In conclusion, the proposed method could be used to improve the image quality of ultrasonic elastography and displacement-based modulus reconstruction.
		2024 Vol. 43 (1): 39-48 [Abstract] ( 153 ) HTML (1 KB) PDF (6829 KB) ( 86 )

	49	Three-Dimensional Reconstruction Method of Lesion Slices Based on Refined Isosurface
		Tan Ling, Liang Ying, Ma Wenjie, Xia Jingming, Zhu Jining
		DOI: 10.3969/j.issn.0258-8021.2024.01.006
		Three-dimensional reconstruction of brain tissue lesion slices is of great significance for understanding status of glioblastoma, and can be used for various clinical applications including differential diagnosis and surgical simulation. The marching cubes (MC) algorithm is a classic polygonal surface reconstruction algorithm with the advantages of simplicity and ease of implementation, however, its efficiency is low with obvious cascade phenomena. To solve this problem, this study proposed a lesion slice spatial stacking reconstruction method (SSR-RI) refining the isosurface topology, aiming to achieve optimization of topology configuration and operational efficiency. SSR-RI was applied to process adjacent images of MRI slices by constructing a spatial coordinate system. In order to improve the vulnerability of image components in the bilinear interpolation method, an adaptive spatial interpolation method was proposed, which adaptively selected interpolation points according to the change of gray value to expand the surrounding. On the basis of combining isonormal vertices, a stacked reconstruction method for refining isosurface extraction was designed to improve stacking speed and effectively reduce cascade problems. In order to further optimize the iterative reconstruction effect of SSR-RI, an improved local reflection illumination (PR) was proposed to draw 3D lesions, and the reconstruction volume was optimized by using specular color reflection (SCR) and specular exponent (SE). There were 618 cases of brain tumor segmentation dataset BraTS used to carry out iterative reconstruction experiments to verify the performance of this research method. Experimental results showed that the reconstruction time of the proposed algorithm was only 2.124 seconds, with an F-score value of 0.845 and an SSIM value of 0.81. Compared to the MC algorithm, the reconstruction time was reduced by 38%, and the F-score value and SSIM value were increased by 30.89% and 38.4%, respectively. The structure of the reconstructed volume sequence was compact, and the visual effect was more stereoscopic and textured, which effectively improved the rendering efficiency of iterative reconstruction.
		2024 Vol. 43 (1): 49-59 [Abstract] ( 137 ) HTML (1 KB) PDF (9451 KB) ( 127 )

	60	Research on Trachea Segmentation Algorithm Based on Unet+Attention from Chest CT Images
		Zhang Ziming, Zhou Qinghua, Xue Hongsheng, Qin Wenjun
		DOI: 10.3969/j.issn.0258-8021.2024.01.007
		There are many challenges in current lung trachea segmentation including complex grayscale distribution of CT images, segmentation target pixel approximation, easy to cause over-segmentation, fewer lung trachea pixels, and difficult to get more target features, therefore, fine lung trachea is easy to be ignored. In this paper, we studied a lung trachea segmentation algorithm combining Unet network and attention mechanism, which a convolutional block attention model CBAM focusing on the channel domain and spatial domain was used in the attention mechanism, which improved the tracheal feature weights. In terms of loss function, for the problem of imbalance between positive and negative samples in the original data, this paper used the focal loss function to improve the standard cross-entropy loss function, so that the hard-to-classify samples got more attention in the training process. Finally, the isolated points were removed by eight connected domains judgment, and several larger connected domains were retained, i.e., the last pulmonary trachea part. Twenty-four sets of CT images and 43 sets of CTA images provided by the partner hospitals, totaling 26 157 slice images, were selected as the data set for segmentation experiments. The results showed that the segmentation accuracy reached 0.86, and the mean values of over-segmentation rate and under-segmentation rate were 0.28 and 0.39 respectively. After the ablation experiments of attention module and loss function, the accuracy, over-segmentation rate and under-segmentation rate before improvement were 0.81, 0.30 and 0.40, respectively, indicating the segmentation effect was inferior to the method proposed in this paper. Compared with other commonly used methods under the same conditions, the proposed method reached the highest accuracy rate under the condition that the over-segmentation rate, and under-segmentation rate were guaranteed to be unchanged. The above experiments proved the accuracy of the algorithm in this paper, and successfully solved the problem of inaccurate segmentation of fine trachea.
		2024 Vol. 43 (1): 60-69 [Abstract] ( 208 ) HTML (1 KB) PDF (2587 KB) ( 311 )

	70	Image Recognition of Small Intestinal Ulcer Based on MobileNetV2
		Liu Zhang, Guo Xudong, Li Shengnan
		DOI: 10.3969/j.issn.0258-8021.2024.01.008
		Ulcer lesions under enteroscopy are complex in shape and difficult to differentiate and diagnose. To realize the artificial intelligence-assisted recognition of small intestinal ulcer lesions and improve the diagnosis efficiency and accuracy, a small intestinal ulcer lesion recognition algorithm based on the MobileNetV2 network was constructed. The MobileNetV2 was used as the backbone feature extraction network, and the output feature map was extracted in space at multiple scales and then input to the channel attention module for feature recalibration, and the features on multiple scales were fused and output classification, in order to alleviate the impact of data imbalance, an improved loss function was proposed. The data set used were collected from a total of 2124 enteroscopy clinical images of 282 patients in Shanghai Changhai Hospital. The proposed method was used to test the data set, and the recognition accuracy was 87.86%, the average accuracy of 5-fold cross-validation was 87.24%. The gradient weighted class activation map was used for visual verification. At the same time, the proposed modules were applied to different backbone network architectures, which reached an improvement to a certain extent, and displayed good versatility. Experimental results showed that the network model extracted more information of lesion, strengthened the identification of lesion characteristics, and had a higher recognition accuracy for small intestinal ulcer images, and was able to initially realize the automatic identification of small intestinal ulcer types.
		2024 Vol. 43 (1): 70-79 [Abstract] ( 146 ) HTML (1 KB) PDF (5981 KB) ( 297 )

Reviews

	80	A Review of Methods for Constructing Brain Functional Atlas Based on Neuroimaging Data and Machine Learning
		Yang Mengting, Zhang Daoqiang, Wen Xuyun
		DOI: 10.3969/j.issn.0258-8021.2024.01.009
		Brain atlas is an important tool in brain science research, including brain function exploration, neuroscience and cognitive science, and clinical diagnosis and treatment, which can be constructed using machine learning based on neuroimaging data. The parcellation patterns generated by brain atlas provide the foundation for understanding brain structure and function, and are frequently used for defining nodes in brain networks to reduce the impact of imaging noise on analysis results. Compared to structural atlases, functional atlases have a later development but demonstrate higher functional consistency, gradually gaining widespread attention and application in various brain function-related studies. In order to reveal the development path of functional atlases, based on the investigation of different types and methods of brain functional atlases constructed using neuroimaging data and machine learning, this article first classified and summarized the atlases according to multiple attribute features such as cortical surface and voxel, individual and population, and imaging modalities, providing detailed information for each atlas. After that, according to machine learning methods, we reviewed the construction methods of brain atlases based on graph clustering and time series clustering separately. Finally, we outlined the challenges faced in the field of brain atlas research and prospects for future research directions.
		2024 Vol. 43 (1): 80-97 [Abstract] ( 190 ) HTML (1 KB) PDF (2093 KB) ( 468 )

	98	Brain Age Prediction Methods and Applications Based on Multimodal Neuroimaging Data
		Liu Shuang, Yu Jing, Chen Yuanyuan, Fan Qiuyun, Zhao Xin, Ming Dong
		DOI: 10.3969/j.issn.0258-8021.2024.01.010
		Brain age can predict the degree of brain maturity and aging by modeling and analysis of neuroimaging data. With the development of artificial intelligence algorithms, the related research on brain age prediction have demonstrated a rapid emerging trend. It is generally recognized that brain age can be an effective biomarker for monitoring abnormal development and aging, which can assess individual brain health, and has great potential to detect abnormal aging and disease. In the context of rapid growing of research interests on the brain age prediction, this review summarized the latest achievements from the aspects of brain age classification, brain age model, clinical application, and further discusses challenges and developing directions of brain age in the future studies.
		2024 Vol. 43 (1): 98-105 [Abstract] ( 215 ) HTML (1 KB) PDF (800 KB) ( 348 )

	106	Research Progress of Test-Retest Reliability Based on Scalp EEG
		Qin Huiyi, Wang Yulin, Lei Xu
		DOI: 10.3969/j.issn.0258-8021.2024.01.011
		In recent years, the test-retest reliability of EEG has been increasingly emphasized in the exploration of the cognitive neural mechanisms of psychological processes based on scalp electroencephalography (EEG). When developing experimental methods based on scalp EEG to evaluate human brain function, researchers are increasingly concerned about whether these methods are highly reliable. Starting from the main factors that affect the test-retest reliability of EEG, we focused on the impact of the basic experimental process of EEG on the test-retest reliability. Firstly, the commonly used methods for measuring the test-retest reliability of EEG were introduced. Then, from the perspective of experimental methods, the research progress in this field was reviewed, including experimental design, data processing, feature selection, subject group, etc. The main focus was on the impact of data analysis methods on the test-retest reliability. It was pointed out that improving the signal-to-noise ratio of EEG signals and standardizing the preprocessing process are important ways to improve the test-retest reliability of EEG. Finally, ways to improve the test-retest reliability of EEG research were proposed based on the selection of experimental paradigms and features or indicators, and prospects for the development of EEG test-retest research were presented.
		2024 Vol. 43 (1): 106-116 [Abstract] ( 138 ) HTML (1 KB) PDF (946 KB) ( 247 )

	117	Research Progress of Microcurrent Healing Promotion
		Tang Zhongyu, Dan Nianhua, Chen Yining
		DOI: 10.3969/j.issn.0258-8021.2024.01.012
		In recent years, promoting the healing of skin wounds and bone injuries has been one of the focuses in the field of promoting healing. Microcurrent healing has attracted researchers’ attention because of its effective and rapid healing properties and great healing effect. The current research on microcurrent healing can be divided into two categories: the active and the passive. Active microcurrent is widely used to promote chronic wound healing because the current, voltage and other parameters are easy to control. The passive microcurrent is mainly used to promote the healing of bone injury as it has the ability of transforming mechanical energy into electrical energy. In the part of active microcurrent promoting healing, the similarities and differences of direct current, alternating current and mixed electric stimulation were described in the context of different external power supply. Based on the different sources of passive microcurrent, the principles and differences of chemical microbatteries, piezoelectric ceramics, piezoelectric flexible materials and nano-generators were described in the section of passive microcurrent promoting healing. The internal mechanism was also discussed and the development prospect was prospected as well.
		2024 Vol. 43 (1): 117-128 [Abstract] ( 173 ) HTML (1 KB) PDF (10109 KB) ( 273 )

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