Home    About Journal    Editorial Board    Instruction    Subscribe    Download    Messages Board    Contact Us    中文     Quick Search

Office Online     Current Issue Accepted Current Issue Archive Adv Search Read Articles Download Articles Email Alert     Download More>>     Links More>>

2025 Vol. 44, No. 6 Published: 2025-12-20 Reviews Regular Papers Indexes        Regular Papers 641 Cognitive Status Prediction for Alzheimer′s Disease via the Attention and LSTM-enhanced Neural Network Yang Zelin, Yang Kai, Ren Xiaomei, Xu Jun, Zeng Xian, Wan Zhuo, Huang Yunzhi DOI: 10.3969/j.issn.0258-8021.2025.06.001 Alzheimer′s disease (AD) is an irreversible neurodegenerative disease, and mild cognitive impairment (MCI) is the intermediate transition state from normal control (NC) to AD, thereby MCI is a crucial monitoring state of early intervention treatment for AD. This motivates us to propose a model for predicting the cognitive status and the possible translation in the future. Our model employed the multilayer perceptron (MLP) for cognitive status recognition based on the longitudinal electronic healthy records (EHR). As the recorded data contained longitudinal changes, we integrated the attention block and long-short-term memory network (LSTM) module with the backbone MLP. In this way, our model captured both the spatial- and temporal-relationship across features at multiple timepoints. Two type experiments were conducted on 1 313 cases from ADNI dataset and 5 209 cases NACC dataset, consisting of a baseline visiting record and the following-up records every 6 months: (1) single-time-point cognitive status prediction with multiple previous visiting records, and (2) multiple-time-point cognitive statuses simultaneously prediction with merely previous single-time-point visiting record. The proposed method made a desirable prediction on the cognitive status. Especially, based on the trained models over the 2nd, 3rd, 5th, and 6th time point records, we achieved average F2 scores of 0.87, 0.88, 0.91, and 0.91 when predicting the corresponding subsequential single-time-point cognitive status, respectively; and achieved average F2 scores of 0.84, 0.86, 0.88, and 0.89 when predicting the corresponding next two-, three-, and four-time points cognitive status, respectively. Compared to the gated recurrent unit, the corresponding F2 scores were improved by 0.03, 0.01, 0.03, and 0.02 when predicting the corresponding subsequential single-time-point cognitive status, respectively, and the average F2 scores were improved by 0.04, 0.05, 0.02, and 0.02 when predicting the corresponding next two-, three-, and four-time points cognitive status, respectively. These results demonstrated a superior performance of the proposed method in predicting cognitive status as compared to the state-of-the-art methods, e.g., MLP, gated recurrent unit and the traditional machine learning classifiers, and can be useful for early intervention treatment for AD patients. 2025 Vol. 44 (6): 641-653 [Abstract] ( 47 ) HTML (1 KB)  PDF (11993 KB)  ( 32 ) 654 Research on Complex Image Contour Detection Model Based on Avian Visual Features Wu Wei, Lou Kaiwen, Fan Yingle, Fang Tao DOI: 10.3969/j.issn.0258-8021.2025.06.002 This work aimed to address the limitations of traditional contour detection methods in extracting contour information from complex image objects on the basis of superior mechanisms of bird visual system in the image information processing. Firstly, by utilizing the preferred orientation mechanism of birds, the optimal direction selection for the receptive field was determined to enhance contour responses and reduce noise interference. Secondly, a color pre-segmentation model is constructed based on the specific wavelength filtering mechanism of bird oil droplets, which enhanced the color difference between the target and the background. Additionally, a color difference calculation model based on the spectral response characteristics of bird photoreceptors was proposed to avoid fragmentation of color channels in the image. Finally, inspired by the excitatory input mechanism of the avian optic tectum (TeO) for targets and the inhibitory projection mechanism of avian thalamic nuclei for backgrounds, the fine extraction of complex image contour and the suppression of image texture were achieved. Comparative experiments were conducted on 1449 images from the NYUD dataset and 500 images from the BSDS500 dataset, with ODS, OIS, and AP values of 0.68, 0.69, and 0.68, respectively, demonstrating superior performance compared to mainstream comparison methods. In conclusion, this study applied the excellent visual information processing mechanisms of birds to contour detection tasks, improving existing computational models and providing new ideas and methods for visual tasks. These findings are of significance in promoting the development of bioinspired visual models. 2025 Vol. 44 (6): 654-665 [Abstract] ( 23 ) HTML (1 KB)  PDF (8915 KB)  ( 12 ) 666 Multi-label ECG Classification Based on Knowledge Distillation and Label Relevance Zhou Jinghao, Wang Xingyao, Zhang Shuo, Zhao Lulu, Li Jianqing, Liu Chengyu DOI: 10.3969/j.issn.0258-8021.2025.06.003 In electrocardiogram (ECG) signal classification, patients often have multiple coexisting diseases, making multi-label classification more challenging and clinically meaningful. In deep learning, large discrepancies in the number of samples often lead to label imbalance. The model tends to underperform on labels with fewer examples. Label correlation can help alleviate this issue. This paper compressed a Teacher Model into a Student Model. The loss function of the Student Model was based on both ground-truth labels and intermediate outputs from the Teacher Model. This paper proposed an approach to constructing a label correlation matrix. The label correlation matrix used cosine similarity to capture inter-label relationships. The correlation matrix was then used to calibrate the predicted probabilities. Hefei-Tianchi contains 18 232 raw ECG recordings,containing 38left bundle branch conduction block data. PTB_XL contains 21 836 raw ECG data,containing 51 left bundle branch conduction block raw ECG data. The new model helped to alleviate the problem of label imbalance. For left bundle branch block, F1 score on the Hefei-Tianchi dataset increased from 0.799 to 0.837. For pacing heart rhythm, it increased from 0.862 to 0.876. For First-degree atrioventricular block, it increased from 0.862 to 0.876, higher than ECGNet′s 0.535 and Acharya′s 0.810. For PTB_XL dataset, First-degree atrioventricular block, increased from 0.679 to 0.857. For Right ventricular hypertrophy, it increased from 0.698 to 0.745. For Complete left bundle branch block, it increased from 0.649 to 0.887. Meanwhile, on the Hefei-Tianchi dataset, the F1 score of the Student Model was 0.816, while that of the Teacher Model was 0.824. On the PTB_XL dataset, the F1 score of the Student Model was 0.876, while that of the Teacher Model was 0.873. The size of the Student Model was only 0.62 times of the Teacher Model. The established model addressed the long-tail distribution issue, improving predictive capability for few-shot labels, while simplifying parameters and saving computational resources. 2025 Vol. 44 (6): 666-673 [Abstract] ( 25 ) HTML (1 KB)  PDF (2420 KB)  ( 26 ) 674 Parameter Identification and Comparative Verification of Different COPD Respiratory MechanicsModels Based on Constrained Optimization Method Xu Liqiang, Qiao Huiting, Zhang Chi, Xia Jingen, Li Deyu DOI: 10.3969/j.issn.0258-8021.2025.06.004 In intelligent closed-loop ventilation control, the physiological model of the respiratory system must be consistent with the patient′s actual condition, and its parameter estimation must be real-time and effective.Using the parameter estimation method of constrained optimization, models that can reflect the pathological conditions of chronic obstructive pulmonary disease (COPD) patients as well as can be used for real-time estimation were compared and verified. In this study, we used numerical simulation data and clinical data to estimate parameters of respiratory mechanics models with different complexities. The estimated results were used to reconstruct airway pressure. By comparing the estimation results and reconstruction results among different models, the mathematical description of respiratory mechanics problems suitable for mechanical ventilation COPD was selected. For four models with the different complexities, the corresponding parameter estimation methods were established. By using the simulation and clinical data, a constrained optimization parameter estimation method was applied to different models. Comparative analysis between the reconstructed airway pressure (using estimated resistance and elasticity) and input airway pressure demonstrated that the elastic nonlinear and resistance nonlinear models exhibited the smallest root mean square error (RMSE) in simulation data. For the 12 respiratory cycle data of 6 clinical cases, the resistance nonlinear model achieved the lowest RMSE (M=1.55 cmH2O, SD=0.69 cmH2O) among the evaluated models. The multiple comparison results of repeated measures ANOVA showed that the resistance nonlinear model statistically significantly outperformed the other models (P<0.05). The results showed that according to the simulation data and clinical data in this study, the resistance nonlinear model was more appropriate model of ventilated COPD. Based on the application of constraint optimization method for real-time parameter estimation of mechanical ventilation, the resistance nonlinear model reflected the respiratory mechanics characteristics of COPD better. The results can be used for monitoring respiratory mechanics of COPD, and also provide a basis for the intelligent parameter setting and setting of ventilators. 2025 Vol. 44 (6): 674-684 [Abstract] ( 20 ) HTML (1 KB)  PDF (1057 KB)  ( 9 ) 685 Macroscale Neural Computational Modeling Simulation of Propofol Incorporating Pharmacological Effects Wang Dihuan, Liang Zhenhu, Li Xiaoali DOI: 10.3969/j.issn.0258-8021.2025.06.005 To investigate the neural action mechanism of propofol at the macroscale,an unscented Kalman filter framework based on the incorporating pharmacological effects neural mass model (PENMM) was proposed in this paper. The framework regulated the neural effect parameters (G, b, and g) with the optimal weights of effector compartment concentrations, while tracking the real electroencephalogram (rEEG) and estimating the neuromodulation parameters (A, B, and a) of the model for fitting the simulated electroencephalogram (sEEG). The cumulative errors, correlations, and coefficients of determination of rEEG and sEEG were calculated, which indicated that rEEG and sEEG were well correlated, proving that the framework could effectively estimate NMM model parameters and generate sEEG. The results showed that the framework was able to effectively track EEG characteristics of nine volunteers in the effect of propofol. The highest correlation between rEEG and sEEG was 0.91. Parameter B and a curves were significantly different (p<0.001) in different anesthesia states and can be used to differentiate between different anesthesia states. In conclusion, this method holds a potential to be developed as a depth of anesthesia monitoring tool. 2025 Vol. 44 (6): 685-699 [Abstract] ( 24 ) HTML (1 KB)  PDF (4987 KB)  ( 8 ) 700 Research of Contact Near-Infrared Diffuse Correlation Tomography Data Processing andCerebral Blood Flow Reconstruction Zhang Xiaojuan, Li Zicheng, Wei Jiahui, Cao Xiangqian, Shang Yu DOI: 10.3969/j.issn.0258-8021.2025.06.006 Near-infrared diffusecorrelationblood flow tomography can provide important information for the trend of tumor. In response to the drawback of large fluctuations and abnormal g1(τ) curve slopes in phantom and clinical tests, a continuous descent-threshold method was proposed to screen the g1(τ) curves and fit the slope of the g1(τ) curve by the N-order linear algorithm. In addition, to solve the “concave artifact” caused by ambient light in the BFI image, the slope was corrected by the adjusted matrix, and finally BFI reconstruction was realized by the Bregman-TV reconstruction algorithm. Furthermore, voxels at the bottom edge were removed to reduce the ill-condition of equations, because little amount of S-D can detect them. Phantom experiments verified that the “concave artifact” was eliminated in reconstructed BFI images. The edge of the quasi-solid cross-shaped phantom anomaly were sharp and the contrast of the second layer (depth 10-15 mm) was 0.77. Both the contrasts of two layers were above 3 for 500 mL/h current velocity tubular anomaly. Clinical testing revealed that the 95% confidence interval of the t-distribution for BFI differences (supine rest and 70° head-up tilt) in 20 volunteers was [1.10 cm2/s, 1.44 cm2/s], excluding zero. The median contrast ratio obtained through Bootstrap resampling was 1.42 (95% CI [1.35, 1.50]), which was significantly greater than 1. These findings collectively indicated that the NL-DCT system with g1(τ) correction improved the sensitivity of in-vivo tissue microhemodynamic detection, enabling more precise measurement of posture-induced BFI changes. The system exhibited excellent hemodynamic imaging and abnormality detection performance, establishing its potential as a novel diagnostic tool for localized perfusion-related diseases. 2025 Vol. 44 (6): 700-710 [Abstract] ( 22 ) HTML (1 KB)  PDF (6272 KB)  ( 11 ) 711 An Intelligent Decision-Making Method for Rehabilitation Assistive Device Matching inSpinal Cord Injury Patients Based on K-Modes Clustering and Weighted K-Nearest Neighbor Li Sujiao, Shao Jiang, Liu Qi, Chen Shizheng, Yu Hongliu DOI: 10.3969/j.issn.0258-8021.2025.06.007 The injury segments, nature, and limb muscle strength characteristics of patients with spinal cord injuries are complex and variable. Inappropriate rehabilitation assistive device fitting has long hindered the overall rehabilitation process and efficacy for such patients. This study proposed a smart assistive device fitting decision-making method (KKA) that integrated K-modes clustering with weighted KNN, aiming to enhance the precision and efficiency of assistive device fitting. The study collected clinical data from 600 spinal cord injury patients across different hospitals. A 13-member expert panel comprising cross-disciplinary, cross-professional, and frontline assistive device fitting expert′s consensus determined nine feature attributes, including injury segment, injury nature, limb muscle strength and different fitting schemes. The KAA method comprised three key steps: (1) using entropy weighting and the analytic hierarchy process to determine feature weights and construct a weighted feature space; (2) optimizing the case library structure using the K-modes clustering algorithm to effectively reduce classification bias; and (3) combining weighted KNN case retrieval with a rule-based inference model to generate personalized adaptation prescriptions. The algorithm′s performance was validated using 600 static medical records and 40 medical records tracked dynamically over a one-month period. The study demonstrated that KAA performed exceptionally well in the adaptation of four types of assistive devices, achieving average accuracy and recall rates of 94.6-97.9% and 89.5-97.2%, respectively, indicating that the KAA algorithm possessed high effectiveness and reliability in the field of rehabilitation assistive product recommendation. Additionally, dynamic clinical tracking assessments indicated that the optimal threshold for SCIM improvement in the KAA algorithm was 24.2%, while the optimal threshold for ICF improvement was 2.3%. The AUC values for both SCIM and ICF improvement levels were greater than 0.7, indicating that the KAA method had high diagnostic value in assessing the effectiveness and applicability of rehabilitation assistive product fitting for patients. In conclusion, the KAA method demonstrated significant advantages in terms of fitting accuracy, efficiency, and rehabilitation outcomes, and is expected to provide reliable technical support for precise rehabilitation in spinal cord injury patients. 2025 Vol. 44 (6): 711-719 [Abstract] ( 15 ) HTML (1 KB)  PDF (1363 KB)  ( 12 ) 720 Experimental Study on the Mechanical and Electrical Properties of Defective Bone-CartilageUnits Xue Yanru, Zhong Hao, Guo Li, Feng Haoyu, Wu Xiaogang, Chen Weiyi DOI: 10.3969/j.issn.0258-8021.2025.06.008 The health of the bone-cartilage unit is crucial for maintaining normal joint movement and function. By measuring and analyzing changes in mechanical and electrical properties of the bone-cartilage unit, it is possible to achieve early diagnosis of joint diseases, which has potential value for developing new treatment methods to promote cartilage repair. However, current research on the mechanical and electrical properties of the bone-cartilage unit is relatively limited, and the mechanisms of changes in these properties are not well understood. In this work, we investigated the mechanical properties, electrical properties, and electro-mechanical effects of defective bone-cartilage units.For the bone-cartilage unit harvested from medial tibial plateau defects of adult New Zealand rabbits (n=36, allocated into 4 groups), compression-relaxation tests were performed using a universal material testing machine,double-electrode method was used to measure the electrical impedance of bone-cartilage units with different defect radius (R), the force-electric effect was tested by Instron 3343 universal material testing machine, multichannel USB data acquisition system, DH3820N distributed stress-strain test system and computer for data collection. The results showed that as the R increased to 0.75 mm, the maximum load-bearing capacity of the bone-cartilage unit decreased to 116 N, and the relaxation time gradually rose to 14.26 s. The peak voltage of the force-electric effect at a loading rate of 6 N/s droped to 67.57 mV; the impedance gradually increased, and the absolute value of the phase angle gradually decreased. The peak voltage of the bone-cartilage unit with different R was proportional to the peak load and the loading rate. Consequently, there existed significant differences in the mechanical properties, electrical characteristics, and force-electric effects between different R bone-cartilage units. 2025 Vol. 44 (6): 720-729 [Abstract] ( 17 ) HTML (1 KB)  PDF (5241 KB)  ( 9 )        Reviews 730 Research Advances in DCE-MRI-Based Breast Lesion Segmentation and Benign-Malignant Classification Methods Huang Kaiyang, Li Xiujuan, Xie Yuanzhong, Hou Jixue, Han Baosan, Nie Shengdong DOI: 10.3969/j.issn.0258-8021.2025.06.009 Breast cancer is one of the most common malignant tumors threatening women′s health, and early and accurate diagnosis plays an important role in the prognosis of breast cancer patients. Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) provides an important technical support for segmentation of breast lesions and discrimination of benign and malignant by virtue of its high resolution and dynamic flow analysis capability.In this article, the research progress of breast lesion segmentation and classification techniques based on DCE-MRI was reviewed, starting from non-automated traditional image segmentation and machine learning models based on manual feature extraction, moving on to deep learning techniques represented by convolutional neural networks, and further extending to analysis strategies integrating multi-modal imaging data as well as other emerging technologies. The technical characteristics and clinical application value of various methods were elaborated in detail. The current status of breast lesion segmentation and benign-malignant classification technologies was summarized, and future development directions were outlined. Specifically, emphasis should be placed on multi-center data collaboration, model interpretability, and development strategies integrating artificial intelligence with clinical practice, so that a safe and efficient intelligent auxiliary diagnosis system could be established to provide reliable support for the precise diagnosis and treatment of breast cancer. 2025 Vol. 44 (6): 730-737 [Abstract] ( 12 ) HTML (1 KB)  PDF (3124 KB)  ( 12 ) 738 Frontier Research and Applications of Surface Electromyography in Stroke Rehabilitation Liu Jiawei, Ge Xuanxuan, Liu Kai, Zhou Ping, Bao Tianzhe, Gong Weijun DOI: 10.3969/j.issn.0258-8021.2025.06.010 Rehabilitation assessment and intervention after stroke are major issues of clinical research and significantly challenging. Surface electromyography (sEMG) is a non-invasive electrophysiological technique that can effectively capture muscle electrophysiological data, providing objective and quantitative metrics for rehabilitation assessment and intervention. This article highlighted recent advancements in sEMG research and applications within stroke rehabilitation, exploring its essential roles in detecting post-stroke neuro-muscular changes, supporting rehabilitation therapies, and evaluating treatment efficacy, aiming to offer new insights for inpatient, community, and home-based stroke rehabilitation. In the first part, the application of analytical methods such as clustering index analysis and muscle synergy was discussed to demonstrate sEMG’s feasibility in analyzing neuro-muscular changes. The article then reviewed the research of sEMG on the movement therapy, botulinum toxin guidance, and prevention and treatment of complications, addressing advancements in efficacy assessment and clinical scale optimization. Finally, we indicated current limitations and clinical challenges of sEMG research and outlined future research directions. 2025 Vol. 44 (6): 738-747 [Abstract] ( 17 ) HTML (1 KB)  PDF (1553 KB)  ( 22 ) 748 Research Advances in 3D Bio-printing for Breast Cancer Shi Panpan, Zhai Weijia, Wang Xiaofeng, Zhang Yiqing, Deng Lili, Tang Ningning, Wang Yunlong DOI: 10.3969/j.issn.0258-8021.2025.06.011 Breast cancer is one of the most common malignant tumors among women worldwide, and its high heterogeneity poses significant challenges for precision therapy in medical research. The 3D bioprinting technology enables precise spatial control of cells, biomaterials, and biomolecules at the microscale, constructing complex tissue structures or microenvironments, thereby providing a novel platform for precision therapy of breast cancer. This paper systematically reviewed recent advancements in 3D bioprinting applications for breast cancer research, with particular focus on its implementation in constructing in vitro tumor models, guiding precision drug therapy, facilitating personalized surgical planning, and post-operative breast reconstruction. Current technical challenges and future development directions were critically discussed. This review aimed to provide valuable references for researchers and clinicians engaged in precision therapy of breast cancer for promoting the clinical translation of 3D bioprinting technology in breast cancer treatment. 2025 Vol. 44 (6): 748-758 [Abstract] ( 19 ) HTML (1 KB)  PDF (2120 KB)  ( 14 ) 759 Regulation and Mechanism of Ultrasound Bubble Technology on Tumor Microenvironment Xu Wenqi, Yan Yifei, Wang Shiwei, Xu Heyang, Chen Tiandong, Yang Fang DOI: 10.3969/j.issn.0258-8021.2025.06.012 Tumor blood vessels have an important impact on tumor growth and development as an important part of the tumor microenvironment and can induce tumor resistance to treatment as a key factor in tumor metastasis. The complex and variable heterogeneity of tumor blood vessels poses significant challenges to tumor imaging and treatment. Ultrasonic bubble technology is a relatively new treatment method based on micro/nanobubbles and ultrasound. Compared with traditional treatment methods, ultrasonic bubble technology has advantages of being non-invasive and more accurate, which can improve the tumor blood vessels and effectively treat the tumors. In recent years, this technology has shown great potential in tumor therapy and diagnosis, especially in the regulation of tumor microenvironment. This review aimed to discuss effects and mechanisms of ultrasound bubble technology on tumor blood vessels, and analyzed the different mechanisms in tumor vascular therapy applications. Furthermore, this review proposed promising prospects for the future of ultrasound bubble technology as a tool to modulate the tumor microenvironment. 2025 Vol. 44 (6): 759-768 [Abstract] ( 22 ) HTML (1 KB)  PDF (984 KB)  ( 16 )        Indexes 769 Contents Index 2025 Vol. 44 (6): 769-769 [Abstract] ( 14 ) HTML (1 KB)  PDF (413 KB)  ( 10 ) 770 Author Index 2025 Vol. 44 (6): 770-770 [Abstract] ( 9 ) HTML (1 KB)  PDF (381 KB)  ( 6 )

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