Chinese Journal of Biomedical Engineering

Home

Editorial Board

Download

Quick Search

Office Online

Current Issue

Download Articles

Download

Links

2026 Vol. 45, No. 1 Published: 2026-02-20

Reviews Communications Regular Papers

Regular Papers

	1	Identification of Children with Autism Spectrum Disorder via Multi-Modal HyperdimensionalComputing with EEG and Eye-Tracking Data
		Wang Sha, Jiang Guoqian, Han Junxia, Xie Ping, Li Xiaoli
		DOI: 10.3969/j.issn.0258-8021.2026.01.001
		Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by core deficits in social communication and the presence of restricted, repetitive behaviors. Its high heterogeneity poses substantial challenges for early identification and diagnosis. Traditional approaches often rely on single-modality data and are constrained by small sample sizes and poor generalization, limiting their utility in clinical decision support. To address these issues, this study proposed a multimodal fusion recognition framework based on hyperdimensional computing (MMHDC). The framework encoded electroencephalography (EEG) and eye-tracking data from 40 children with ASD and 50 typically developing children into hyperdimensional vectors, leveraging the complementary information between neurophysiological and behavioral signals through fusion modeling to enhance recognition accuracy and model robustness. Experimental results showed that the MMHDC model achieved an accuracy of 86.82% using only 5% of the training data, significantly outperforming mainstream methods such as support vector machines, naive Bayes,extreme gradient boosting, and multimodal stacked denoising autoencoders. Further ablation studies demonstrated that combining EEG and eye-tracking features effectively improved discriminative power, highlighting the advantages of a multimodal strategy. By integrating small-sample learning with hyperdimensional encoding theory, this work provided a lightweight and efficient new approach for the early identification of children with ASD, with strong potential for practical deployment and clinical application.
		2026 Vol. 45 (1): 1-10 [Abstract] ( 72 ) HTML (1 KB) PDF (4626 KB) ( 45 )

	11	Retinal Segmentation Method Based on Multi-Scale Bitemporal Fusion Module and GlobalGrouped Coordinate Attention
		Liu Xuepeng, Xu He, Ji Yimu, Li Peng
		DOI: 10.3969/j.issn.0258-8021.2026.01.002
		Retinal vessel segmentation plays a crucial role in diagnosing retinal vascular diseases. Recent studies based on UNet have demonstrated promising performance in this task. However, segmenting thin retinal vessels with low contrast often leads to the loss of spatial information at certain stages. To address this issue, this paper proposed the novel network architecture, BFM-GGCA-UNet. The model employed convolutional kernels of different sizes at various stages to process feature maps, effectively capturing multi-scale features and forming a comprehensive feature representation. Furthermore, a multi-resolution convolutional interaction mechanism was introduced to expand the receptive field in both horizontal and vertical directions while maintaining the full image resolution. Additionally, to enhance this directional expansion, a global grouped attention mechanism was proposed. This mechanism leverages shared convolutional layers and an attention module to generate attention maps for the height and width dimensions, which weight the input feature maps to accentuate crucial features for a more precise prediction map. The proposed model was evaluated on five public datasets: DRIVE, STARE, CHASE_DB1, HRF and ARIA. The results showed that our model achieved AUCs of 98.89%, 99.50%, 99.13%,98.87%, and98.72%, and accuracies of97.25%, 98.06%,97.49%, 97.96% and 96.81% on these datasets, respectively, outperforming most existing methods. In conclusion, the designed BFM-GGCA-UNet effectively improved segmentation accuracy and delivered superior performance on these five fundamental retinal vessel segmentation datasets.
		2026 Vol. 45 (1): 11-24 [Abstract] ( 55 ) HTML (1 KB) PDF (15490 KB) ( 16 )

	25	Liver Vessel Segmentation in CT Images Based on Vascular Skeleton Feature Constraints
		Tao Siyu, Ji Xu, Liu Qiegen, Chen Yang, Tang Hui
		DOI: 10.3969/j.issn.0258-8021.2026.01.003
		Computed tomography (CT) is widely used in the preoperative planning stage of liver ablation surgery to assist doctors in making surgical plans. It is important to segment liver vessels from abdominal CT images. Due to the complex structure of liver blood vessels and the low contrast between blood vessels and surrounding tissues, it is difficult to accurately segment liver blood vessels from CT images. Most of the current segmentation methods only focus on the pixel distribution of blood vessels and ignore the structure information of blood vessels, so that the segmentation results often have fractures and holes. To solve the above problems, this paper proposed a liver vessel segmentation method based on vascular skeleton feature constraint and applies vascular skeleton features to the segmentation neural network. The detail process of the method included following steps: 1) A blood vessel enhancement step was added to the data preprocessing, then the blood vessel enhanced image and the original image were used as the input of the neural network, so as to introduce vascular spatial attention into the network input. 2) A skeletonization module based on Euler features was added to the post-processing procedure of the segmentation network and the output vascular skeleton feature was added to the loss function to design a joint loss function (Dice_MS_Loss) including Dice loss and morphological skeleton loss (MS_Loss), which serves as a constraint to promote the network’s topology preserve ability. In this study, 9 cases were selected from the public dataset IRCAD, and 29 cases were selected from the dataset MSD8. A total of 38 cases were used for method evaluation. The results of five-fold cross experiment showed that the proposed method was superior to other SOTA methods in terms of quantitative evaluation indicators, with the Dice coefficient of 0.749, the centerline Dice (clDice) of 0.79 and the sensitivity (Sen) of 0.754. The visual effects of the experiments showed that the proposed method was able to effectively segment liver blood vessel structures with less fractures and deficiencies.
		2026 Vol. 45 (1): 25-37 [Abstract] ( 43 ) HTML (1 KB) PDF (7473 KB) ( 15 )

	38	Breast Tumor Prediction Based on Multiscale Residual Spatiotemporal Model
		Zhang Liangliang, Zheng Hang, Liu Jiawei, Wang Zhenzhen, Bi Kejian
		DOI: 10.3969/j.issn.0258-8021.2026.01.004
		Spatial and temporal features obtained from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) are both used for the prediction of breast tumors. Previous studies mainly focused on the shallow spatiotemporal features, and the effectiveness of the deep spatiotemporal features is still unclear. To this end, a multiscale residual spatiotemporal model (MRSM) based on ResNet (2+1)D was proposed for breast tumor prediction. This study included 232 patients from Taizhou Central Hospital, with 85 benign cases and 147 malignant cases. The DCE-MRI data consisted of one precontrast image and eight postcontrast images.Compared with the original ResNet (2+1)D, MRSM method added a multiscale module and residual module. Specifically, the multi-scale module employed two independent dual-path spatial convolutions and dual-path temporal convolutions to effectively capture the spatio-temporal features of the breast tumor. The performance of the MRSM method was evaluated using the area under the receiver operating characteristic (ROC) curve (AUC). Compared to the ResNet (2+1)D method (AUC=0.974 0) or existing methods (The best AUC=0.960 8), this method achieved a higher AUC of 0.987 0.In conclusion, the MRSM method is an effective approach for accurate prediction of breast tumors.
		2026 Vol. 45 (1): 38-46 [Abstract] ( 41 ) HTML (1 KB) PDF (4101 KB) ( 13 )

	47	Prediction of Multi-Ethnic Pathological Information Based on Multi-Parameter MRI
		Liu Jiabao, Fan Ming, Li Lihua
		DOI: 10.3969/j.issn.0258-8021.2026.01.005
		Magnetic resonance imaging (MRI) plays a crucial role in the diagnosis and treatment of breast cancer. Among its modalities, dynamic contrast-enhanced MRI (DCE-MRI), apparent diffusion coefficient (ADC) imaging, and T2-weighted imaging (T2WI) provide comprehensive imaging information that offers valuable insights for predicting the pathological characteristics of breast cancer. In this study, we integrated multi-parametric imaging features derived from DCE-MRI, ADC, T2WI, delayed-phase imaging, DCE-MRI-based tumor subregions, tumor pharmacokinetics, and quantitative background parenchymal enhancement (BPE) features to conduct predictive analyses on both multi-ethnic classification and Ki-67 expression levels. A total of 781 breast cancer cases were collected. Following preprocessing, tumor regions and fibroglandular tissue (FGT) on multi-parametric MR images were delineated. Tumor subregions were segmented using pixel-based enhancement clustering into three enhancement patterns. Radiomics features were extracted from both tumors and FGT. In addition, pharmacokinetic features of the tumors on DCE-MRI, the percentage of FGT volume within the entire breast (%BPEF/B) and the proportion of enhancing FGT volume exceeding different signal intensity thresholds within the FGT (%BPEF/F) were calculated. An unsupervised feature selection strategy was employed to eliminate irrelevant features. A genetic algorithm combined with cross-validation was used to identify the optimal feature subset. A random forest classifier was constructed for prediction, and model performance was evaluated using the area under the receiver operating characteristic curve (AUC) and ROC curve plotting. Furthermore, a feature-level fusion approach was adopted to integrate radiomic, kinetic, and BPE features, thereby constructing a multi-parametric imaging fusion model. In single-modality models for the multi-ethnic classification task (691 Han, 70 Tibetan, and 20 Yi patients), ADC images achieved the best AUCs of 0.841±0.024, 0.819±0.031, and 0.775±0.092 for Han, Tibetan, and Yi ethnic groups, respectively. For the Ki-67 expression prediction task (103 low expression, 498 high expression cases), the Fast Flow subregion yielded the best AUC of 0.756, with a sensitivity of 0.845 and specificity of 0.684. Upon fusion of multi-parametric imaging and BPE features, the AUCs for the multi-ethnic prediction task improved to 0.909±0.027 (Han), 0.890±0.031 (Tibetan), and 0.863±0.070 (Yi), which improved by 8.1%, 8.7%, and 11.3%, respectively compared to the best single-parameter model. For Ki-67 prediction, the fusion model incorporating ADC, DCE-MRI, and BPE features yielded an improved AUC of 0.770, with sensitivity and specificity increasing to 0.866 and 0.728, respectively. In summary, the integration of multi-parametric MRI features, quantitative BPE metrics, and tumor subregion enhancement patterns provides a more comprehensive characterization of breast cancer imaging and underlying biology. This approach enhances diagnostic accuracy and offers valuable guidance for personalized treatment and prognostic assessment in breast cancer patients across different ethnic groups.
		2026 Vol. 45 (1): 47-60 [Abstract] ( 44 ) HTML (1 KB) PDF (7980 KB) ( 8 )

	61	Study on Temperature Field of Biological Tissues During Low-Frequency Ultrasound TransdermalDrug Delivery
		Zhao Ke, He Bingbing, Zhang Yufeng
		DOI: 10.3969/j.issn.0258-8021.2026.01.006
		In the process of low-frequency ultrasound transdermal drug delivery, effects of the ultrasound frequency parameters on the temperature of biological tissues are still unclear. In this paper, finite element analysis (finite element analysis and in vitro experiments of biological tissues) was used to study the effects of low-frequency ultrasound of different frequencies on the temperature distribution of biological tissues, aiming to achieve the therapeutic effect without causing thermal damage to the skin layer of biological tissues. In order to verify the effectiveness of the model, a self-developed low-frequency ultrasonic energy output device was used to perform anex vivo pork test, and the temperature field distribution and temporal temperature changes of the isolated tissues under low-frequency ultrasound were monitored by a FLIR infrared thermal imager. The results of finite element analysis showed that after 20 minutes of low-frequency ultrasound, the temperature inside the biological tissue decreased with the increase of ultrasound frequency, and the temperature of the fat layer was the highest, reaching 40.89℃, and the temperature change of the skin layer displayed by the domain probe met the safety requirements of transdermal drug administration of low-frequency ultrasound. Different ultrasound frequency parameters have a significant effect on the temperature change of biological tissues under the action of low-frequency ultrasound. The in vitro experiments under the same ultrasonic parameter settings showed that the temperature field distribution and its maximum temperature value in the biological tissue were basically consistent with the values detected by the domain point probe of the simulation experiment, and the temperature difference was within 1℃, which is within a reasonable error range. From the simulation results and in vitro experimental results, it could be seen that the above ultrasound parameter settings did not cause thermal damage to skin tissue and achieved the purpose of effective treatment. The model established in this study is expected to predict the temperature change in tissues and provide a reference for the action time and parameter setting of low-frequency ultrasound transdermal drug delivery.
		2026 Vol. 45 (1): 61-69 [Abstract] ( 35 ) HTML (1 KB) PDF (3884 KB) ( 11 )

	70	Biomechanical Effect of Different Widths of Tibial Component on Two-Component Total AnkleProstheses Post Total Ankle Arthroplasty
		Liu Jing, Xu Yangyang, Lu Da, Wu Yong, Pei Baoqing
		DOI: 10.3969/j.issn.0258-8021.2026.01.007
		The two-component total ankle prosthesis has shown favorable short-term clinical outcomes in treating patients with end-stage ankle osteoarthritis; however, its compatibility with the tibial anatomy remains a critical determinant of clinical performance. This study aimed to investigate the biomechanical effects of different tibial component widths in the two-component total ankle prosthesis on ankle joint mechanics. Based on weight-bearing foot and ankle CT imaging data, five three-dimensional (3D) finite element models of ankle joint replacement were constructed, with tibial component widths of 20, 22, 24, 26, 28 mm. Finite element analyses were conducted under various loading conditions representative of those experienced during the gait cycle. The validity of the ankle joint model was confirmed through mesh convergence analysis and by comparing stress distributions in the tibia, fibula, and talus with values reported in the literature. Results showed that increasing the tibial component width from 20 mm to 26 mm progressively reduced the peak tibial stress under gait loading, with a maximum decrease of 32.94%, thereby promoting a more uniform stress distribution. At this width, prosthesis micromotion was consistently below 50 μm, indicating optimal initial stability of the model. Increasing the tibial component width to 28 mm led to a peak stress rise of up to 36.3%, accompanied by a reduction in prosthesis stability. At a tibial component width of 26 mm, the cross-sectional area ratio at the top tangent plane of the distal tibial articular surface was 71.91%. Therefore, when selecting the width of a two-component total ankle prosthesis, it is advisable to approximate this ratio and match the corresponding prosthesis model to enhance stability and ensure favorable clinical outcomes.
		2026 Vol. 45 (1): 70-78 [Abstract] ( 41 ) HTML (1 KB) PDF (3702 KB) ( 18 )

Reviews

	79	Research Progress of Brain Tumor MRI Image Diagnosis Based on Deep Learning
		Jiang Liang, Ma Xingmin, Wang Hua
		DOI: 10.3969/j.issn.0258-8021.2026.01.008
		Brain tumors are malignant diseases caused by abnormal proliferation of brain cells, seriously threatening the life of patients. Magnetic resonance imaging (MRI), as a non-invasive and clear diagnostic tool, is widely used in the diagnosis of brain tumors. In recent years, deep learning technology has made breakthrough progress in the field of medical image analysis, providing new approaches for the diagnosis and lesion localization of brain tumors. This article reviewed the application progress of deep learning in MRI images of brain tumors, mainly elaborating from three aspects: multi-scale feature extraction, lesion segmentation and localization, and classification and grading. The application of generative models in alleviating the scarcity of MRI data was summarized, and the advantages of federated collaborative learning in multi-institution and multi-data fusion were introduced as well. This paper pointed out that deep learning is still faceing challenges such as insufficient model interpretability and scarce data in the analysis of brain tumor images, at the same time, its future development directions were discussed.
		2026 Vol. 45 (1): 79-86 [Abstract] ( 47 ) HTML (1 KB) PDF (814 KB) ( 22 )

	87	Analytical Methods of EEG Microstates and Applications in Cognitive Impairment Patient
		Li Zipeng, Li Xin, Qu Zhongjie, Su Rui, Yin Bowen, Yin Liyong
		DOI: 10.3969/j.issn.0258-8021.2026.01.009
		Electroencephalogram (EEG) microstates can characterize the spatiotemporal dynamic network reorganization in the brain by capturing transient stable patterns of global potential distribution. The core value lies in resolving the discontinuous transitions and non-stationary conversion characteristics of functional brain states. However, conventional analysis methods for the microstates face challenges in standardization due to technical heterogeneity in workflows and classification uncertainty, limiting the comparability of cross-study results. This review systematically elaborated the fundamental principles and extraction procedures of microstate analysis, covering key technical steps including signal preprocessing, frequency band optimization, and clustering method selection. Rresearch advancements in traditional feature parameters and dynamic syntactic characteristics were summarized. Focusing on cognitive impairment research, this article discussed current applications of the microstate analysis in patients with Alzheimer′s disease (AD) and mild cognitive impairment (MCI), including feature characterization, disease diagnosis, and cognitive assessment. Finally, potential challenges and future research directions were discussed, with the aim of providing valuable references for researchers in this field.
		2026 Vol. 45 (1): 87-99 [Abstract] ( 40 ) HTML (1 KB) PDF (3815 KB) ( 18 )

	100	Application of Nanomedical Sensors in Tumor Marker Detection
		Qin Wei, Li Jiahui, Chang Chunrui, Zhao Lijun
		DOI: 10.3969/j.issn.0258-8021.2026.01.010
		The high mortality rate of cancer underscores the critical importance of early diagnosis, with biomarker detection serving as a pivotal approach that plays a vital role in routine physical examinations, preventive screening, recurrence monitoring, and other aspects of tumor-related healthcare. This article focuses on the applications of nanomedical sensors in tumor biomarker detection. Following a brief introduction to tumor biomarkers, medical sensors, and nanomaterials, we provided detailed discussions on biosensors based on various nanomaterials, including metallic nanoparticles, graphene, and magnetic nanoparticles, which utilize electrochemical or fluorescent signaling mechanisms. These sensors are applied to detect glycoprotein antigens, carcinoembryonic antigens, enzyme-based biomarkers, circulating tumor cells, nucleic acidsand other molecular markers. We highlighted their potential in early tumor diagnosis, disease progression monitoring, and prognostic follow-up, aiming to provide actionable insights for identifying patients′ latent malignancies and assisting clinicians in diagnosis. Looking ahead, nanomedical sensors are expected to integrate with artificial intelligence technologies to enhance detection performance, reduce costs, simplify operational workflows, and prioritize biocompatibility and safety of materials, thereby advancing the development of early-stage cancer diagnosis and treatment.
		2026 Vol. 45 (1): 100-113 [Abstract] ( 39 ) HTML (1 KB) PDF (3948 KB) ( 13 )

Communications

	114	Construction of a BREP Curved Surface Phantom for a Chinese Pregnant Woman Based on MRIImages
		Zhang Haowei, Sun Yueyang, Zhang Tiangui, Xia Mingchen, Liu Ying, Lu Heqing
		DOI: 10.3969/j.issn.0258-8021.2026.01.011

		2026 Vol. 45 (1): 114-118 [Abstract] ( 32 ) HTML (1 KB) PDF (3198 KB) ( 10 )

	119	A Study on the Assessment of Flight Personnel′s Cognitive Abilities Using a Multiple-ObjectTracking Experimental Paradigm
		Tian Wei, Wang Xue, Zhang Chi , He Yiting, Du Changwei, Cong Fengyu
		DOI: 10.3969/j.issn.0258-8021.2026.01.012

		2026 Vol. 45 (1): 119-123 [Abstract] ( 43 ) HTML (1 KB) PDF (4237 KB) ( 6 )

	124	Design and Performance Evaluation of a Bio-inspired Prosthetic Knee Joint with a VariableTransmission Ratio Mechanism for Active-Passive Hybrid Drive
		Du Yanchen, Li Yuanhua, Wang Xiaoming, Ye Yingguo, Wang Xu, Yu Hongliu
		DOI: 10.3969/j.issn.0258-8021.2026.01.013

		2026 Vol. 45 (1): 124-128 [Abstract] ( 47 ) HTML (1 KB) PDF (4577 KB) ( 10 )

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