Chinese Journal of Biomedical Engineering

Home

Editorial Board

Download

Quick Search

Office Online

Current Issue

Download Articles

Download

Links

2025 Vol. 44, No. 5 Published: 2025-10-20

Reviews Communications Regular Papers

Regular Papers

	513	Adaptive Multi-Band Instantaneous Dynamic Functional Connectivity Method Based on Hilbert-Huang Transform
		Li Muchen, Song Zirui, Xu Yuyang, Zhou Yang, Wang Yanxi, Qing Zhao
		DOI: 10.3969/j.issn.0258-8021.2025.05.001
		Functional connectivity (FC) is an important indicator commonly used in the functional magnetic resonance imaging (fMRI) research to study neural synchrony between brain regions. Recent studies have indicated that FC exhibits frequency and time-varying characteristics, the latter being referred to as “dynamic functional connectivity” (dFC). This study addressed the lack of fixed standards for important parameters such as frequency bands and sliding window widths in the studies of FC frequency and temporal characteristics. We proposed to apply the Hilbert-Huang transform (HHT) to establish an adaptive cross-frequency FC time-frequency characteristic analysis method. The study utilized an fMRI dataset from the “Consortium for Reliability and Reproducibility” (CORR) at the Institute of Psychology, Chinese Academy of Sciences, which includes 50 participants with 2 scans each. After conventional preprocessing, the fMRI signals from 246 brain regions were extracted using the human brain connectome atlas. The HHT method was then applied to the fMRI signals to adaptively decompose them into intrinsic mode functions (IMFs) with different frequency characteristics. Using the instantaneous phases defined by HHT, dFC matrices were constructed for each time point based on the instantaneous phase differences between pairs of the 246 brain regions, and conventional dFC analysis was performed. The results showed that within the frequency band representing resting-state neural activity, namely 0.01~0.1 Hz, there were three IMFs (IMF1~IMF3) in the functional network connectivity pattern of the human brain, with Hilbert-weighted frequency distributions of (0.050±0.007), (0.022±0.005), and (0.011±0.004)Hz. K-means clustering classified the dFC matrices at different time points for each IMF into five brain connection states. The occurrence times of these five states in any two frequency bands exhibited significant clustering effects (chi-square test, P<0.001). The connection patterns of states that tend to occur simultaneously in different frequency bands also exhibited high spatial similarity (for example, the spatial similarity of the average connection matrices for the first connection state of IMF1 and IMF2 reaches R=0.90, P<0.001). In conclusion, this study revealed the existence of multi-frequency dynamic network patterns in the human brain and provided a multi-frequency instantaneous brain network analysis with higher temporal resolution, offering new insights into the analysis of fMRI data.
		2025 Vol. 44 (5): 513-522 [Abstract] ( 50 ) HTML (1 KB) PDF (2873 KB) ( 33 )

	523	MACD-Net: Spine Image Segmentation Based on Multi-Dimensional Attention and Cascaded Decoding
		Mei Xiajin, Ma Yuliang, Zhang Wenxin, Lv Qiang
		DOI: 10.3969/j.issn.0258-8021.2025.05.002
		Computed Tomography (CT) technology is crucial for the diagnosis and treatment of spinal diseases. However, the complex structure and unclear boundaries of the spine in spinal CT images lead to poor segmentation accuracy. To address these issues, a spinal image segmentation network based on multi-dimensional attention and cascaded decoding was proposed in this work. First, a multi-dimensional attention module was added to the encoder, which fully extracted and fused features from multiple dimensions, thereby enhancing the model's representation ability. Second, a cascaded refinement decoding structure was constructed in the decoder. Different segmentation tasks were designed for the two decoders, which decoded the encoded features successively to refine the segmentation boundaries step by step and ensured the recovery of fine features. Finally, the proposed MACD-Net was validated on the VerSe2019 dataset (comprising 160 CT images), with Dice similarity coefficient (DSC) values reaching 91.94 %, 91.59 %, and 72.69% for the cervical, thoracic, and lumbar vertebrae, respectively, and 95^th percentile Hausdorffdistance (HD95) values of 2.24 mm, 2.93 mm, and 6.90 mm, respectively. Compared with other models, the proposed model achieved significantly better segmentation results. Additionally, experiments conducted on the CTSpine1K dataset (with 300 randomly selected CT images) further verified the effectiveness of the model. In conclusion, this study demonstrated that MACD-Net achieved precise spinal image segmentation, thereby assisting doctors in diagnosis and treatment planning.
		2025 Vol. 44 (5): 523-532 [Abstract] ( 51 ) HTML (1 KB) PDF (3390 KB) ( 30 )

	533	Level II Ultrasound Standard Plane Recognition for Mid Pregnancy by a Specific ResNet20
		Xiong Runqing, Cai Jiaxin, Zheng Liping, Ma Duo, Dai Chenquan
		DOI: 10.3969/j.issn.0258-8021.2025.05.003
		To address the problems of experience-dependency, low efficiency and wrong diagnosis risks in standard section recognition during second-trimester fetal level II prenatal ultrasound screening, this study explored the application of convolutional neural network (CNN) technology for automatic recognition of multi-section ultrasound images, aiming to improve screening accuracy and real-time performance. Ultrasound images of second-trimester fetuses from 200 pregnant women were collected using GE E8 and Samsung WS80A color Doppler devices, covering 10 standard fetal sections (1 869 images in total). All images were normalized to 227 pixels×227 pixels. A specially designed ResNet20 model was constructed, featuring a seven-layer architecture with bottleneck residual modules, and the softmax function was used for classification probability output. During the training, 90% of the images were used as the training set, and 10% of them as the test set. The optimization employed a multi-class cross-entropy loss with a learning rate of 0.1 and a momentum optimizer. The ResNet20 model achieved an overall recognition accuracy of 88.2%, with 100% precisions in identifying the cerebellar transverse section, bilateral eye transverse section, and spinal sagittal section. In comparison, AlexNet and VGG16 showed accuracies of 80.2% and 73.3%, respectively, with statistically significant differences (P<0.001). The average testing time was 33 seconds per case, meeting clinical real-time requirements. The proposed ResNet20 model enabled efficient automatic recognition of second-trimester fetal level II standard sections, demonstrating good performance and clinical application prospects, contributes to the intelligent and standardized improvement of ultrasound screening. The code is open-source at https://github.com/Chenan7/Resnet-19.
		2025 Vol. 44 (5): 533-540 [Abstract] ( 39 ) HTML (1 KB) PDF (4391 KB) ( 24 )

	541	Multi-Ethnic Mammographic Diagnosis Analysis Based on Deep Learning
		Wang Yunfei, Fan Ming, Zhou Peng, Li Lihua
		DOI: 10.3969/j.issn.0258-8021.2025.05.004
		Breast cancer is a highly prevalent malignancy among women globally and exhibits significant clinical characteristic differences across ethnic groups. However, studies on the imaging features of breast cancer in different ethnic populations remain limited. This study included 366 breast cancer patients from multiple ethnic groups (153 Han, 134 Tibetan, and 79 Yi) and analyzed the clinical and imaging characteristics of each group, aiming to explore cross-ethnic differences and provide evidence for personalized diagnosis and treatment. Using clinical pathological data and mammography data, deep learning models based on ResNet networks of varying depths were developed for multi-ethnic classification and Ki-67 prediction. The results indicated that, compared to Han patients, Tibetan and Yi patients were diagnosed at younger ages (P<0.05) and exhibited significantly higher breast density. Tibetan patients also showed a higher proportion of HER2 positivity. In the multi-ethnic classification task, the ResNet34 model demonstrated the best classification performance, with a macro-average AUC of 0.880 on test data (Han: 0.911, Tibetan: 0.974, Yi: 0.742). Model visualization revealed that the model focused more on glandular regions for Tibetan patients, whereas it paid more attention to tumor regions for Han patients. In the multi-ethnic Ki-67 prediction task, the ethnicity-specific models (Han AUC=0.820, Tibetan AUC=0.842, Yi AUC=0.970) significantly outperformed the mixed-ethnic model (AUC=0.818), with the Yi model showing the most substantial improvement in predictive performance. In conclusion, significant differences existed in clinical and imaging features of the breast cancer patients across ethnic groups. The ethnicity-specific deep learning prediction models showed superior performance in assessing the Ki-67 proliferation index, offering valuable insights for the development of personalized diagnostic and treatment strategies for breast cancer patients from different ethnic backgrounds.
		2025 Vol. 44 (5): 541-550 [Abstract] ( 35 ) HTML (1 KB) PDF (1954 KB) ( 13 )

	551	Assessment of Pulmonary Artery Pressure and Pulmonary Capillary Wedge Pressure Based on Cardio-Electrical and-Mechanical Signals
		Zheng Quanzhong, Wang Xingyao, Li Jianqing, Liu Chengyu, Yang Chenxi
		DOI: 10.3969/j.issn.0258-8021.2025.05.005
		Monitoring patients′ hemodynamics and providing timely interventions can significantly enhance the treatment outcomes for heart failure. Recent research has demonstrated the potential of seismocardiogram (SCG) and electrocardiogram (ECG) in heart failure management. To enable more patients to monitor their hemodynamic changes outside the hospital for prompt treatment, this study attempted to utilize cardio-electrical and -mechanical signals to assess pulmonary artery pressure (PAP) and pulmonary capillary wedge pressure (PCWP). The study proposed a prediction method for PAP and PCWP based on multi-scale feature fusion within a multi-task learning framework. This method employed dilated convolutions of varying scales to extract features from individual heartbeat SCG and ECG signals at different scales for fusion and encoding. Subsequently, global information of the encoded data was embedded to adjust the weights of various components, and finally, the encoded signals were decoded to predict PAP and PCWP. Training and validation on data from 66 subjects showed that the MAE for predicting PAP and PCWP were 4.31 and 3.20 mmHg, respectively with corresponding R² scores of 0.736 and 0.771. The experimental results indicated that the proposed method achieved non-invasive prediction of PAP and PCWP, presenting broad prospects for hemodynamic monitoring out of the clinic.
		2025 Vol. 44 (5): 551-559 [Abstract] ( 32 ) HTML (1 KB) PDF (3839 KB) ( 22 )

	560	Multi-FHRNet: Multimodal Fetal Diagnosis Based on Image and Text with Adaptive Weighting Fusion
		Zhang Ling, Zhao Zhidong, Zhang Yiefei, Jiao Pengfei, Deng Yanjun, Zhang Xianfei
		DOI: 10.3969/j.issn.0258-8021.2025.05.006
		Distress or acidosis resulted from hypoxia can lead to irreversible consequences such as fetal organ damage even death. Intelligent cardiotocography (ICTG) is an important tool for monitoring fetal health by continuously and synchronously recording fetal heart rate (FHR) signals in late pregnancy. The existing FHR-based unimodal ICTG combined with machine learning algorithms for assisted diagnosis method neglects the complementarity between different modalities data of FHR and the influence of clinical physiological factors on the fetus. In this study, we proposed a multimodal learning model with adaptive weighted fusion of images and text. In particular, we designed a multimodal feature extraction network consisting of a Vision Transformer (ViT) image encoder and a convolutional neural networks (CNN) text encoder. Adaptive weighting fusion (AWF) was proposed to fuse multimodal features, instead of direct concatenates. Considering the impact of realistic risk factors, textual data can be constructed by combining the clinical data and the morphological features of FHR signals. Meanwhile, the Markov transfer field was employed to convert signals into images as complementary data. Using 200 sets of public clinical real-world FHR signals, multiple performance comparisons, parameter optimization, and ablation experiments were carried out. The experiment results indicated that the Multi-FHRNet outperformed traditional unimodal ICTG methods, withthe highest accuracy, precision, recall, and F1-score of 96.02%, 93.10%, 99.29%, 95.45% and 93.48%, respectively. The algorithm in this paper may help to detect and treat abnormal fetuses during labor.
		2025 Vol. 44 (5): 560-569 [Abstract] ( 31 ) HTML (1 KB) PDF (2374 KB) ( 28 )

	570	Research on Federated Learning-Based SSVEP Frequency Recognition Algorithm
		Li Jinfei, Chen Jianbo, Zhang Yangsong, Xu Peng
		DOI: 10.3969/j.issn.0258-8021.2025.05.007
		Steady-state visual evoked potential (SSVEP) is widely used in brain-computer interface (BCI) applications due to its high signal-to-noise ratio and stability. Currently, frequency recognition algorithms for SSVEP based on deep learning are facing the challenge of insufficient data. Effective training models using multi-center data has emerged as a potential solution. In this paper, a federated learning algorithm framework FLSSVEP for SSVEP recognition algorithm was proposed, which realized the joint training of models on different clients, and avoided the violation of data privacy while using more training data, so as to alleviate the impact of insufficient data or data dispersion on the model training. To verify the effectiveness of the proposed algorithm framework, comprehensive experiments were conducted on two public datasets of Benchmark and BETA by using EEGNet, CCNN, and SSVEP Former as the baseline models. Experimental results showed that the proposed FLSSVEP led to an average improvement of 33.82% in classification accuracy for these baseline models on one client, while the other two clients achieved average improvements of 8.13% and 6.05% in classification accuracy, respectively, better than those obtained from traditional local data training methods. This study demonstrated the effectiveness of federated learning in designing algorithms for SSVEP-based BCI, providing a theoretical reference for future research.
		2025 Vol. 44 (5): 570-578 [Abstract] ( 47 ) HTML (1 KB) PDF (2135 KB) ( 30 )

	579	Research on Detection of Interictal Epileptiform Discharges in Children Based on Deep Learning
		Rao Wenhao, Chen Duo, Zhang Ling, Jiang Jun
		DOI: 10.3969/j.issn.0258-8021.2025.05.008
		Interictal epileptiform discharges (IED) are crucial for epilepsy diagnosis, but the non-stationarity of EEG signals make IED detection complicated. Traditional manual interpretation of EEG is subjective and time-consuming. With the development of machine learning and deep learning, computer-assisted models have been proposed in the field of IED detection. CNN-based IED detection methods have achieved promising results, but CNN is less effective in capturing long-range dependencies within time series data. Transformer is good at processing sequential data by adopting a self-attention mechanism, which enables it to capture long-term dependencies. This study proposes a novel Transformer-based IED detection method, which first uses simple convolution to extract local features of IEDs and then employs a Transformer to further model the long-range dependencies of these features. To address the scarcity of IED data, a new Transformer-based generative adversarial network (GAN) is also designed to augment the IED data. Based on an analysis of 11 pediatric epilepsy patients, the new method achieved an average accuracy of 96.11%, average recall of 97.08%, and average precision of 93.85% in the binary classification task on the augmented dataset. In the multi-class classification task, the average recall reached 93.47%, and the average precision was 93.84%. This study provides valuable reference for the future application of deep learning in automatic IED detection.
		2025 Vol. 44 (5): 579-590 [Abstract] ( 35 ) HTML (1 KB) PDF (5500 KB) ( 25 )

Reviews

	591	The Application of Body Area Communication Technology in Programming/Telemetry of CardiacPacemakers
		Guo Binbin, Huang Yanqi, Yan Shengjie, Wu Xiaomei
		DOI: 10.3969/j.issn.0258-8021.2025.05.009
		Aimed to integrate key achievements of body area networks in programming/telemetry technology of implantable cardiac pacemakers,this article made a systematic review and multi-dimensional analysis approach. Firstly, we elaborated on the working principle of implantable cardiac pacemakers, particularly the programming/telemetry functions. Then, the development history of cardiac pacemaker programming/telemetry technology and sorts out the key technical threads in the development process were briefly reviewed. On this basis, this article conducted a comparative analysis of communication technologies including inductive coupling communication, current coupling communication, capacitive coupling communication, and microwave communication, to evaluate their core indicators such as transmission efficiency, biocompatibility, and electromagnetic safety in the programming/telemetry of cardiac pacemakers, and deeply explore the technical bottlenecks and potential solutions faced by each technology in clinical applications. Finally, a summary and comparative study of the common modulation and coding mechanisms of various communication standards were conducted from the perspective of information theory. In conclusion, this review was expected to provide a reference for the research on the programming/telemetry technology of implantable medical electronic devices represented by cardiac pacemakers.
		2025 Vol. 44 (5): 591-603 [Abstract] ( 32 ) HTML (1 KB) PDF (3451 KB) ( 14 )

	604	Research Advances and Challenges in Proton Flash Radiotherapy
		Chen Tianmei, Li Yanyan, Zhang Huojun
		DOI: 10.3969/j.issn.0258-8021.2025.05.010
		The objective of modern radiotherapy is to achieve an optimal balance between tumour control and the protection of normal tissues. Proton FLASH radiotherapy is a therapeutic method that outputs proton beams at an ultra-high dose rate. Its unique FLASH effect and proton properties enable it to efficiently kill the tumor while reducing radiation toxicity to the normal tissue. Despite the promising results observed in several studies, the clinical translation of proton FLASH-RT still faces several challenges and limitations. This review will introduce the biological mechanisms of FLASH-RT, provide a systematic analysis and discussion of the clinical advantages of proton FLASH-RT based on existing relevant studies, and summarize the difficulties and challenges in the clinical translation of FLASH-RT from four parts: proton acceleration equipment and delivery technology, dosimetric studies, dosimetry, and biological mechanisms, aiming to provide comprehesive information for the transition of proton FLASH-RT from theory to practice.
		2025 Vol. 44 (5): 604-612 [Abstract] ( 39 ) HTML (1 KB) PDF (849 KB) ( 24 )

	613	Current Status and Progress of Laboratory Automation in Clinical Microbiology
		Ge Yang, Cao Wei, Wu Yanfan, Feng Yongtong, Lin Wenqi, Liu Han, Meng Jiao, Xu Zhengping, Liu Yi
		DOI: 10.3969/j.issn.0258-8021.2025.05.011
		With the global rise in multi-drug resistant organism (MDRO) infections, rapid and accurate diagnosis is crucial for effective treatment. However, with testing volumes increasing annually, clinical microbiology laboratories face challenges like staff shortages and cost constraints, creating a strong demand for automation. Total laboratory automation (TLA) in microbiology has developed significantly, offering benefits such as standardized testing, improved efficiency, enhanced safety, and lower long-term costs. This article reviewed the development of TLA, covering sample processing, culture, imaging, identification, and various leading automation systems worldwide. Despite TLA′s advantages, challenges remain, including costs, software/hardware integration, and adapting to diverse workflows. Therefore, the implementation of clinical microbiology TLA should not pursue complete replacement of manual operation,instead, it is necessary to adopt a gradual and human-machine collaborative strategy. Overall, the application of TLA is promising, and with continued technological progress, its role in clinical microbiology will likely expand further.
		2025 Vol. 44 (5): 613-621 [Abstract] ( 30 ) HTML (1 KB) PDF (2294 KB) ( 24 )

	622	Research Progress on Melanoma Postoperative Hydrogel Therapy Platform
		Li Jinhua, Sun Honggang
		DOI: 10.3969/j.issn.0258-8021.2025.05.012
		Melanoma has remarkably gained extensive attention owing to its high morbidity and mortality. Surgical resection is the mainstay for melanoma therapy in clinic. Inhibiting the recurrence of melanoma and healing after-surgical wounds have been great challenges in clinical studies. Hydrogels is a class of three-dimensional network of biomaterials with unique porous structures, and hydrogel therapy platforms can offer solutions to these challenges. In this review, we summarized recent developments of polymer-based hydrogels as therapy platforms for local delivery of therapeutic agents for postoperative melanoma therapy. According to the mechanisms that induce cancer cell death in melanoma, we focused on the latest research achievements of hydrogel chemotherapy, photodynamic therapy (PDT), photothermal therapy (PTT), immunotherapy, and combination therapies. Furthermore, the merits of postoperative therapy platform of the hydrogels were summarized and existing challenges and further perspectives of this platform for melanoma postoperative therapy were discussed.
		2025 Vol. 44 (5): 622-630 [Abstract] ( 40 ) HTML (1 KB) PDF (2413 KB) ( 13 )

Communications

	631	SignBrain Viewer: a High Performance EEG Analysis Software
		Chang Dongming, Meng Qingtong, Hu Ruochen, Jia Ziyu, Jiang Tianzi, Zuo Nianming
		DOI: 10.3969/j.issn.0258-8021.2025.05.013

		2025 Vol. 44 (5): 631-635 [Abstract] ( 29 ) HTML (1 KB) PDF (3448 KB) ( 21 )

	636	A Novel Method for Cell Distribution Using an Integrated Superhydrophobic Microwell Array Chip
		Jiao Pengcheng, Zhao Tian, Ji Jiaojiao, Yu Bin, Wu Rong, Liu Peng, Feng Qianqian
		DOI: 10.3969/j.issn.0258-8021.2025.05.014

		2025 Vol. 44 (5): 636-640 [Abstract] ( 32 ) HTML (1 KB) PDF (2869 KB) ( 13 )

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