生物医学工程学杂志最新文献

Cone-beam computed tomography (CBCT) is widely used in dentistry, surgery, radiotherapy and other medical fields. However, repeated CBCT scans expose patients to additional radiation doses, increasing the risk of secondary malignant tumors. Low-dose CBCT image reconstruction technology, which employs advanced algorithms to reduce radiation dose while enhancing image quality, has emerged as a focal point of recent research. This review systematically examined deep learning-based methods for low-dose CBCT reconstruction. It compared different network architectures in terms of noise reduction, artifact removal, detail preservation, and computational efficiency, covering three approaches: image-domain, projection-domain, and dual-domain techniques. The review also explored how emerging technologies like multimodal fusion and self-supervised learning could enhance these methods. By summarizing the strengths and weaknesses of current approaches, this work provides insights to optimize low-dose CBCT algorithms and support their clinical adoption.

The orientation of the acetabular cup in hip joint anteroposterior radiograph is a key factor in evaluating the postoperative outcomes of total hip arthroplasty (THA). Currently, measurement of the acetabular cup anteversion angle primarily relies on manual drawing of auxiliary lines by orthopedic surgeons and calculations using scientific calculators. This study proposes an automated computer-aided measurement method for the acetabular cup anteversion angle based on hip joint anteroposterior radiograph. The proposed method segments hip prosthesis images using an improved Otsu algorithm, identifies feature points at the acetabular cup opening by combining circle-fitting theory and the cup's geometric characteristics, and fits an ellipse to the cup opening to calculate the anteversion angle. A total of 104 hip joint anteroposterior radiographs, including 71 right-sided and 81 left-sided prostheses, were analyzed. Two orthopedic surgeons independently measured the postoperative anteversion angles, and the results were compared with computer-generated measurements for correlation analysis. Spearman and Pearson correlation analyses demonstrated significant correlations between the proposed method and manual measurements for both the right group ( r = 0.795, P < 0.01) and the left group ( r = 0.859, P < 0.01). This method provides a reliable reference for orthopedic surgeons to assess postoperative prognosis.

Stroke causes abnormality of brain physiological function and limb motor function. Brain-computer interface (BCI) connects the patient's active consciousness to an external device, so as to enhance limb motor function. Previous studies have preliminarily confirmed the efficacy of BCI rehabilitation training in improving upper limb motor function after stroke, but the brain mechanism behind it is still unclear. This paper aims to review on the brain mechanism of upper limb motor dysfunction in stroke patients and the improvement of brain function in those receiving BCI training, aiming to further explore the brain mechanism of BCI in promoting the rehabilitation of upper limb motor function after stroke. The results of this study show that in the fields of imaging and electrophysiology, abnormal activity and connectivity have been found in stroke patients. And BCI training for stroke patients can improve their upper limb motor function by increasing the activity and connectivity of one hemisphere of the brain and restoring the balance between the bilateral hemispheres of the brain. This article summarizes the brain mechanism of BCI in promoting the rehabilitation of upper limb motor function in stroke in both imaging and electrophysiology, and provides a reference for the clinical application and scientific research of BCI in stroke rehabilitation in the future.

Speech imagery is an emerging brain-computer interface (BCI) paradigm with potential to provide effective communication for individuals with speech impairments. This study designed a Chinese speech imagery paradigm using three clinically relevant words-"Help me", "Sit up" and "Turn over"-and collected electroencephalography (EEG) data from 15 healthy subjects. Based on the data, a Channel Attention Multi-Scale Convolutional Neural Network (CAM-Net) decoding algorithm was proposed, which combined multi-scale temporal convolutions with asymmetric spatial convolutions to extract multidimensional EEG features, and incorporated a channel attention mechanism along with a bidirectional long short-term memory network to perform channel weighting and capture temporal dependencies. Experimental results showed that CAM-Net achieved a classification accuracy of 48.54% in the three-class task, outperforming baseline models such as EEGNet and Deep ConvNet, and reached a highest accuracy of 64.17% in the binary classification between "Sit up" and "Turn over". This work provides a promising approach for future Chinese speech imagery BCI research and applications.

This study aims to explore whether Guzheng playing training has a positive impact on the brain functional state of children with Autism Spectrum Disorder (ASD) based on power spectral and sample entropy analyses. Eight ASD participants were selected to undergo four months of Guzheng playing training, with one month as a training cycle. Electroencephalogram (EEG) signals and behavioral data were collected for comparative analysis. The results showed that after Guzheng playing training, the relative power of the alpha band in the occipital lobe of ASD children increased, and the relative power of the theta band in the parietal lobe decreased. The differences compared with typically developing (TD) children were narrowed. Moreover, some channels exhibited a gradual increase or decrease in power with the extended training period. Meanwhile, the sample entropy parameter also showed a similar upward trend, which was consistent with the behavioral data representation. The study shows that Guzheng training can enhance the brain function of ASD patients, with better effects from longer training. Guzheng playing training could be used as a daily intervention for autism.

In recent years, the research on artificial intelligence medical devices has risen markedly along with the expanding application scenarios, exhibiting prominent interdisciplinary characteristics. From 2000 to 2024, the variety of research in artificial intelligence medical devices has significantly increased, while the balance of disciplines has slightly declined, and Simpson's diversity index has continuously increased. Medicine and biology are the main research themes and supportive disciplines in this field. Knowledge from computer science, engineering technology, and mathematics is widely involved and shows an upward trend, while content from the humanities and social sciences is less involved in the research. Compared to the United States and the United Kingdom, China has relatively less biological and chemical knowledge content in the research of this field, but more content related to computer science, engineering technology and material science is involved. This study analyzes the current state and trends of interdisciplinary on artificial intelligence medical devices from the perspective of macro-categories of disciplines, aiming to provide references for research planning, talent training and interdisciplinary cooperation in the field.

In recent years, it has become a new direction in the field of neuroscience to explore the mode characteristics, functional significance and interaction mechanism of resting spontaneous electroencephalography (EEG) and task-evoked EEG. This paper introduced the basic characteristics of spontaneous EEG and task-evoked EEG, and summarized the core role of spontaneous EEG in shaping the adaptability of the nervous system. It focused on how the spontaneous EEG interacted with the task-evoked EEG in the process of task processing, and emphasized that the spontaneous EEG could significantly affect the performance of tasks such as perception, cognition and movement by regulating neural activities and predicting external stimuli. These studies provide an important theoretical basis for in-depth understanding of the principle and mechanism of brain information processing in resting and task states, and point out the direction for further exploring the complex relationship between them in the future.

Post-stroke spasticity, a common sequelae of upper motor neuron lesions, results in motor control deficits and pathological hypertonia that not only reduce patients' activities of daily living but may also cause impairment of adaptive neuroplasticity. Repetitive peripheral magnetic stimulation (rPMS), as a novel non-invasive neuromodulation technique, demonstrates unique clinical potential through targeted modulation of electromagnetic coupling effects in the peripheral neuromuscular system. Although current international studies have validated the therapeutic potential of rPMS for spasticity, significant heterogeneity persists in elucidating its mechanisms of action, optimizing parameter protocols, and standardizing outcome assessment systems. This review innovatively synthesized recent randomized controlled trials (RCTs) and mechanistic evidence, systematically summarizing rPMS-mediated multidimensional intervention paradigms for upper- and lower-limb spasticity. It rigorously examined the correlations between stimulation frequency parameters (low-frequency vs. high-frequency), anatomical targeting (nerve trunk vs. motor point), and clinical outcomes including spasticity severity, motor function, and quality of life. Crucially, the analysis reveals that rPMS may ameliorate spasticity after stroke through dual mechanisms involving local neuroelectrophysiological modulation and central sensorimotor network reorganization, thereby providing a theoretical foundation for developing individualized rPMS clinical protocols and establishing precision treatment strategies.

To address the high computational complexity of the Transformer in the segmentation of ultrasound thyroid nodules and the loss of image details or omission of key spatial information caused by traditional image sampling techniques when dealing with high-resolution, complex texture or uneven density two-dimensional ultrasound images, this paper proposes a thyroid nodule segmentation method that integrates the receiving weighted key-value (RWKV) architecture and spherical geometry feature (SGF) sampling technology. This method effectively captures the details of adjacent regions through two-dimensional offset prediction and pixel-level sampling position adjustment, achieving precise segmentation. Additionally, this study introduces a patch attention module (PAM) to optimize the decoder feature map using a regional cross-attention mechanism, enabling it to focus more precisely on the high-resolution features of the encoder. Experiments on the thyroid nodule segmentation dataset (TN3K) and the digital database for thyroid images (DDTI) show that the proposed method achieves dice similarity coefficients (DSC) of 87.24% and 80.79% respectively, outperforming existing models while maintaining a lower computational complexity. This approach may provide an efficient solution for the precise segmentation of thyroid nodules.

To address the issues of difficulty in preserving anatomical structures, low realism of generated images, and loss of high-frequency image information in medical image cross-modal translation, this paper proposes a medical image cross-modal translation method based on diffusion generative adversarial networks. First, an unsupervised translation module is used to convert magnetic resonance imaging (MRI) into pseudo-computed tomography (CT) images. Subsequently, a nonlinear frequency decomposition module is used to extract high-frequency CT images. Finally, the pseudo-CT image is input into the forward process, while the high-frequency CT image as a conditional input is used to guide the reverse process to generate the final CT image. The proposed model is evaluated on the SynthRAD2023 dataset, which is used for CT image generation for radiotherapy planning. The generated brain CT images achieve a Fréchet Inception Distance (FID) score of 33.159 7, a structure similarity index measure (SSIM) of 89.84%, a peak signal-to-noise ratio (PSNR) of 35.596 5 dB, and a mean squared error (MSE) of 17.873 9. The generated pelvic CT images yield an FID score of 33.951 6, a structural similarity index of 91.30%, a PSNR of 34.870 7 dB, and an MSE of 17.465 8. Experimental results show that the proposed model generates highly realistic CT images while preserving anatomical accuracy as much as possible. The transformed CT images can be effectively used in radiotherapy planning, further enhancing diagnostic efficiency.