Abstract:To improve the classification accuracy of benign and malignant pulmonary nodules,we proposed an improved residual network model LNC-Net.Firstly,small convolution kernels were cascaded to instead of large convolution kernels to improve computa-tional efficiency.Secondly,the input backbone and residual module of ResNet50 were reconstructed,and the output part of ResNet50 was replaced by the global average pooling layer to enhance the feature learning ability of the network and reduce the model parameters.Finally,the feature extraction capability of the network was enhanced by fusing the residual channel attention module(RCAM)and the receptive field block-small(RFB-s).Experiments on LIDC dataset showed that the accuracy rate,precision rate,recall rate,F1 score and AUC of the model reached 0.983,0.984,0.987,0.985 and 0.999,respectively.This model can effectively achieve the auxilia-ry diagnosis of benign and malignant pulmonary nodules.

Abstract:To enhance blood pressure measurement accuracy,we proposed a Korotkoff sound phase classification model based on deep learning,and designed a blood pressure measurement method based on Korotkoff sound phase classification.Firstly,369 pieces of Korotkoff sound data from 102 healthy volunteers were collected,and manual auscultation was used to label different time phases.Sec-ondly,the log-mel spectrogram and Hilbert envelope features of the Korotkoff sound signal were extracted,and combined with ResNet18,convolutional block attention module(CBAM),bidirectional long short-term memory network(BiLSTM)and multi-head self-attention module,the features of the Korotkoff sound signal were fully learned.Finally,on the basis of the Korotkoff sound phase classi-fication,blood pressure measurement was completed.The experimental results showed that the average classification accuracy of the Korotkoff sound phase reached 88.9%.The blood pressure measurement method in the research met the A-level standard set by the British Society of Hypertension(BHS)under the four blood pressure measurement standards,and the intraclass correlation coefficient(ICC)was greater than 0.95,providing reference significance for the research of automatic blood pressure measurement methods.

Abstract:To enhance the standardization of traditional Chinese medicine(TCM)acupuncture treatment by applying modern intel-ligent technologies,we developed a six-degree-of-freedom robotic arm system for acupoint positioning and manipulation.Firtly,the system combined the meridian theory of TCM,and realized acupoint recognition through RealSense depth camera and YOLOv8n-pose model.Secondly,the acupuncture manipulation simulation device was designed,the stepping motor and the lead screw slides were used to master the lifting-thrusting and twisting manipulation.Finally,the six-DOF robotic arm of myCobot 280 M5 was used,and the map-ping relationship between the coordinate system of the camera and that of the robotic arm through the nine-point calibration was estab-lished to realize the automated point searching.The accuracy test results showed that the average error of positioning was within 3 mm,the average deviation of lifting-thrusting amplitude was 2 mm,the twisting angle was negligible,and the average errors of the left and right arm point searching was 8.64 mm and 9.28 mm,respectively.In the stability test of the system,the overall success rate of the sys-tem reached 98.70%in 4 h of continuous operation.The system positioning accuracy,manipulation simulation accuracy,and system stability are good,can provide solutions for the modernization and intelligence of TCM acupuncture.

Abstract:In order to improve the antibacterial and hemostatic properties of gelatin nasal hemostatic sponge,we cross-linked and modified gelatin using squid chitosan,then compounded it with polyvinyl alcohol,and finally prepared the modified gelatin nasal hemo-static sponge through freeze-drying.The effect of the hemostatic sponge was explored through tests on its physical and chemical proper-ties,antibacterial properties,hemostatic performance and cytotoxicity.The results showed that the breaking force of the modified gelatin nasal hemostatic sponge was 14.78 N,the expansibility was 3.27 times,the water absorption ratio was 36.5 times,and the average an-tibacterial rates against Staphylococcus aureus and Escherichia coli reached 100%and 98.80%,respectively.The cell survival rate ex-ceeds 90%and the effective hemostasis time is 14 s.The modified gelatin nasal hemostatic sponge prepared exhibits rapid hemostasis,good antibacterial performance,no cytotoxicity and broad clinical application value.

Abstract:Chest medical imaging techniques such as X-rays,computed tomography(CT),and magnetic resonance imaging(MRI)are widely used for disease screening and diagnosis.The development of artificial intelligence(AI)technology enables the con-version of images of different modalities,thereby filling the data gap in medical imaging diagnosis and treatment,and improving the ef-ficiency and accuracy of medical services.The paper analyzes the research progress of cross-modal generation technology in improving the diagnosis and treatment of chest diseases,and elaborates on the application of cross-modal generation technology for chest imaging in medical imaging diagnosis and treatment,to provide guidance for subsequent research.

Abstract:To address the bias issues caused by uneven data distribution in large vision-language models(LVLMs)for medical auxiliary diagnosis,we proposed a universal medical debiasing method called bias fairness enhancement(BFE).The effectiveness of BFE in mitigating bias was validated by constructing a benchmark to evaluate bias issues in medical LVLMs.This benchmark coverd bi-nary classification,multi-classification,and open-ended question tasks,while incorporating two parameters that control output ran-domness to ensure the method's generality.Experimental results demonstrated that BFE outperformed other mainstream debiasing meth-ods in the medical LVLMs(LLaVA-Med,SkinGPT).Notably,in open-ended medical question-answering tasks,LLaVA-Med's per-formance improved by 6.3%(cd_beta=0.1)and 6.7%(cd_beta=0.5).These findings indicate that BFE can effectively alleviate bias in medical LVLMs,and provide significant support for improving the accuracy and reliability of medical auxiliary diagnosis.

Abstract:In order to solve the problems of low specificity of current pre-admission stroke detection and inconvenient placement of large equipment,we combined the microwave propagation characteristics with microwave imaging,designed an acute stroke detection device.Firstly,the finite element simulation method was used to simulate the microwave detection effect and antenna characteristics.Then,the microwave detection circuit was designed and a stroke detection device with STM32 as the core processor was constructed.The simulation result showed that when the device operated within the frequency range of 0.8～6 GHz,the maximum measurement gain exceeded 10 dB.The validation results in the skull model indicated that the positioning error of the device for the bleeding center point was 14.8 mm,and the positioning error for the ischemic center point was 16.2 mm.The device can rapidly distinguish between ischemic and hemorrhagic stroke before hospitaladmission,which is of great significance for precise treatment and postoperative rehabilitation of stroke patients.

Abstract:Aiming at the problems of difficult data acquisition and high costs with manual annotation in the actual assessment of electroencephalogram(EEG)signal quality,we proposed an EEG signal quality assessment method based on data augmentation and transfer learning.Firstly,the autoregressive model was used to fit the real and pure EEG signals.Secondly,by adding different levels of simulated artifacts to the EEG signals,a multi-quality distribution of simulated EEG was formed to construct the source domain dataset.Finally,multi-dimensional features were extracted and the support vector machine(SVM)was trained on the source domain model and feature alignment was achieved through the association alignment transfer learning method.Experimental results showed that the accura-cy,macro average precision,macro-average recall and macro-average F1-score of the transfer learning-enhanced SVM achieved 84.00%,81.06%,85.76%and 82.85%,respectively,significantly outperforming the baseline approach without transfer learning.The research combines data augmentation with transfer learning for EEG quality evaluation,can provide a new low-cost cross-scenario method for EEG signal quality assessment.

Abstract:To explore the relationship between muscle synergy patterns and firefighters'injury risk,we constructed a risk assess-ment model for injuries,analyzed the muscle patterns of the lower limbs of firefighters during jumping,and evaluated the risk of injury.Firstly,10 firefighters were recruited and divided into injured and non-injured groups.The injury status of the firefighters was then re-corded one year after the experiment.Subsequently,muscle synergy characteristics were analyzed via non-negative matrix factorization and least squares to develop the injury risk assessment model.The results showed that 4 and 3 synergy patterns in the injured group and non-injured group respectively,which proved the phenomenon of pattern decomposition.Differences in force-generating characteristics between the two groups were identified based on the temporal feature H vector.The risk assessment model(R2＝0.6899,P＝0.0006)indicated that firefighters with fewer muscle synergy patterns had a lower risk of injury.The result indicates that enhancing muscle strength can reduce the risk of training injuries.

Abstract:To examine the sleep-modulatory effects of transcutaneous electrical nerve stimulation(TENS)based on neural nuclei resonance,30 healthy participants were assigned to either a sham stimulation group or a TENS(123~127 Hz)group,each group was intervened 30 min.Sleep electroencephalogram(EEG)was recorded for 60 min and evaluated using both objective and subjective measures.The results showed that among participants with poor sleep quality,the TENS group exhibited bidirectional emotional im-provements,with increased positive emotions(P＝0.048)and decreased negative emotions(P＝0.002).TENS decreased the gravity frequency of EEG(P＝ 0.009),and reduced neuronal discharge frequency with stable dominance of δ and θ activity during sleep.Mo-reover,sleep efficiency increased,the proportion of N3 sleep was enhanced,the stability of N2 and N3 stages improved,and abnormal arousals were reduced.The research demonstrate that TENS with specific frequency waveforms effectively promotes low-frequency EEG activity,optimizes sleep architecture and improves emotional state,offering a novel non-pharmacological intervention for insomnia.

Abstract:To address the issues of blurred boundaries and low detection accuracy for subtle fractures in Child's wrist X-ray ima-ges,as well as the high computational cost faced by some computer aided diagnosis(CAD)methods,we proposed a lightweight frac-ture detection algorithm based on improved YOLOv12.Firstly,a synergistic multi-scale backbone(SMSB)was constructed to enhance the collaborative extraction of shallow spatial details and deep semantic information.Secondly,a contextual detail alignment(CDA)module was designed to efficiently fuse multi-scale features.Finally,a lightweight bounding box quality prediction head(LBBQP-Head)was proposed to mitigate the mismatch between classification confidence and localization accuracy.Experimental results demon-strated that the proposed method achieved a mAP50 of 64.98%on the GRAZPEDWRI-DX dataset,outperformed the baseline and main-stream models including YOLOv8 and YOLOv11,while reducing model parameters by 73.41%.Furthermore,experiments on the HBFMID dataset validated the generalization capability of the proposed algorithm.This study can provide a high-precision and efficient technical solution for intelligent computer aided fracture diagnosis,specifically tailored for primary healthcare institutions.

Abstract:Addressing the critical limitation of traditional computational fluid dynamics(CFD)simulations in evaluating airway ste-nosis—specifically,the neglect of the alveolar pressure feedback mechanism.We developed a multi-scale numerical model based on dynamic alveolar pressure feedback.Utilizing a 0D-3D coupling strategy,the model integrated three-dimensional upper airway geome-try with lumped parameter models representing the peripheral lung units,successfully replicating the physiological feedback inherent in spontaneous breathing.The simulation results accurately captured core pathophysiological phenomena,including"pendelluft"(22.2 mL reflux per cycle)and the dynamic accumulation of intrinsic positive end-expiratory pressure(PEEPi,steady-state value:6.97 Pa),alongside a marked reduction in breathing efficiency.Furthermore,we mechanistically revealed that pressure imbalance,driven by regional heterogeneity in time constants,was the key factor deteriorating respiratory function.This research provides an innovative numerical simulation framework for the functional assessment of airway stenosis and personalized treatment planning.

Abstract:The ankle-foot orthosis(AFO)achieves ankle joint movement control through mechanical support and biomechanical design,and is widely used in the rehabilitation of neurological muscle and orthopedic diseases.In recent years,the advancements in sensors,control strategies,data-driven technologies and additive manufacturing have provided support for the functionalization and lightweight design of AFO.This article systematically reviews the research progress of AFO in aspects such as sensors,control strate-gies,driving technologies,as well as manufacturing material innovations,and analyzes the application scenarios including neurological rehabilitation,orthopedic rehabilitation and elderly assistive walking.Finally,the application prospects of AFO technology in motion in-tention recognition and prediction,deep learning-driven personalized intervention and multi-driver efficiency optimization are outlined to provide a reference basis for the miniaturization,personalization and multi-scenario application of AFO.

Abstract:To address the critical issue of internal coagulation within the dialyzer that affects the safety and efficiency of hemodialysis,we designed a helical flow guide vane integrated into the arterial end of the dialyzer.Firstly,based on the computational fluid dynamics(CFD)method,the coefficient of variation(CV)of blood flow at the end face of the fiber bundle and the wall shear stress(WSS)were used as the key hemodynamic indicators for predicting the risk of coagulation to conduct fluid dynamics simulation on the helical flow guide vanes.Sec-ondly,the optimal configuration of the helical flow guide vanes was determined through multi-objective optimization design,and 3D printing processing was carried out.Finally,in vitro bovine blood circulation experiment was conducted.The rate of increase in transmembrane pres-sure(TMP)was used as a macroscopic indicator to characterize the overall coagulation process.The anticoagulation performance of the exper-imental group and the control group was tested.The simulation results showed that the optimized helical flow guide vanes could induce helical flow in the fiber bundle area.The CV of the optimized model was reduced by 59.00%,and the proportion of the low WSS area at the end face of the fiber bundle was reduced by 69.24%,indicating the elimination of the low-speed stagnation zone in the end cap and the reduction of the initial risk of coagulation.In vitro anticoagulation experiments also indicated that the rate of TMP increase in the experimental group was 9.37%,slower than that in the control group,proving that the coagulation problem was improved.The structure of the helical flow guide vanes can delay the coagulation process within the dialyzer by improving the uniformity of blood flow and WSS of the dialyzer.The research can pro-vide new ideas and theoretical basis for the development of high-performance hemodialyzers with low anticoagulant dependence.

Abstract:To address the challenges of precise segmentation of the liver,tumors and vessels in preoperative planning for primary liver cancer ablation,we proposed a point cloud guided segmentation network(PCG-Net).Firstly,by using HU value thresholding and gradient computation,the three-dimensional CT image was transformed into a structured point cloud.Then,an improved RandLA-Net was utilized to encode the topological information of the point cloud,from which a feature map rich in cross-slice contextual infor-mation was generated through differentiable Gaussian splatting(GS).Finally,this feature map was subsequently fused with the original CT slices and fed into a U-Net integrated with a convolution block attention module(CBAM)to accomplish the segmentation task.On the 3D-IRCADb-01 dataset,PCG-Net achieved Dice similarity coefficient(DSC)of 0.934,0.618 and 0.725,and 95%Hausdorff distances(HD95)of 1.618,9.019 and 5.251 for the liver,tumors and vessels,respectively,superior overall performance against com-parative methods.The experimental results indicated that PCG-Net could effectively preserve the topological integrity of vascular seg-mentation and improve the accuracy and robustness of multi-target segmentation.The research can provide reliable technical support for precise surgical planning.

Abstract:Parkinson's disease(PD)is a complex neurodegenerative disorder.The lack sufficient specificity of available clinical indicators for early diagnosis,and the limited pharmacological treatment options,creating a pressing need for reliable biomarkers for early diagnosis and specific therapeutic targets.Studies have revealed significantly different microribonucleic acid(miRNA)expression profiles in the bodily fluids of PD patients compared to healthy individuals,suggesting their potential diagnostic value in distinguishing PD.MiRNA participats in the pathogenesis of PD by regulating genes involved in oxidative stress,neuroinflammation,mitochondrial homeostasis and apoptosis and shows promise as diagnostic markers and therapeutic targets for PD.This review summarizes recent ad-vances in the use of miRNA for the early diagnosis and treatment in PD.By discussing the mechanisms of action and reviewing strate-gies for brain delivery,we aim to offer new perspectives for the clinical diagnosis and treatment of PD.

Abstract:To address the complex microenvironment of diabetic wounds,a multi-enzyme cascade system composed of glucose oxi-dase and horseradish peroxidase(GOX-HRP)was incorporated into hyaluronic acid-modified silk fibroin(SF)hydrogel to prepare hy-drogel dressings capable of promoting wound healing in diabetes.Through examinations of the mechanical properties,catalytic perform-ance and cytotoxicity of the hydrogels,as well as animal experiments,the therapeutic efficacy of this hydrogel on diabetic wounds was investigated.The results demonstrated that the adhesion strength of the hydrogel tissue reached 29.9 kPa,which was 4.75 times higher than the unmodified hydrogels,and the elastic modulus was 14.8 kPa.In terms of effectively generating oxygen,reducing glucose levels and scavenging reactive oxygen species,the duration of action all reached 24 h(those performance absent in the enzyme-free hydro-gels).The cell survival rate reached 95%.The wound healing rate of mice reached 81%within two weeks,which was 12%higher than that of the control group.The multi-enzyme cascade SF hydrogel can effectively facilitate the healing of diabetic wounds,exhibits no cytotoxicity and broad prospects for clinical applications.

Abstract:Hydrogen sulfide(H2 S)as an important gas signaling molecule,is widely involved in various physiological and patho-logical processes of living organisms.Surface-enhanced Raman spectroscopy(SERS)technology with its advantages of high sensitivity,high specificity and resistance to photobleaching,has shown great potential in the field of H2 S detection.This paper systematically re-views the research progress of H2 S detection based on SERS technology in the past years,with a focus on the innovative research in di-rect detection strategies,molecular-mediated indirect detection strategies,SERS combined with other technologies,and the construc-tion methods of SERS active substrates.Finally,the challenges currently faced in the field are summarized and the future development directions are prospected.

Abstract:To address the problem that existing deep learning models hard to fully utilize the complementary information between the slow-changing hemodynamic trends and fast-changing morphological features in physiological signals,we proposed a hypotension prediction network based on trend-residual decomposition and cross-component attention.Firstly,the trend-residual decomposition strategy was employed to explicitly decompose the continuous physiological signals into low-frequency trend components and high-fre-quency residual components,which correspond to the macroscopic evolution of blood pressure and the microscopic pulsation patterns,respectively.Then,a cross-component attention mechanism was designed to construct bidirectional interactions between the two sets of features,capture long-term trends while sensitively detecting early compensatory signs hidden within the high-frequency waveforms.Experiments on the VitalDB dataset demonstrated that the area under the receiver operating characteristic(AUROC)curve was 89.81%,81.30%and 80.59%for 5,10 and 15 min prediction windows,respectively,significantly outperforming the traditional meth-ods.This model can provide doctors with continuous and accurate intraoperative risk quantitative assessment.

Abstract:To explore the physiological mechanism of the electrophysiological model of pulsed radiofrequency(PRF)and neuronal coupling in pain treatment,based on the Hodgkin-Huxley(HH)model,the electrophysiological models of healthy and pathological states(Na1.7 and Na1.8 channel conductance increased by 125%and 115%,respectively,and K channel conductance decreased by 85%)were constructed.At the same time,time-varying electric field stimulation combined with PRF(frequency 2 Hz,pulse width 20 ms,voltage 45 V and time 120 s)were coupled.The results showed that before treatment,the pathological model discharged 6 times every 500 ms without stimulation,and the discharge was persistent and irregular.When the stimulus was increased to 0.02 and 0.05 μA/cm2,discharges of 15 and 28 times per 500 ms occured,respectively,and the firing frequency increased with the increase of stimu-lus intensity.After treatment,there was no spontaneous discharge when no stimulation,and the resting membrane potential was-65 mV.However,when the external stimulation current density was 0.02 μA/cm2,discharges of 10 times per 500 ms were all close to the healthy model.This study can provide a key theoretical tool for the optimization of PRF treatment parameters and the study of pain treatment regimens.