Objective To apply the image recognition method based on distance graph convolutional network to the image processing of malaria-infected erythrocytes for realizing the multi-stage recognition of malaria and improving the diagnostic efficiency of malaria.Methods A multi-stage malaria recognition model based on distance graph convolutional network was proposed.A radial basis function was firstly added in KNN graph construction algorithm to construct adjacency matrix and assign weights to the nearest-neighbor nodes according to the similarity between nodes,so as to weaken the effects of the distant nearest-neighbor nodes on the central node.Then,attention mechanism was introduced to update adjacency matrix dynamically in the graph convolutional network for making the model pay attention to near-neighbor nodes with higher similarity,and finally the multi-stage image recognition of malaria-infected erythrocytes was completed.Results Validated on the Malaria-MIT dataset,the experimental results show that compared with original model,the proposed method improved accuracy,precision,recall rate and F1-score to 96.18%,96.23%,96.18%and 96.18%,respectively.Conclusion The proposed approach can effectively accomplish the task of multi-stage image recognition of malaria-infected erythrocytes.

Objective To analyze the dosimetric differences between Monaco Monte Carlo(MC)algorithm and Pinnacle collapse cone convolution(CCC)algorithm for the same machine model using the 6 MV flatten-filter free(FFF)mode,thus providing a reference for the clinical application of these two treatment planning systems.Methods According to the MPPG5 and TRS430 reports,the acceptance and commissioning of Monaco MC algorithm model and Pinnacle CCC algorithm model in Department of Radiation Oncology,Shenzhen Hospital,Cancer Hospital of Chinese Academy of Medical Sciences were performed.A retrospectively analysis was conducted on 30 cases,including 10 cases of nasopharyngeal cancer,6 cases of lung cancer,4 cases of esophageal cancer,and 10 cases of cervical cancer.For each case,a 6 MV FFF plan was designed in Pinnacle using CCC algorithm.A 2.5 mm dose grid was selected for plan optimization and dose calculation.The plan was then exported to Monaco,where dose to medium was calculated using MC algorithm and a 2.5 mm dose grid with a 1%uncertainty for each control point.Both calculations in Pinnacle and Monaco took into account the impact of the treatment couch and immobilization devices on dose attenuation.The dose differences to the target volumes and major organs at risk between Pinnacle and Monaco for 3 different body parts including the head,chest and abdomen were compared.Results(1)For nasopharyngeal cancer,compared with Pinnacle CCC algorithm,Monaco MC algorithm lowered the Vpd of PTVp,PTVn,PTVrpn,PTV1 and PTV2 by 9.98%,2.64%,15.0%,1.93%and 8.01%,elevated the Dmax of PTVp,PTVn and PTVrpn by 2.98%,5.62%and 2.39%,increased the Dmean of PTVn and PTV1 by 1.87%and 0.72%,respectively;and the Dmax of the brainstem,the Dmax of the optic chiasm,and the Dmean of the right parotid gland were 3.83%lower,7.03%higher and 1.32%higher in Monaco MC algorithm as compared with Pinnacle CCC algorithm,respectively.(2)For lung cancer and esophageal cancer,Monaco MC algorithm showed increases of 2.37%,4.18%,15.30%,6.36%and 1.04%in the Dmax of PGTV,the V20,V5 and Dmean of lungs,and the V30 of heart as compared with Pinnacle CCC algorithm,respectively.(3)For cervical cancer,the Vpd of PTV_LR,Dmax of PTV_LR,the V40 of the rectum,the Dmax of the small intestine,and the Dmean of humeral head were 7.70%lower,3.70%higher,4.31%lower,3.05%higher and 2.07%higher in Monaco MC algorithm than in Pinnacle CCC algorithm,respectively.These differences were statistically significant(P<0.05).Conclusion For the above 3 treatment sites,significant differences are found in dose calculations for target areas and organs at risk between Monaco MC algorithm and Pinnacle CCC algorithm for the same treatment plan.Attention should be paid to the differences in dose algorithms for different treatment planning systems in clinical applications,which may have impacts on patient survival rates and organ toxicity.

Objective To explore aging effect on healthy cerebral white matter using fixel-based analysis(FBA)on diffusion tensor imaging.Methods Eighty-seven healthy participants were divided into younger group and older group according to their ages,and all of whom were examined with diffusion tensor imaging at 3.0T.The fractional anisotropy(FA)of two groups was compared using whole brain voxel-based analysis(VBA).Then,FBA was used to calculate complexity(CX),fiber density(FD),fiber cross-section(FC)and fiber density and fiber cross-section(FDC)in voxels where older group exhibited significantly lower FA values than younger group.Finally,certain tracts of interest were chosen for tract-specific analyses.Results Compared with younger group,older group had significantly lower FA values in the genu of corpus callosum,splenium of corpus callosum,fornix,and posterior limb of internal capsule(P<0.05).Besides the above regions,older group showed significantly lower FD,FC and FDC in the anterior thalamic radiation,anterior limb of internal capsule,cerebral peduncle,superior longitudinal fasciculus,inferior longitudinal fasciculus,cingulum gyrus,forceps-minor and uncinate fasciculus(P<0.05).In voxels where FA values of older group were significantly lower than those of younger group,strong negative correlation was observed between average CX and average FA,while positive correlation was observed between average FD/FDC and average FA.Quantitative tract-specific analyses showed that older group had showed lower FD,FC and FDC in regard of superior longitudinal fasciculus,inferior longitudinal fasciculus,cingulum,anterior thalamic radiation,forceps-minor and uncinate fasciculus(P<0.05).Conclusion FBA reveals the characteristics of specific fibers and can be used to perform comprehensive and effective evaluation of aging effect on healthy cerebral white matter.

Objective To investigate the low-dose CT image denoising and generalization performance of the existing mainstream deep learning based denoising networks.Methods The public AAPM Mayo challenge dataset was used to train the denoising network using 3 image-domain methods(REDCNN,WGAN-VGG,CTformer)and 2 projection-image dual-domain methods(VVBP-UNet,CLEAR),separately.The denoising networks were evaluated quantitatively for peak signal-to-noise ratio(PSNR),structural similarity index,root mean square error,number of network parameters and floating point operations,and their generalization performance was analyzed on the AbdomenCT-1K Dataset.Results Image-domain denoising networks effectively suppressed low-dose CT image noise,with REDCNN demonstrating the best denoising performance and achieving a PSNR of 42.0988 dB.The dual-domain denoising networks were better at preserving tiny tissue structures while removing image noise,with VVBP-UNet performing the best and increasing PSNR to 42.150 9 dB.Conclusion The projection-image dual-domain method exhibits superior denoising and generalization performances than the image-domain method,despite requiring a relatively large amount of network parameters and computations.When computing resources are sufficient,the denoising results obtained by dual-domain method better fulfill the requirements for clinical diagnosis.

Objective To construct YOLOv10 based artificial intelligence(AI)models for the automatic detection in small bowel capsule endoscopy(SBCE)images.Methods SBCE data from two centers was collected,including 23 115 images and 35 412 annotated labels covering 11 categories of small bowel lesions.The images were annotated using the LabelMe tool and converted into the YOLO format required for deep learning model development.The pre-trained YOLOv10 and YOLOv8 models were used for transfer learning training on the constructed dataset.Model performance was comprehensively evaluated using metrics such as precision,accuracy,sensitivity,specificity,false-positive rate,and detection speed.Finally,the models were deployed on local computers for real-time detection of SBCE images and videos.Results Six different versions of YOLO object detection models were developed,namely YOLOv8n,YOLOv8s,YOLOv8m,YOLOv10n,YOLOv10s,and YOLOv10m.On the validation set,YOLOv10s model achieved the best mAP50(0.795);although its inference latency was not the fastest(4.803 ms/img),it met the requirements for clinical application.On the test set,YOLOv10s performed well,with an accuracy of 92.69%,a sensitivity of 89.23%,and a false-positive rate of 4.78%.Especially,in category-specific inference,the highest sensitivity was for"bleeding"at 96.41%,while the lowest was for"narrowing"at 82.29%.Conclusion The model constructed based on YOLOv10 neural network can rapidly and accurately detect and classify various small bowel lesions,exhibiting significant clinical application potential.

Objective To test the hypothesis that backward crawling and forward crawling share similar inter-joint coordination patterns,thus providing potential evidence for the application of backward crawling in rehabilitation training.Methods The acceleration signals in the X,Y,and Z directions for 9 joints(including bilateral wrists,elbows,shoulders,knees,and hips)in 9 volunteers during forward and backward crawling were collected using a custom signal acquisition system,and the pressure signals were also recorded when the palms contacted the ground.The collected acceleration signals were preprocessed,segmented into cycles,and vectorized.Based on the pressure signals,a single crawling cycle was divided into support phase and swing phase.In addition,principal component analysis was applied to extract inter-joint coordination in limbs at various scales(sagittal,coronal,and transverse planes).Pearson correlation coefficients of inter-joint coordination patterns were compared between forward and backward crawling in support period,swing period,and full cycle.Results The correlation coefficients for coordination patterns in the full cycle at the transverse plane scale were 0.813 5(PC1)and 0.837 5(PC2),and the correlation coefficient of the support period PC2 was 0.901 8.At the sagittal plane scale,the correlation coefficient of the support period PC1 was 0.948 5.Conclusion The study provides preliminary evidence that limb motion coordination patterns during backward crawling are similar to those observed during forward crawling.Future research will further explore the effects of backward crawling on functional rehabilitation in individuals with motor impairments.

The study introduces the technical characteristics(product type,usage mode,temperature control mode,use environment,etc.)of temperature control equipment for extracorporeal membrane oxygenation,and evaluates effectiveness and safety of the product based on its technical characteristics.Main contributions of the study are as follows:(1)clarifying key technical indicators for effectiveness and safety,as well as common risk sources;(2)putting forward the requirements for the researches on product performance,cleaning and disinfection,and stability.

As a non-invasive physiological indicator,electroencephalography signal provides an objective assessment of psychological stress levels among nursing staff in major public health emergencies,offering a scientific basis for targeted psychological interventions while overcoming the limitations of subjective bias inherent in traditional questionnaire-based methods.A psychological stress classification model based on bidirectional long short-term memory and attention mechanism is proposed to classify the psychological stress of clinical nurses more effectively by analyzing their electroencephalography signals.Experimental results show that the proposed model exhibits better classification performance than the traditional long short-term memory model on the DREAMER dataset,the Feeling Emotions dataset and the self-built dataset.This study provides a novel approach for assessing psychological stress,which is helpful to improve the pertinence and effectiveness of clinical nursing work.

Due to the constraints of the surgical environment and operational space,laparoscopic surgical instruments employ wire-driven mechanisms.However,factors such as wire rigidity,hysteresis,and motor drive limitations result in the end-effector accuracy of surgical instruments failing to meet ideal requirements.To address the shortcomings of existing multi-degree-of-freedom laparoscopic surgical instruments in achieving end-effector precision,a motion compensation algorithm based on the Autogluon algorithm for a 4-degree-freedom laparoscopic surgical instrument is proposed.A single-degree-of-freedom surgical instrument driven by wire ropes was constructed,and machine learning was utilized to estimate the end-effector position.This estimated position served as a feedback compensation condition to control the end-effector of the surgical instrument.To validate the correctness of this method,it was compared with approaches such as neural networks,linear regression,decision trees,Gaussian processes,and support vector machines.The results demonstrated that the proposed method achieved the smallest mean squared error,maximum error,and mean absolute error,thereby verifying its effectiveness.

Objective To improve the three-dimensional gait and walking function of hemiplegic patients and enhance their quality of life through training with an end-driven lower limb rehabilitation robot.Methods Seventy elderly stroke patients with hemiplegia were enrolled and randomly divided into control group(n=35)and observation group(n=35).The patients in control group received conventional rehabilitation therapy after their conditions were stable,while the patients in observation group were trained by therapists with an end-driven lower limb rehabilitation robot besides the same treatment as control group.The rehabilitation efficacy was analyzed in terms of National Institute of Health Stroke Scale(NIHSS)score,Barthel Index(BI)score,Berg Balance Scale(BBS)score,Fugl-Meyer Assessment:Lower Extremity(FMA-LE),three-dimensional gait and satisfaction with nursing care.Results Before intervention,there was no significant difference between two groups in NIHSS score,BI score,BBS score,FMA-LE score,gait spatiotemporal parameters,gait temporal phase parameters,and ranges of motion of lower limb joints(P>0.05).After intervention,NIHSS score decreased,while BI score,BBS score and FMA-LE score increased in both groups(P<0.05),and the improvements were more obvious in observation group than in control group(P<0.05).The interventions in both groups increased step speed,step frequency and stride length,while reducing step width(P<0.05),with more significant improvements observed in observation group than in control group(P<0.05).Compared with those before intervention,the percentage of support period of the affected lower limb increased,and the ratio of the support period of the healthy side to that of the affected side and the percentage of double support period decreased(P<0.05),and these indexes were improved more prominent in observation group(P<0.05).The ranges of motion of lower limb joints in both groups were significantly greater than those before intervention(P<0.05),with greater ranges of motion in observation group than in control group(P<0.05).The total satisfaction with nursing care in observation group was higher than that in control group(97.14%vs 82.86%,P<0.05).Conclusion End-driven lower limb rehabilitation robot-assisted training can further improve the neurological function,lower limb function and balance ability in hemiplegic patients,with high patient satisfaction.The rehabilitation efficacy can be objectively assessed through three-dimensional gait analysis.