Video-based intelligent action recognition remains challenging in the field of computer vision.The review analyzes the state-of-the-art methods of video-based intelligent action recognition,including machine learning methods with handcrafted features,deep learning methods with automatically extracted features,and multi-information fusion methods.In addition,the important medical applications and limitations of these technologies in the past decade are introduced,and the interdisciplinary views on the future application to improve human health are also shared.

Although it has high resolution for soft tissues,magnetic resonance imaging(MRI)is not the standard for chest imaging,which results in an insufficient amount of expert-annotated MRI data.Therefore,CT image is usually converted into MRI image.To overcome the difficulty of obtaining the corresponding modal CT and MRI images,a CSCGAN model with CycleGAN as the framework is proposed based on the structural characteristics of generative adversarial networks.Considering the possibility of mode collapse in CycleGAN,StyleGan2 which can control the style and feature details of the synthetic image and realize the synthesis of high-resolution images is integrated into CycleGAN for reconstructing the generator.A noise module is introduced to reduce external interference.In addition,in order to prevent the loss of tumors during conversion,the discriminator structure of the network is modified,and a mixed attention mechanism is added.Experimental results show that compared with the images generated by other methods,those generated by the proposed model are improved in Dice similarity coefficient,Hausdorff distance,volume ratio and mean intersection over union,indicating that the proposed method can effectively realize the mode conversion of liver tumor images,and that the generated data can improve the segmentation accuracy.

Objective To explore the predictive value of CT radiomics and morphological features for the prognosis and survival in non-small cell lung cancer(NSCLC)patients.Methods The clinic data of 300 NSCLC patients(300 lesions)were downloaded from the Cancer Imaging Archive,with 210 randomly selected as the training set and 90 as the test set.According to the prognosis and survival,the patients were divided into two groups with survival period≤3 and＞3 years.3D Slicer software was used to delineate the regions of interest layer by layer in CT images,and the radiomics features were extracted from each region of interest.Both t-test and least absolute shrinkage and selection operator were utilized for radiomics feature screening.Three types of prediction models,namely radiomics model,morphological model and combined model,were constructed with Logistic regression,whose performances were evaluated using the receiver operating characteristic(ROC)curve.Results The differences in radiomics labels and mediastinal lymph node metastasis between the training set and the test set were statistically significant.For radiomics model,morphological model and combined model,the area under the ROC curve was 0.784(95%CI:0.722-0.847),0.734(95%CI:0.664-0.804)and 0.748(95%CI:0.680-0.815)in the training set,and 0.737(95%CI:0.630-0.844),0.665(95%CI:0.554-0.777)and 0.687(95%CI:0.578-0.797)in the test set,which demonstrated that radiomics model had the best diagnostic performance.Conclusion The CT radiomics model can effectively predict the prognosis and survival in NSCLC patients.

Objective To compare the performances of 3D MERGE sequence and 3D SPACE STIR sequence in detecting lumbar disc herniation(LDH).Methods The clinical data and MRI data of 135 LDH patients admitted between January 2020 and November 2022 were analyzed retrospectively.All patients were examined using conventional MRI,3D MERGE sequence and 3D SPACE STIR sequence.The consistency of 3D MERGE sequence and 3D SPACE STIR sequence in measuring the diameter of nerve root was analyzed,and the image quality parameters[signal-to-noise ratio(SNR),contrast-to-noise ratio(CNR)]and image definition score of the two sequences were evaluated.Results There were no statistically significant differences in L3-S1 nerve root diameters measured by 3D MERGE sequence and 3D SPACE STIR sequence(P＞0.05),and the diameters of L3,L4,L5 and S1 measured by the two sequences showed high correlations(r=0.957,0.986,0.975,0.972,P＜0.05).Compared with 3D SPACE STIR sequence,3D MERGE sequence had higher SNR and CNR,scored better on image definition,and displayed nerve root more clearly(P＜0.05).Conclusion 3D MERGE sequence and 3D SPACE STIR sequence have high consistency in the measurement of LDH nerve root diameter.3D MERGE sequence can display the anatomical morphology of nerve root more clearly as compared with 3D SPACE STIR sequence,and the former one has higher image quality.

A U-Net incorporating improved Transformer and convolutional channel attention module is designed for biventricular segmentation in MRI image.By replacing the high-level convolution of U-Net with the improved Transformer,the global feature information can be effectively extracted to cope with the challenge of poor segmentation performance due to the complex morphological variation of the right ventricle.The improved Transformer incorporates a fixed window attention for position localization in the self-attention module,and aggregates the output feature map for reducing the feature map size;and the network learning capability is improved by increasing network depth through the adjustment of multilayer perceptron.To solve the problem of unsatisfactory segmentation performance caused by blurred tissue edges,a feature aggregation module is used for the fusion of multi-level underlying features,and a convolutional channel attention module is adopted to rescale the underlying features to achieve adaptive learning of feature weights.In addition,a plug-and-play feature enhancement module is integrated to improve the segmentation performance which is affected by feature loss due to channel decay in the codec structure,which guarantees the spatial information while increasing the proportion of useful channel information.The test on the ACDC dataset shows that the proposed method has higher biventricular segmentation accuracy,especially for the right ventricle segmentation.Compared with other methods,the proposed method improves the DSC coefficient by at least 2.83%,proving its effectiveness in biventricular segmentation.

A lightweight residual attention enhanced Unet(LRAE-Unet)is designed for the fully automatic brain tumor segmentation.LRAE-Unet uses lightweight residual module to solve the problems of gradient disappearance and network degradation when the network layers increases,lightweight self-attention module to suppress the irrelevant areas and highlight the significant features of specific local areas,and enhanced average pooling module with a larger field of perception to reduce the space of feature map,save computing resources and avoid over-fitting.The experiment on BraTS 2019 dataset shows that the proposed method has a Dice similarity coefficient of 91.24%,88.64%and 88.32%in the segmentations of the whole tumor,tumor core and enhanced tumor,which proves its feasibility and effectiveness for brain tumor segmentation.

Objective To analyze the diagnostic utility of combining susceptibility-weighted imaging(SWI)with arterial spin labeling(ASL)in patients with acute ischemic stroke(AIS).Methods Fifty AIS patients who admitted to Yongchuan Hospital,Chongqing Medical University from July 2020 to July 2021 were selected.Scans were performed using a 3.0T MRI scanner,including sequences such as FLAIR,DWI,3D-TOF-MRA,3D-ASL,and SWI.The perfusion status of the infarction core,the grading of draining veins around the infarction core,compensation by collateral circulation,the occurrence of hemorrhagic transformation,and prognosis were assessed.Results The grading of draining veins around the infarction core was significantly correlated with NIHSS scores(r=0.869,P＜0.05)and prognosis(r=0.825,P＜0.05).In addition,significant correlations were found between the perfusion status of the infarction core and the occurrence of hemorrhagic transformation(r=0.873,P＜0.05),compensation by collateral circulation and prognosis(r=0.883,P＜0.05).Conclusion The combination of SWI and ASL provides accurate indications of the hemodynamic conditions around the infarction core in AIS patients,and it can accurately assess the prognosis of AIS patients,contributing valuable information for clinical diagnosis and the selection of treatment strategies.

A dose reconstruction algorithm for electrionic portal imaging device(EPID)based on calibration and calculation is developed.The raw data of EPID in continuous acquisition mode are corrected for dark field and gain,and the gray level features of bright field are used to determine the field boundary.Subsequently,MU calibration,off-axis calibration and field size calibration are performed on the EPID data,and dose reconstruction is carried out based on the calibrated superimposed flux and the Monte Carlo model of the linac head.Nine cases of IMRT plans are selected for verification and measurement using EPID and MapCheck separately,and the passing rates between the two tools are compared under different gamma criteria(3%/3 mm and 2%/2 mm).For a planned case,the average passing rates of multiple cases verified by MapCheck under the two criteria were 99.02%±1.28%and 90.84%±4.49%,and the average passing rates of the EPID reconstruction models were 98.86%±1.19%and 91.39%±4.80%.Compared with MapCheck,the EPID reconstruction algorithm based on calibration and calculation has no significant difference in the passing rate of IMRT plan verification(P＞0.05),which meets the clinical requirements of dose verification.

When dealing with public health emergencies,telemedicine can optimize the allocation of medical resources of primary healthcare institutions quickly.Therefore,a blood oxygen saturation monitoring system based on cellular internet of things is designed in the study.Compared with the traditional medical blood oxygen saturation monitors,the system is wearable,low-cost and easy-to-operate,and it is more suitable for the scenario of rapid detection at the primary healthcare institutions or user monitoring at home.The in-ear earphone model makes the detection module innovatively.Both blood oxygen saturation and body temperature can be obtained simultaneously,and the monitoring data are transmitted to the database through narrow band internet of things.The accumulated data provides effective support for personalized diagnosis and treatment.

A non-invasive deep learning method is proposed for reconstructing arterial blood pressure signals from photoplethysmography signals.The method employs U-Net as a feature extractor,and a module referred to as bidirectional temporal processor is designed to extract time-dependent information on an individual model basis.The bidirectional temporal processor module utilizes a BiLSTM network to effectively analyze time series data in both forward and backward directions.Furthermore,a deep supervision approach which involves training the model to focus on various aspects of data features is adopted to enhance the accuracy of the predicted waveforms.The differences between actual and predicted values are 2.89±2.43,1.55±1.79 and 1.52±1.47 mmHg on systolic blood pressure,diastolic blood pressure and mean arterial pressure,respectively,suggesting the superiority of the proposed method over the existing techniques,and demonstrating its application potential.