摘要
OBJECTIVE@#To propose an deep learning-based algorithm for automatic prediction of dose distribution in radiotherapy planning for head and neck cancer.@*METHODS@#We propose a novel beam dose decomposition learning (BDDL) method designed on a cascade network. The delivery matter of beam through the planning target volume (PTV) was fitted with the pre-defined beam angles, which served as an input to the convolution neural network (CNN). The output of the network was decomposed into multiple sub-fractions of dose distribution along the beam directions to carry out a complex task by performing multiple simpler sub-tasks, thus allowing the model more focused on extracting the local features. The subfractions of dose distribution map were merged into a distribution map using the proposed multi-voting mechanism. We also introduced dose distribution features of the regions-of-interest (ROIs) and boundary map as the loss function during the training phase to serve as constraining factors of the network when extracting features of the ROIs and areas of dose boundary. Public datasets of radiotherapy planning for head and neck cancer were used for obtaining the accuracy of dose distribution of the BDDL method and for implementing the ablation study of the proposed method.@*RESULTS@#The BDDL method achieved a Dose score of 2.166 and a DVH score of 1.178 (P < 0.05), demonstrating its superior prediction accuracy to that of current state-ofthe-art (SOTA) methods. Compared with the C3D method, which was in the first place in OpenKBP-2020 Challenge, the BDDL method improved the Dose score and DVH score by 26.3% and 30%, respectively. The results of the ablation study also demonstrated the effectiveness of each key component of the BDDL method.@*CONCLUSION@#The BDDL method utilizes the prior knowledge of the delivery matter of beam and dose distribution in the ROIs to establish a dose prediction model. Compared with the existing methods, the proposed method is interpretable and reliable and can be potentially applied in clinical radiotherapy.
Subject(s)
Humans , Deep Learning , Head and Neck Neoplasms/radiotherapy , Algorithms , Neural Networks, Computer摘要
OBJECTIVE@#To propose a coupled convolutional and graph convolutional network (CCGCN) model for diagnosis of Alzheimer's disease (AD) and its prodromal stage.@*METHODS@#The disease-related brain regions generated by group-wise comparison were used as the input. The convolutional neural networks (CNNs) were used to extract disease-related features from different locations on brain magnetic resonance (MR) images. The generated features via the graph convolutional network (GCN) were processed, and graph pooling was performed to analyze the inherent relationship between the brain topology and the diagnosis task adaptively. Through ADNI dataset, we acquired the accuracy, sensitivity and specificity of the diagnosis tasks for AD and its prodromal stages, followed by an ablation study on the model structure.@*RESULTS@#The CCGCN model outperformed the current state-of-the-art methods and showed a classification accuracy of 92.5% for AD with a sensitivity of 88.1% and a specificity of 96.0%.@*CONCLUSIONS@#Based on the structural and topological features of the brain MR images, the proposed CCGCN model shows excellent performance in AD diagnosis and is expected to provide important assistance to physicians in disease diagnosis.
Subject(s)
Humans , Alzheimer Disease , Diagnostic Imaging , Brain , Magnetic Resonance Imaging , Neural Networks, Computer摘要
OBJECTIVE@#To establish an algorithm based on 3D convolution neural network to segment the organs at risk (OARs) in the head and neck on CT images.@*METHODS@#We propose an automatic segmentation algorithm of head and neck OARs based on V-Net. To enhance the feature expression ability of the 3D neural network, we combined the squeeze and exception (SE) module with the residual convolution module in V-Net to increase the weight of the features that has greater contributions to the segmentation task. Using a multi-scale strategy, we completed organ segmentation using two cascade models for location and fine segmentation, and the input image was resampled to different resolutions during preprocessing to allow the two models to focus on the extraction of global location information and local detail features respectively.@*RESULTS@#Our experiments on segmentation of 22 OARs in the head and neck indicated that compared with the existing methods, the proposed method achieved better segmentation accuracy and efficiency, and the average segmentation accuracy was improved by 9%. At the same time, the average test time was reduced from 33.82 s to 2.79 s.@*CONCLUSIONS@#The 3D convolution neural network based on multi-scale strategy can effectively and efficiently improve the accuracy of organ segmentation and can be potentially used in clinical setting for segmentation of other organs to improve the efficiency of clinical treatment.
Subject(s)
Humans , Head , Image Processing, Computer-Assisted , Neck , Neural Networks, Computer , Organs at Risk , Tomography, X-Ray Computed摘要
OBJECTIVE@#To train convolutional networks using multi-lead ECG data and classify new data accurately to provide reliable information for clinical diagnosis.@*METHODS@#The data were pre-processed with a bandpass filter, and signal framing was adopted to adjust the data of different lengths to the same size to facilitate network training and prediction. The dataset was expanded by increasing the sample size to improve the detection rate of abnormal samples. A depth-wise separable convolution structure was used for more specific feature extraction for different channels of twelve-lead ECG data. We trained the two classifiers for each label using the improved DenseNet to classify different labels.@*RESULTS@#The propose model showed an accuracy of 80.13% for distinguishing between normal and abnormal ECG with a sensitivity of 80.38%, a specificity of 79.91% and a F1 score of 79.35%.@*CONCLUSIONS@#The model proposed herein can rapidly and effectively classify the ECG data. The running time of a single dataset on GPU is 33.59 ms, which allows real-time prediction to meet the clinical requirements.
Subject(s)
Humans , Algorithms , Arrhythmias, Cardiac , Diagnosis , Databases as Topic , Electrocardiography , Classification , Methods , Neural Networks, Computer , Sensitivity and Specificity摘要
OBJECTIVE@#To establish a cone beam computed tomography (ECBCT) system for high-resolution imaging of the extremities.@*METHODS@#Based on three-dimensional X-Ray CT imaging and high-resolution flat plate detector technique, we constructed a physical model and a geometric model for ECBCT imaging, optimized the geometric calibration and image reconstruction methods, and established the scanner system. In the experiments, the pencil vase phantom, image quality (IQ) phantom and a swine feet were scanned using this imaging system to evaluate its effectiveness and stability.@*RESULTS@#On the reconstructed image of the pencil vase phantom, the edges were well preserved with geometric calibrated parameters and no aliasing artifacts were observed. The reconstructed images of the IQ phantom showed a uniform distribution of the CT number, and the noise power spectra were stable in multiple scanning under the same condition. The reconstructed images of the swine feet had clearly displayed the bones with a good resolution.@*CONCLUSIONS@#The ECBCT system can be used for highresolution imaging of the extremities to provide important imaging information to assist in the diagnosis of bone diseases.
Subject(s)
Animals , Algorithms , Artifacts , Calibration , Cone-Beam Computed Tomography , Methods , Equipment Design , Extremities , Diagnostic Imaging , Image Processing, Computer-Assisted , Methods , Phantoms, Imaging , Radiographic Image Enhancement , Methods , Swine摘要
<p><b>OBJECTIVE</b>Accurate segmentation of lung fields in chest radiographs (CXR) is very useful for automatic analysis of CXR. In this work, we propose to use dense matching of local features and label fusion to automatically segment the lung fields in CXR.</p><p><b>METHODS</b>For an input CXR, the dense Scale Invariant Feature Transform (SIFT) descriptors and raw image patches were extracted as the local features for each pixel. The nearest neighbors of the local features were then quickly searched by dense matching directly from the whole feature dataset of the reference images. The dense matching included three steps: limited random initialization, propagation of nearest neighbor field, and limited random search, with iteration of the last two steps for several times. The label image patches for each pixel were extracted according to the nearest neighbor field and weighted by the matching similarity. Finally, the weighted label patches were rearranged as the label class probability image of the input CXR, from which thresholds were obtained for segmentation of the lung fields.</p><p><b>RESULTS</b>The Jaccard index of the proposed method reached 95.5% on the public JSRT dataset.</p><p><b>CONCLUSION</b>A high accuracy and robustness can be obtained by adopting dense matching of local features and label fusion to segment the lung fields in CXR, and the result is better than that of current segmentation method.</p>
Subject(s)
Humans , Algorithms , Cluster Analysis , Lung , Radiographic Image Interpretation, Computer-Assisted , Radiography, Thoracic摘要
We propose a multi-weighted probabilistic atlas to obtain accurate, robust, and reliable segmentation. The local similarity measure is used as the weight to compute the probabilistic atlas, and the distance field is used as the weight to incorporate the locality information of the atlas; the self-similarity is used as the weight to incorporate the local information of target image to refine the probabilistic atlas. Experimental results with brain MRI images showed that the proposed algorithm outperforms the common brain image segmentation methods and achieved a median Dice coefficient of 87.1% on the left hippocampus and 87.6% on the right.
Subject(s)
Humans , Algorithms , Brain , Magnetic Resonance Imaging , Neuroimaging摘要
<p><b>OBJECTIVE</b>We propose a cross-correlation method for automatic extraction of the pennation angle (PA) of the gastrocnemius (GM) muscle from ultrasound radiofrequency (RF) signals.</p><p><b>METHODS</b>The ultrasound RF signals of the GM muscles in tension condition from normal subjects and the simulated ultrasound signals were collected. After the starting point of tracking, a fascicle was selected in the reconstructed GM ultrasound image from the RF signals, and the fascicle and deep aponeurosis could be automatically tracked using the cross-correlation algorithm. The lines of the fascicle and deep aponeurosis were then drawn and the PA was calculated. The reproducibility of the proposed method and its consistency with the manual measurement method were tested.</p><p><b>RESULTS</b>The angles of the simulated fascicles were precisely extracted automatically. The difference between the experimental measurement and the theoretical values was less than 1°. The PA measured automatically and manually was 20.48°∓0.47° and 21.49°∓1.79°, respectively. The coefficient of variation (CV) of the two methods was less than 3% and the root-mean square error (RMSE) was less than 1°. Bland-Altman plot showed a good agreement between the proposed automatic method and the manual method.</p><p><b>CONCLUSION</b>The proposed cross-correlation automatic measurement method can detect the orientation of the fascicle and deep aponeurosis and measure the PA based on ultrasound RF signals with serious speckle noise.</p>
Subject(s)
Humans , Algorithms , Image Processing, Computer-Assisted , Muscle, Skeletal , Diagnostic Imaging , Radio Waves , Reproducibility of Results , Ultrasonography摘要
AIM:To explore the therapeutic effect of a novel Rho kinase inhibitor WAR 5 on the experimental autoimmune encephalomyelitis (EAE) and its possible mechanism.METHODS: Female C57BL/6 mice were randomly divided into EAE group and WAR5 group.EAE model was induced by the application of MOG 35-55 peptide.WAR5 was in-jected intraperitoneally every other day from post-immunization (PI) day 3 to PI day 27.The clinical score and body weight were recorded every other day .On PI day 28, the animals were sacrificed and spinal cords were obtained for HE and mye-lin staining .The splenocytes were isolated and the expression of CD 16/32 and CD206 were analyzed by flow cytometry . The protein extracts from the brains and spinal cords were collected for the measurement of inducible nitric oxide synthase ( iNOS) by Western blotting .RESULTS:The administration of WAR 5 delayed the onset of EAE and attenuated the clini-cal symptoms .The results of the pathological examination revealed that WAR 5 inhibited the infiltration of inflammatory cells and improved myelination in spinal cords , accompanied with the poralization of M 1 macrophages to M2 phenotype in the spleen.WAR5 inhibited the expression of iNOS in the central nervous system , especially in the spinal cords .CON-CLUSION:The therapeutic effect of WAR5 on EAE may be related to the shift of M1 macrophages to M2 phenotype and inhibition of inflammation in the central nerve system .
摘要
We propose an effective algorithm for accurate 3D segmentation of CT liver images based on statistical and specific information. We present a new intensity model which combines patient-specific intensity information of boundary with the statistical information for liver segmentation. Compared to the traditional methods, our approach not only produces excellent segmentation accuracy, but also increases the robustness.
Subject(s)
Humans , Algorithms , Imaging, Three-Dimensional , Methods , Liver , Diagnostic Imaging , Liver Diseases , Diagnostic Imaging , Liver Neoplasms , Diagnostic Imaging , Models, Statistical , Radiographic Image Interpretation, Computer-Assisted , Methods , Tomography, X-Ray Computed , Methods摘要
Objective: 3D segmentation is an important part of medical image analysis and visualization. It also continues to be large challenge in the medical image segmentation. While level sets have demonstrated a great potential for 3D medical image segmentation, these algorithms have a large computational burden thus are not suitable for real time processing requirement. To solve this problem, we propose a parallel accerelated method based on CUDA. Methods: We implement C-V level set algorithm in the CUDA environment which is the NVIDIA's GPGPU model.The segmentation speed can greatly improved by using independence of image pixel and concurrence of partial differential equation .The paper shows the flow chart of the parallel computing and gives the detailed introduction of the C-V level set algorithm which is implemented in the CUDA environment. Results: Realizing the C-V level set parallel accerelated algorithm. This method has faster segmentation speed while preserving the qualitative results, Conclusions: This method is viable and makes the fast 3D medical image segmentation come hue.
摘要
Objective: Real time medical image registration technique is one of the key techniques in image based surgery navi-gation system. While in medical image analysis, image registration is usually a very time-cousuming operation, and this is not conducive to clinical real-time requirements. This paper studies and realizes the acceleration of the process of image registra-tion. Methods: In order to improve the regisWation rate, in this paper, we propose a new technology which is based on CUDA (Compute Unified Device Architecture) programming model to accelerate the process of registration in hardware, using paral-lel methods to achieve pixel coordinate transformation, linear interpolation, and calculate the corresponding pixel gray value residuals. Results: The registration is up to the sub-pixel level and the GPU-based registration is dozens or even hundreds of times faster than CPU-based registration. Conclusions: This method greatly enhances the speed of rigid registration without changing the alignment accuracy.
摘要
Objective To investigate the evolution of monkey brain isehemic penumbra(IP)area.Methods Seyen monkeys were used to make middle cerebral artery occlusion(MCAO)model by interventional methods.CT perfnsion imaging,MR diffusion weighted-imaging (DWI),perfusion weighted imaging(PWI)and T2W1 were performed at 1,5,10;15,20 and 24 h after MCAO respectively.Four regions of interest of infarct lesion were measured.Point 1 was at the infarct core.point 3 was at the infamt margin,and point 2 was at the midpoint between point 1 and 3.Point 4 demonstrated normal signal intensity adjacent to high signal intensity.Parameters measured included cerebral blood flow (CBF), cerebral blood volume (CBV), mean transit time (MTT), apparent diffusion coefficient (ADC) and negative enhancement integral (NEI).The relative ratios between infarct lesions and the corresponding contralateral normal brain were calculated(rCBF, rCBV, rMTY, rADC and rNEI).The IP areas were calculated by two methods: IP thresholds combined with self-made computer software, and PWI( MrlT)-DWI mismatch.ANOVA and ROC analysis were used. Results Five of 7 monkey MCAO models were made successfuUy. There were signitlcanfly difference of rCBF and rNEI within 20 h, of rCBV within 15 h, of rADC within 10 h, of rMTT at24 h (P<0.05).ROI 1,2 and 3 values as following: rCBF: 1 h(0.160 ±0.034, 0.310 ±0.037,0.540±0.107), 5 h(0.098±0.029, 0.157 ± 0.052, 0.427 ±0.116), 10 h(0.072 ±0.023, 0.097 ±0.028, 0.209 ± 0.070), 15 h(0.054 ± 0.017, 0.069 ± 0.015, 0.166 ± 0.049), 20 h(0.038 ± 0.011,0.026± 0.007, 0.092±0.013); rNEI: 1 h(0.219 ± 0.085, 0.303 ± 0.099, 0.463 ± 0.132), 5 h (0.143±0.057, 0.195± 0.055, 0.348±0.127), 10 h(0.127 ± 0.029, 0.171 ± 0.058, 0.259 ±0.079), 15 h(0.128 ±0.024, 0.164 ±0.031, 0.217 ±0.030), 20 h(0.075±0.019, 0.147±0.058,0.129 ±0.045) ; rCBV: 1 h(0.594 ± 0.199, 0.804 ± 0.099, 1.021 ±0.169), 5 h(0.457±0.103,0.462±0.145, 0.815±0.201), 10 h(0.222 ±0.046, 0.249±0.065, 0.529 ±0.135), 15 h(0.201 ±0.047, 0.187 ±0.055, 0.361 ±0.083) ; rADC: 1 h(0.515 ±0.115, 0.667±0.097, 0.761 ±0.106),5 h(0.488 ±0.100, 0.539 ±0.107, 0.674 ±0.099), 10 h(0.456 ±0.057, 0.549 ±0.049, 0.590 ±0.081 ) ; 24 h rMTT(4.163 ± 1.179, 4.192± 1.607, 2.397±0.909).The thresholds of IP were 0.203 of rCBF, 0.483 of rCBV, 0.571 of rADC and 0.250 of rNEI respectively.The values measured using the method of IP thresholds combined with software were larger than PWI(MTr)-DWI mismatch region before 15 hours, but were smaller at 20 h and 24 h. The area of IP was 20%-38% of infarct area at 1 h,15%-35% at 5-10 h, 13%-25% at 15 h, 9%-15% at 20 h, and 3%-12% at 24 h.Conclusion The time window of IP in monkey MCAO model was 15%-20 hours.At the early phase of infarction, IP was present within the region of high signal intensity on DWI.PWI-DWI mismatch method could not estimate IP areas accurately.Areas evaluated with CT perfusion (MrIT) and DWI mismatch were much closer to the actual IP areas.
摘要
Objective To develop a virtual endoscopy system which can be integrated into PACS.Methods Key techniques on virtual endoscopy were researched and we implemented a virtual endoscopy system with the help of the Visualization Toolkit VTK.Results The Virtual endoscopy system was integrated into PACS and the post-processing function of PACS was advanced.Conclusion As a novel medical image post-processing technology,virtual endoscopy provides a completely non-invaded inspection,so it has broad application prospects in the computer-aided medical teaching,surgical navigation,surgical planning and clinical diagnosis.
摘要
Objective To compare the difference in bioavailability of microemulsion and pill by detecting the concentration of andrographolide in rabbit plasma.Methods RP-HPLC was used to detect the concentration of andrographolide in rabbits plasma at different time,and software 3p87 was used to analyze the pharmacokinetics parameter.Results The pharmacokinetics parameters of andrographolide oral microemulsion and andrographolide pill were as follows::AUC0-7=1 406.72 ? g? mL-1? min,Tpeak=27.08 min,Cmax=5.37 ? g/mL for microemulsion;AUC0-7=877.37 ? g? mL-1? min,Tpeak=61.04 min,Cmax=3.06 ? g/mL for the pill.Conclusion Oral microemulsion of andrographolide has a shorter peak time and higher biological availability than the pill.
摘要
Objective To study the pharmacokinetics and the relative bioavailabi lity of micro-emulsified genistein in rabbits. Methods Rabbits received gastric gavage of micro-emulsified genistein and CMC-Na suspension of genistein. Then the genistein content in rabbit plasma at different time was determined by HPLC ,concentration-time curve was drafted,and the pharmacokinetic parameters and the relative bioavailability were calculated. Results The main parameters of mic ro-emulsified genistein and CMC-Na suspension of genistein were as follows:AU C0→10h being(24.90?1.24)and(10.71?0.86)?g?h?mL-1,Cmax being(4.02?0.20)a nd(0.99?0.04)?g?mL-1,Tmax being 2 h and 4 h,respectively. The relative bio availability of micro-emulsified genistein was 232.49 %. Conclusion Micro-emu lsified genistein system can improve the bioavailability of genistein evidently.
摘要
VTK is not only an open-source code but also a powerful toolkit for visualization.According to the requirement of medical image processing and on the basis of visualization and display function of VTK,VC++6.0 is used to design and implement a three-dimensional system which can be integrated into PACS.This system is useful for advancing the post-processing function of PACS.
摘要
Objective To present a new algorithm for multidimensional medical image registration from global registration to local registration in sequence. Methods Firstly, the global registration was achieved by the method of affine transformation composed of B-splines,whose knots were the four vertexes of the medical image. Then the knots of the B-splines were increased, and the transformation function was more complex and elastic than ever,which completed the elastic aligning for the detail of the medical image. Results The whole registration algorithm represented the principle aligning from global registration to local registration. Conclusion It is proved by experiments that the presented algorithm can decrease the time of calculation and increase the robustness of registration.