Single-photon avalanche diode(SPAD)arrays are solid-state detectors that offer imaging capabilities at the level of individual photons,with unparalleled photon counting and time-resolved performance.This fascinating t...Single-photon avalanche diode(SPAD)arrays are solid-state detectors that offer imaging capabilities at the level of individual photons,with unparalleled photon counting and time-resolved performance.This fascinating technology has progressed at a very fast pace in the past 15 years,since its inception in standard CMOS technology in 2003.A host of architectures have been investigated,ranging from simpler implementations,based solely on off-chip data processing,to progressively“smarter”sensors including on-chip,or even pixel level,time-stamping and processing capabilities.As the technology has matured,a range of biophotonics applications have been explored,including(endoscopic)FLIM,(multibeam multiphoton)FLIM-FRET,SPIM-FCS,super-resolution microscopy,time-resolved Raman spectroscopy,NIROT and PET.We will review some representative sensors and their corresponding applications,including the most relevant challenges faced by chip designers and end-users.Finally,we will provide an outlook on the future of this fascinating technology.展开更多
This article studies numerically a familiar important phenomenon in spray combustion which is deformation and breakup of liquid drops in gas flow. The SIMPLER method is used to solve the two-dimensional （2D） unstead...This article studies numerically a familiar important phenomenon in spray combustion which is deformation and breakup of liquid drops in gas flow. The SIMPLER method is used to solve the two-dimensional （2D） unsteady axisymmetric Navier-Stokes equations for both the drop and the ambient gas flow. The level set method is applied to capturing the liquid/gas interface. Through calculation are obtained four typical breakup modes--oscillation, bag breakup, sheet stripping breakup and shear breakup governed by four non-dimensional numbers which are gas Weber number （Weg）, liquid Reynolds number （Rel）, gas Reynolds number （Reg） and density ratio （γ）. Their effects upon each mode are analyzed. The results indicate that among the four numbers, Weg is of the highest importance with Rel, Reg and γfollowing up. By widening the range of the density ratio up to 1 000, the breakup mode is discovered to be so complicated that a new one called multimode breakup mode turns up. This mode contains the shearing breakup and piercing breakup, which successively happen. The calculation results agree well with what is observed from the experiments.展开更多
Since establishing a business in her rural hometown in 2017,Shi Jia,an ethnic Miao entrepreneur,has been committed to exploring ethnic resources to develop local industries.“Compared with other embroidery styles,the ...Since establishing a business in her rural hometown in 2017,Shi Jia,an ethnic Miao entrepreneur,has been committed to exploring ethnic resources to develop local industries.“Compared with other embroidery styles,the Miao embroidery focuses on vivid expression.At the beginning,it was used to record big events in place of writing,and distinguish tribes by different totems.Its color is brighter,and the design is simpler.There are also many auspicious patterns to pray for good luck,”Shi Jia,a deputy to the National People’s Congress(NPC)and president of a Miao clothing company,told China Today.展开更多
Despite ML’s impressive performance in commercial applications,several unique challenges exist when applying ML in materials science applications.In such a context,the contributions of this work are twofold.First,we ...Despite ML’s impressive performance in commercial applications,several unique challenges exist when applying ML in materials science applications.In such a context,the contributions of this work are twofold.First,we identify common pitfalls of existing ML techniques when learning from underrepresented/imbalanced material data.Specifically,we show that with imbalanced data,standard methods for assessing quality of ML models break down and lead to misleading conclusions.Furthermore,we find that the model’s own confidence score cannot be trusted and model introspection methods(using simpler models)do not help as they result in loss of predictive performance(reliability-explainability trade-off).Second,to overcome these challenges,we propose a general-purpose explainable and reliable machine-learning framework.Specifically,we propose a generic pipeline that employs an ensemble of simpler models to reliably predict material properties.We also propose a transfer learning technique and show that the performance loss due to models’simplicity can be overcome by exploiting correlations among different material properties.A new evaluation metric and a trust score to better quantify the confidence in the predictions are also proposed.To improve the interpretability,we add a rationale generator component to our framework which provides both model-level and decision-level explanations.Finally,we demonstrate the versatility of our technique on two applications:(1)predicting properties of crystalline compounds and(2)identifying potentially stable solar cell materials.We also point to some outstanding issues yet to be resolved for a successful application of ML in material science.展开更多
Nonlinear encoding of optical information can be achieved using various forms of data representation.Here,we analyze the performances of different nonlinear information encoding strategies that can be employed in diff...Nonlinear encoding of optical information can be achieved using various forms of data representation.Here,we analyze the performances of different nonlinear information encoding strategies that can be employed in diffractive optical processors based on linear materials and shed light on their utility and performance gaps compared to the state-of-the-art digital deep neural networks.For a comprehensive evaluation,we used different datasets to compare the statistical inference performance of simpler-to-implement nonlinear encoding strategies that involve,e.g.,phase encoding,against data repetition-based nonlinear encoding strategies.We show that data repetition within a diffractive volume(e.g.,through an optical cavity or cascaded introduction of the input data)causes the loss of the universal linear transformation capability of a diffractive optical processor.Therefore,data repetition-based diffractive blocks cannot provide optical analogs to fully connected or convolutional layers commonly employed in digital neural networks.However,they can still be effectively trained for specific inference tasks and achieve enhanced accuracy,benefiting from the nonlinear encoding of the input information.Our results also reveal that phase encoding of input information without data repetition provides a simpler nonlinear encoding strategy with comparable statistical inference accuracy to data repetition-based diffractive processors.Our analyses and conclusions would be of broad interest to explore the push-pull relationship between linear material-based diffractive optical systems and nonlinear encoding strategies in visual information processors.展开更多
In the latest years,researchers from the industry and academia extensively applied machine learning algorithms in a broad range of domains.The goal of this special issue is to illustrate the most recent applications o...In the latest years,researchers from the industry and academia extensively applied machine learning algorithms in a broad range of domains.The goal of this special issue is to illustrate the most recent applications of deep learning methods in a range of real-life domains and to show the practical utility of these techniques.A particular attention goes towards methods to process network data that is capable of modelling complex artificial and natural systems as the interactions of a multitude of simpler entities.展开更多
1.Elements and human history The rich chemistry of Earth's elements played an inherent and multifaceted role in shaping the course of human history.Generally,the more inert an element is,the simpler it becomes to ...1.Elements and human history The rich chemistry of Earth's elements played an inherent and multifaceted role in shaping the course of human history.Generally,the more inert an element is,the simpler it becomes to extract it from minerals,facilitating its use in the crafting of tools,devices,and weapons.展开更多
Editor’s Note:Peter Barnes offers a critique of the OECD’s Pillars One and Two proposals;he lauds the OECD’s goals but expresses concern that the proposals depend on unrealistic assumptions.David Rosenbloom,in a co...Editor’s Note:Peter Barnes offers a critique of the OECD’s Pillars One and Two proposals;he lauds the OECD’s goals but expresses concern that the proposals depend on unrealistic assumptions.David Rosenbloom,in a companion commentary,goes further and says the proposals are too complex to work in today’s international tax environment.Both authors1 believe the OECD’s goal of increasing source jurisdiction taxation can be achieved in other simpler ways.展开更多
Objective:This study seeks to explore efficient and multiple-item detection modes in new-style HIV labs,as well as access the accuracy and reliability of CD4 cell count detected by point of care(POC)to analyze POC wor...Objective:This study seeks to explore efficient and multiple-item detection modes in new-style HIV labs,as well as access the accuracy and reliability of CD4 cell count detected by point of care(POC)to analyze POC work feasibility in district or county labs.Methods:POC devices adopted in grassroots-level labs and flow cytometers adopted in prefecture-level labs were used to analyze the same group of blood samples.The individual results were collected and compared for parametric tests in correlation and consistency.Results:The Pearson correlation coefficients(r)between results detected by FACSPresto and those by FACSCalibur,FACSVia,FACSCantoII,and EPICSXL were 0.922,0.938,0.914,and 0.823,respectively;the average deviations were−25.64,24.68,3.05,and 70.97 cells/μL,respectively;the Pearson correlation coefficient(r)between results by Pima and FACSCalibur,FACSVia,FACSCantoII,and EPICSXL were 0.900,0.950,0.954,and 0.876,respectively;and the average deviations were−73.99,−40.78,−29.32,and−22.75 cells/μL,respectively.Discussion:Strong positive correlations and good consistency were observed between the CD4 count tested by POC and flow cytometers.These findings provide theoretical support for new-style HIV labs and one-stop services,which can provide shorter testing duration and simpler testing processes,so that the most comprehensive testing results can be obtained in the shortest amount of time.展开更多
The flow instability through the side branch of a T-junction is analyzed in a numerical simulation. In a previous experimental study, the authors clarified the mechanism of fluid-induced vibration in the side branch o...The flow instability through the side branch of a T-junction is analyzed in a numerical simulation. In a previous experimental study, the authors clarified the mechanism of fluid-induced vibration in the side branch of the T-junction in laminar steady flow through the trunk. However, in that approach there were restrictions with respect to extracting details of flow behavior such as the flow instability and the distribution of wall shear stress along the wall. Here the spatial growth of the velocity perturbation at the upstream boundary of the side branch is investigated. The simulation result indicates that a periodic velocity fluctuation introduced at the upstream boundary is amplified downstream, in good agreement with experimental result. The fluctuation in wall shear stress because of the flow instability shows local extrema in both the near and distal walls. From the numerical simulation, the downstream fluid oscillation under a typical condition has a Strouhal number of 1.05, which approximately agrees with the value obtained in experiments. Therefore, this periodic oscillation motion is a universal phenomenon in the side branch of a T-junction.展开更多
A fuzzy control method was used to accelerate iteration convergence in numerical fluid dynamic simulation using SIMPLER algorithm. The residual ratio of momentum or energy equation between two successive iterations wa...A fuzzy control method was used to accelerate iteration convergence in numerical fluid dynamic simulation using SIMPLER algorithm. The residual ratio of momentum or energy equation between two successive iterations was used as the input variable. A fuzzy logic algorithm was developed in order to obtain the relative increment of the under-relaxation factor and its new value was then used for the next iteration. The algorithm was tested by four benchmark problems. In all cases considered, when the fuzzy control logic was used, convergence was achieved with nearly the minimum number of iterations, showing the feasibility of the proposed method.展开更多
Based on the first-order upwind and second-order central type of finite volume (UFV and CFV) scheme, upwind and central type of perturbation finite volume (UPFV and CPFV) schemes of the Navier-Stokes equations were de...Based on the first-order upwind and second-order central type of finite volume (UFV and CFV) scheme, upwind and central type of perturbation finite volume (UPFV and CPFV) schemes of the Navier-Stokes equations were developed. In PFV method, the mass fluxes of across the cell faces of the control volume (CV) were expanded into power series of the grid spacing and the coefficients of the power series were determined by means of the conservation equation itself. The UPFV and CPFV scheme respectively uses the same nodes and expressions as those of the normal first-order upwind and second-order central scheme, which is apt to programming. The results of numerical experiments about the flow in a lid-driven cavity and the problem of transport of a scalar quantity in a known velocity field show that compared to the first-order UFV and second-order CFV schemes, upwind PFV scheme is higher accuracy and resolution, especially better robustness. The numerical computation to flow in a lid-driven cavity shows that the under-relaxation factor can be arbitrarily selected ranging from (0.3) to (0.8) and convergence perform excellent with Reynolds number variation from 10~2 to 10~4.展开更多
Objective： To investigate the therapeutic efficacy of adenovirus-mediated herpes simplex virus thymidine kinase （HSV-tk） gene transfer under the driving of KDR promoter （AdKDR-tk） in combination of ganciclovir （...Objective： To investigate the therapeutic efficacy of adenovirus-mediated herpes simplex virus thymidine kinase （HSV-tk） gene transfer under the driving of KDR promoter （AdKDR-tk） in combination of ganciclovir （GCV） against human hepatocellular carcinoma in nude mice. Methods： HepG2 cell line was implanted subcutaneously into 32 nude mice, which were subsequently divided into 4 groups （n=8 each group）： Ganciclovir group （Ⅰ）, Ad group （Ⅱ）, AdCMV-tk/GCV group （under the driving of CMV promoter） （Ⅲ） and AdKDR-tk/GCV group （Ⅳ）. Then intratumoral injection of recombinant adenovirus or Ad was performed in all nude mice, and repeated 24 h later. For the following 10 d GCV was given at a dose of 100 mg/（kg· d）, ip. All the treated animals were killed to evaluate the tumor weight and the histopathological changes and the microvessel density of tumors after the treatment was determined. Results Compared with group Ⅰ, the tumor inhibitory rate was 12.3% in group Ⅲ and 24.5% in group Ⅳ; the inhibition rates were significantly different between group Ⅲand Ⅳ （P〈0.05）. The mean MVDs in group Ⅰ, Ⅱ, Ⅲ and Ⅳ were 37.4±8.6, 30.6±7.8, 27.6±7.1, and 10.7±4.1 （microvessels/mm^2）, respectively. Significant differences were found between group Ⅲ and Ⅱ （P〈0.05）, Ⅳ and Ⅱ （P〈0.01）, and Ⅳ and Ⅲ （P〈0.01）. Conclusion： Intratumoral injection of AdKDR-tk results in marked inhibition of HCC growth through inhibition angiogenesis in nude mice. It may be a new treatment approach for human HCC,展开更多
基金supported in part by the Swiss National Science Foundation Grant 166289.The authors would like to thank Scott Lindner and Arin Can Ulku(EPFL AQUA)for their kind assistance.
文摘Single-photon avalanche diode(SPAD)arrays are solid-state detectors that offer imaging capabilities at the level of individual photons,with unparalleled photon counting and time-resolved performance.This fascinating technology has progressed at a very fast pace in the past 15 years,since its inception in standard CMOS technology in 2003.A host of architectures have been investigated,ranging from simpler implementations,based solely on off-chip data processing,to progressively“smarter”sensors including on-chip,or even pixel level,time-stamping and processing capabilities.As the technology has matured,a range of biophotonics applications have been explored,including(endoscopic)FLIM,(multibeam multiphoton)FLIM-FRET,SPIM-FCS,super-resolution microscopy,time-resolved Raman spectroscopy,NIROT and PET.We will review some representative sensors and their corresponding applications,including the most relevant challenges faced by chip designers and end-users.Finally,we will provide an outlook on the future of this fascinating technology.
文摘This article studies numerically a familiar important phenomenon in spray combustion which is deformation and breakup of liquid drops in gas flow. The SIMPLER method is used to solve the two-dimensional （2D） unsteady axisymmetric Navier-Stokes equations for both the drop and the ambient gas flow. The level set method is applied to capturing the liquid/gas interface. Through calculation are obtained four typical breakup modes--oscillation, bag breakup, sheet stripping breakup and shear breakup governed by four non-dimensional numbers which are gas Weber number （Weg）, liquid Reynolds number （Rel）, gas Reynolds number （Reg） and density ratio （γ）. Their effects upon each mode are analyzed. The results indicate that among the four numbers, Weg is of the highest importance with Rel, Reg and γfollowing up. By widening the range of the density ratio up to 1 000, the breakup mode is discovered to be so complicated that a new one called multimode breakup mode turns up. This mode contains the shearing breakup and piercing breakup, which successively happen. The calculation results agree well with what is observed from the experiments.
文摘Since establishing a business in her rural hometown in 2017,Shi Jia,an ethnic Miao entrepreneur,has been committed to exploring ethnic resources to develop local industries.“Compared with other embroidery styles,the Miao embroidery focuses on vivid expression.At the beginning,it was used to record big events in place of writing,and distinguish tribes by different totems.Its color is brighter,and the design is simpler.There are also many auspicious patterns to pray for good luck,”Shi Jia,a deputy to the National People’s Congress(NPC)and president of a Miao clothing company,told China Today.
基金This work was performed under the auspices of the U.S.Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344was supported by the LLNL-LDRD Program under Project No.16-ERD-019 and 19-SI-001(LLNL-JRNL-764864).
文摘Despite ML’s impressive performance in commercial applications,several unique challenges exist when applying ML in materials science applications.In such a context,the contributions of this work are twofold.First,we identify common pitfalls of existing ML techniques when learning from underrepresented/imbalanced material data.Specifically,we show that with imbalanced data,standard methods for assessing quality of ML models break down and lead to misleading conclusions.Furthermore,we find that the model’s own confidence score cannot be trusted and model introspection methods(using simpler models)do not help as they result in loss of predictive performance(reliability-explainability trade-off).Second,to overcome these challenges,we propose a general-purpose explainable and reliable machine-learning framework.Specifically,we propose a generic pipeline that employs an ensemble of simpler models to reliably predict material properties.We also propose a transfer learning technique and show that the performance loss due to models’simplicity can be overcome by exploiting correlations among different material properties.A new evaluation metric and a trust score to better quantify the confidence in the predictions are also proposed.To improve the interpretability,we add a rationale generator component to our framework which provides both model-level and decision-level explanations.Finally,we demonstrate the versatility of our technique on two applications:(1)predicting properties of crystalline compounds and(2)identifying potentially stable solar cell materials.We also point to some outstanding issues yet to be resolved for a successful application of ML in material science.
基金supported by the U.S.Department of Energy(DOE),Office of Basic Energy Sciences,Division of Materials Sciences and Engineering under Award#DE-SC0023088.
文摘Nonlinear encoding of optical information can be achieved using various forms of data representation.Here,we analyze the performances of different nonlinear information encoding strategies that can be employed in diffractive optical processors based on linear materials and shed light on their utility and performance gaps compared to the state-of-the-art digital deep neural networks.For a comprehensive evaluation,we used different datasets to compare the statistical inference performance of simpler-to-implement nonlinear encoding strategies that involve,e.g.,phase encoding,against data repetition-based nonlinear encoding strategies.We show that data repetition within a diffractive volume(e.g.,through an optical cavity or cascaded introduction of the input data)causes the loss of the universal linear transformation capability of a diffractive optical processor.Therefore,data repetition-based diffractive blocks cannot provide optical analogs to fully connected or convolutional layers commonly employed in digital neural networks.However,they can still be effectively trained for specific inference tasks and achieve enhanced accuracy,benefiting from the nonlinear encoding of the input information.Our results also reveal that phase encoding of input information without data repetition provides a simpler nonlinear encoding strategy with comparable statistical inference accuracy to data repetition-based diffractive processors.Our analyses and conclusions would be of broad interest to explore the push-pull relationship between linear material-based diffractive optical systems and nonlinear encoding strategies in visual information processors.
文摘In the latest years,researchers from the industry and academia extensively applied machine learning algorithms in a broad range of domains.The goal of this special issue is to illustrate the most recent applications of deep learning methods in a range of real-life domains and to show the practical utility of these techniques.A particular attention goes towards methods to process network data that is capable of modelling complex artificial and natural systems as the interactions of a multitude of simpler entities.
基金supported by the National Natural Science Foundation of China(Grant No.42020104005)the Disciplinary Development Strategy Research Project of the Division of Earth Sciences,Chinese Academy of Sciences(Grant No.XK2020DXA004)。
文摘1.Elements and human history The rich chemistry of Earth's elements played an inherent and multifaceted role in shaping the course of human history.Generally,the more inert an element is,the simpler it becomes to extract it from minerals,facilitating its use in the crafting of tools,devices,and weapons.
文摘Editor’s Note:Peter Barnes offers a critique of the OECD’s Pillars One and Two proposals;he lauds the OECD’s goals but expresses concern that the proposals depend on unrealistic assumptions.David Rosenbloom,in a companion commentary,goes further and says the proposals are too complex to work in today’s international tax environment.Both authors1 believe the OECD’s goal of increasing source jurisdiction taxation can be achieved in other simpler ways.
文摘Objective:This study seeks to explore efficient and multiple-item detection modes in new-style HIV labs,as well as access the accuracy and reliability of CD4 cell count detected by point of care(POC)to analyze POC work feasibility in district or county labs.Methods:POC devices adopted in grassroots-level labs and flow cytometers adopted in prefecture-level labs were used to analyze the same group of blood samples.The individual results were collected and compared for parametric tests in correlation and consistency.Results:The Pearson correlation coefficients(r)between results detected by FACSPresto and those by FACSCalibur,FACSVia,FACSCantoII,and EPICSXL were 0.922,0.938,0.914,and 0.823,respectively;the average deviations were−25.64,24.68,3.05,and 70.97 cells/μL,respectively;the Pearson correlation coefficient(r)between results by Pima and FACSCalibur,FACSVia,FACSCantoII,and EPICSXL were 0.900,0.950,0.954,and 0.876,respectively;and the average deviations were−73.99,−40.78,−29.32,and−22.75 cells/μL,respectively.Discussion:Strong positive correlations and good consistency were observed between the CD4 count tested by POC and flow cytometers.These findings provide theoretical support for new-style HIV labs and one-stop services,which can provide shorter testing duration and simpler testing processes,so that the most comprehensive testing results can be obtained in the shortest amount of time.
文摘The flow instability through the side branch of a T-junction is analyzed in a numerical simulation. In a previous experimental study, the authors clarified the mechanism of fluid-induced vibration in the side branch of the T-junction in laminar steady flow through the trunk. However, in that approach there were restrictions with respect to extracting details of flow behavior such as the flow instability and the distribution of wall shear stress along the wall. Here the spatial growth of the velocity perturbation at the upstream boundary of the side branch is investigated. The simulation result indicates that a periodic velocity fluctuation introduced at the upstream boundary is amplified downstream, in good agreement with experimental result. The fluctuation in wall shear stress because of the flow instability shows local extrema in both the near and distal walls. From the numerical simulation, the downstream fluid oscillation under a typical condition has a Strouhal number of 1.05, which approximately agrees with the value obtained in experiments. Therefore, this periodic oscillation motion is a universal phenomenon in the side branch of a T-junction.
基金This work was supported by the National Key Project of Fundamental R & D of China (G2000026303).
文摘A fuzzy control method was used to accelerate iteration convergence in numerical fluid dynamic simulation using SIMPLER algorithm. The residual ratio of momentum or energy equation between two successive iterations was used as the input variable. A fuzzy logic algorithm was developed in order to obtain the relative increment of the under-relaxation factor and its new value was then used for the next iteration. The algorithm was tested by four benchmark problems. In all cases considered, when the fuzzy control logic was used, convergence was achieved with nearly the minimum number of iterations, showing the feasibility of the proposed method.
文摘Based on the first-order upwind and second-order central type of finite volume (UFV and CFV) scheme, upwind and central type of perturbation finite volume (UPFV and CPFV) schemes of the Navier-Stokes equations were developed. In PFV method, the mass fluxes of across the cell faces of the control volume (CV) were expanded into power series of the grid spacing and the coefficients of the power series were determined by means of the conservation equation itself. The UPFV and CPFV scheme respectively uses the same nodes and expressions as those of the normal first-order upwind and second-order central scheme, which is apt to programming. The results of numerical experiments about the flow in a lid-driven cavity and the problem of transport of a scalar quantity in a known velocity field show that compared to the first-order UFV and second-order CFV schemes, upwind PFV scheme is higher accuracy and resolution, especially better robustness. The numerical computation to flow in a lid-driven cavity shows that the under-relaxation factor can be arbitrarily selected ranging from (0.3) to (0.8) and convergence perform excellent with Reynolds number variation from 10~2 to 10~4.
基金This project was supported by the National Natural Science Foundation of China (No. 30371386) by a grant from the Natural Science Foundation of Guangdong Province (No. 31010).
文摘Objective： To investigate the therapeutic efficacy of adenovirus-mediated herpes simplex virus thymidine kinase （HSV-tk） gene transfer under the driving of KDR promoter （AdKDR-tk） in combination of ganciclovir （GCV） against human hepatocellular carcinoma in nude mice. Methods： HepG2 cell line was implanted subcutaneously into 32 nude mice, which were subsequently divided into 4 groups （n=8 each group）： Ganciclovir group （Ⅰ）, Ad group （Ⅱ）, AdCMV-tk/GCV group （under the driving of CMV promoter） （Ⅲ） and AdKDR-tk/GCV group （Ⅳ）. Then intratumoral injection of recombinant adenovirus or Ad was performed in all nude mice, and repeated 24 h later. For the following 10 d GCV was given at a dose of 100 mg/（kg· d）, ip. All the treated animals were killed to evaluate the tumor weight and the histopathological changes and the microvessel density of tumors after the treatment was determined. Results Compared with group Ⅰ, the tumor inhibitory rate was 12.3% in group Ⅲ and 24.5% in group Ⅳ; the inhibition rates were significantly different between group Ⅲand Ⅳ （P〈0.05）. The mean MVDs in group Ⅰ, Ⅱ, Ⅲ and Ⅳ were 37.4±8.6, 30.6±7.8, 27.6±7.1, and 10.7±4.1 （microvessels/mm^2）, respectively. Significant differences were found between group Ⅲ and Ⅱ （P〈0.05）, Ⅳ and Ⅱ （P〈0.01）, and Ⅳ and Ⅲ （P〈0.01）. Conclusion： Intratumoral injection of AdKDR-tk results in marked inhibition of HCC growth through inhibition angiogenesis in nude mice. It may be a new treatment approach for human HCC,
