A near-field three-dimensional(3 D)imaging method combining multichannel joint sparse recovery(MJSR)and fast Gaussian gridding nonuniform fast Fourier transform(FGGNUFFT)is proposed,based on a perfect combination of t...A near-field three-dimensional(3 D)imaging method combining multichannel joint sparse recovery(MJSR)and fast Gaussian gridding nonuniform fast Fourier transform(FGGNUFFT)is proposed,based on a perfect combination of the compressed sensing(CS)theory and the matched filtering(MF)technique.The approach has the advantages of high precision and high efficiency:multichannel joint sparse constraint is adopted to improve the problem that the images recovered by the single channel imaging algorithms do not necessarily share the same positions of the scattering centers;the CS dictionary is constructed by combining MF and FGG-NUFFT,so as to improve the imaging efficiency and memory requirement.Firstly,a near-field 3 D imaging model of joint sparse recovery is constructed by combining the MF-based imaging method.Secondly,FGG-NUFFT and reverse FGG-NUFFT are used to replace the interpolation and Fourier transform in MF-based imaging methods,and a sensing matrix with high precision and high efficiency is constructed according to the traditional imaging process.Thirdly,a fast imaging recovery is performed by using the improved separable surrogate functionals(SSF)optimization algorithm,only with matrix and vector multiplication.Finally,a 3 D imagery of the near-field target is obtained by using both the horizontal and the pitching interferometric phase information.This paper contains two imaging models,the only difference is the sub-aperture method used in inverse synthetic aperture radar(ISAR)imaging.Compared to traditional CS-based imaging methods,the proposed method includes both forward transform and inverse transform in each iteration,which improves the quality of reconstruction.The experimental results show that,the proposed method improves the imaging accuracy by about O(10),accelerates the imaging speed by five times and reduces the memory usage by about O(10~2).展开更多
Much of the published literature in Radiology-related Artificial Intelligence(AI)focuses on single tasks,such as identifying the presence or absence or severity of specific lesions.Progress comparable to that achieved...Much of the published literature in Radiology-related Artificial Intelligence(AI)focuses on single tasks,such as identifying the presence or absence or severity of specific lesions.Progress comparable to that achieved for general-purpose computer vision has been hampered by the unavailability of large and diverse radiology datasets containing different types of lesions with possibly multiple kinds of abnormalities in the same image.Also,since a diagnosis is rarely achieved through an image alone,radiology AI must be able to employ diverse strategies that consider all available evidence,not just imaging information.Using key imaging and clinical signs will help improve their accuracy and utility tremendously.Employing strategies that consider all available evidence will be a formidable task;we believe that the combination of human and computer intelligence will be superior to either one alone.Further,unless an AI application is explainable,radiologists will not trust it to be either reliable or bias-free;we discuss some approaches aimed at providing better explanations,as well as regulatory concerns regarding explainability(“transparency”).Finally,we look at federated learning,which allows pooling data from multiple locales while maintaining data privacy to create more generalizable and reliable models,and quantum computing,still prototypical but potentially revolutionary in its computing impact.展开更多
基金supported by the National Natural Science Foundation of China(61771369 61775219+5 种基金 61640422)the Fundamental Research Funds for the Central Universities(JB180310)the Equipment Research Program of the Chinese Academy of Sciences(YJKYYQ20180039)the Shaanxi Provincial Key R&D Program(2018SF-409 2018ZDXM-SF-027)the Natural Science Basic Research Plan
文摘A near-field three-dimensional(3 D)imaging method combining multichannel joint sparse recovery(MJSR)and fast Gaussian gridding nonuniform fast Fourier transform(FGGNUFFT)is proposed,based on a perfect combination of the compressed sensing(CS)theory and the matched filtering(MF)technique.The approach has the advantages of high precision and high efficiency:multichannel joint sparse constraint is adopted to improve the problem that the images recovered by the single channel imaging algorithms do not necessarily share the same positions of the scattering centers;the CS dictionary is constructed by combining MF and FGG-NUFFT,so as to improve the imaging efficiency and memory requirement.Firstly,a near-field 3 D imaging model of joint sparse recovery is constructed by combining the MF-based imaging method.Secondly,FGG-NUFFT and reverse FGG-NUFFT are used to replace the interpolation and Fourier transform in MF-based imaging methods,and a sensing matrix with high precision and high efficiency is constructed according to the traditional imaging process.Thirdly,a fast imaging recovery is performed by using the improved separable surrogate functionals(SSF)optimization algorithm,only with matrix and vector multiplication.Finally,a 3 D imagery of the near-field target is obtained by using both the horizontal and the pitching interferometric phase information.This paper contains two imaging models,the only difference is the sub-aperture method used in inverse synthetic aperture radar(ISAR)imaging.Compared to traditional CS-based imaging methods,the proposed method includes both forward transform and inverse transform in each iteration,which improves the quality of reconstruction.The experimental results show that,the proposed method improves the imaging accuracy by about O(10),accelerates the imaging speed by five times and reduces the memory usage by about O(10~2).
文摘Much of the published literature in Radiology-related Artificial Intelligence(AI)focuses on single tasks,such as identifying the presence or absence or severity of specific lesions.Progress comparable to that achieved for general-purpose computer vision has been hampered by the unavailability of large and diverse radiology datasets containing different types of lesions with possibly multiple kinds of abnormalities in the same image.Also,since a diagnosis is rarely achieved through an image alone,radiology AI must be able to employ diverse strategies that consider all available evidence,not just imaging information.Using key imaging and clinical signs will help improve their accuracy and utility tremendously.Employing strategies that consider all available evidence will be a formidable task;we believe that the combination of human and computer intelligence will be superior to either one alone.Further,unless an AI application is explainable,radiologists will not trust it to be either reliable or bias-free;we discuss some approaches aimed at providing better explanations,as well as regulatory concerns regarding explainability(“transparency”).Finally,we look at federated learning,which allows pooling data from multiple locales while maintaining data privacy to create more generalizable and reliable models,and quantum computing,still prototypical but potentially revolutionary in its computing impact.
