<strong>Background:</strong> High-resolution medical images often need to be downsampled because of the memory limitations of the hardware used for machine learning. Although various image interpolation me...<strong>Background:</strong> High-resolution medical images often need to be downsampled because of the memory limitations of the hardware used for machine learning. Although various image interpolation methods are applicable to downsampling, the effect of data preprocessing on the learning performance of convolutional neural networks (CNNs) has not been fully investigated. <strong>Methods:</strong> In this study, five different pixel interpolation algorithms (nearest neighbor, bilinear, Hamming window, bicubic, and Lanczos interpolation) were used for image downsampling to investigate their effects on the prediction accuracy of a CNN. Chest X-ray images from the NIH public dataset were examined by downsampling 10 patterns. <strong>Results:</strong> The accuracy improved with a decreasing image size, and the best accuracy was achieved at 64 × 64 pixels. Among the interpolation methods, bicubic interpolation obtained the highest accuracy, followed by the Hamming window.展开更多
Owing to its property of applying multi-modality imaging information into the clinical usage has the Medical Image Registration been the research focus. The gray nearest neighbor interpolation and bilinear interpolati...Owing to its property of applying multi-modality imaging information into the clinical usage has the Medical Image Registration been the research focus. The gray nearest neighbor interpolation and bilinear interpolation and cubic convolution interpolation method used to medical image interpolation algorithm. It compares the characteristics of the three gray interpolations. Combined with the characteristics of the above algorithm is proposed based on gray-scale pixel intensity interpolation. The algorithm can be improved in the time and registration accuracy. It combined with the improved optimization algorithm in the proposed image registration of the simulation experiment, experimental precision subpixel image to verify the validity of the method.展开更多
Image interpolation of cross-sections is one of the key steps of medical visualization, and the cubic convolution interpolation is usually employed due to its good tradeoff between computational cost and accuracy, how...Image interpolation of cross-sections is one of the key steps of medical visualization, and the cubic convolution interpolation is usually employed due to its good tradeoff between computational cost and accuracy, however, sometimes its accuracy can still not meet the requirement. Aimed at the problem, in this paper, the interpolation principle based cubic convolution is firstly analyzed systematically, and then essential relationship among the different cubic convolution interpolation methods is clarified. Lastly, a novel cross-section interpolation method for medical images that is based on the optimal parameter of sharp control is presented. The method takes full advantage of the local characteristic of medical images, and the optimized sharp control parameter is obtained by the iterative computation, and then the cross-section interpolation is performed by the cubic convolution with the optimized parameter in one time.The experimental results show that the method presented in the paper not only can improve the interpolation accuracy effectively, but also is robust.展开更多
对于医学序列切片,需要图像分割技术将图像中感兴趣区域ROI(region of interest)提取并进行三维重建.提出一种基于距离场的非线性图像插值分割方法,该方法克服了软组织因灰度值分布不均匀难分割的问题,可以快速提取ROI区域.实验表明用...对于医学序列切片,需要图像分割技术将图像中感兴趣区域ROI(region of interest)提取并进行三维重建.提出一种基于距离场的非线性图像插值分割方法,该方法克服了软组织因灰度值分布不均匀难分割的问题,可以快速提取ROI区域.实验表明用该方法进行医学序列切片的分割,可以得到良好的分割结果,并已成功地应用于医学数据的三维重建系统中.展开更多
文摘<strong>Background:</strong> High-resolution medical images often need to be downsampled because of the memory limitations of the hardware used for machine learning. Although various image interpolation methods are applicable to downsampling, the effect of data preprocessing on the learning performance of convolutional neural networks (CNNs) has not been fully investigated. <strong>Methods:</strong> In this study, five different pixel interpolation algorithms (nearest neighbor, bilinear, Hamming window, bicubic, and Lanczos interpolation) were used for image downsampling to investigate their effects on the prediction accuracy of a CNN. Chest X-ray images from the NIH public dataset were examined by downsampling 10 patterns. <strong>Results:</strong> The accuracy improved with a decreasing image size, and the best accuracy was achieved at 64 × 64 pixels. Among the interpolation methods, bicubic interpolation obtained the highest accuracy, followed by the Hamming window.
文摘Owing to its property of applying multi-modality imaging information into the clinical usage has the Medical Image Registration been the research focus. The gray nearest neighbor interpolation and bilinear interpolation and cubic convolution interpolation method used to medical image interpolation algorithm. It compares the characteristics of the three gray interpolations. Combined with the characteristics of the above algorithm is proposed based on gray-scale pixel intensity interpolation. The algorithm can be improved in the time and registration accuracy. It combined with the improved optimization algorithm in the proposed image registration of the simulation experiment, experimental precision subpixel image to verify the validity of the method.
文摘Image interpolation of cross-sections is one of the key steps of medical visualization, and the cubic convolution interpolation is usually employed due to its good tradeoff between computational cost and accuracy, however, sometimes its accuracy can still not meet the requirement. Aimed at the problem, in this paper, the interpolation principle based cubic convolution is firstly analyzed systematically, and then essential relationship among the different cubic convolution interpolation methods is clarified. Lastly, a novel cross-section interpolation method for medical images that is based on the optimal parameter of sharp control is presented. The method takes full advantage of the local characteristic of medical images, and the optimized sharp control parameter is obtained by the iterative computation, and then the cross-section interpolation is performed by the cubic convolution with the optimized parameter in one time.The experimental results show that the method presented in the paper not only can improve the interpolation accuracy effectively, but also is robust.
文摘对于医学序列切片,需要图像分割技术将图像中感兴趣区域ROI(region of interest)提取并进行三维重建.提出一种基于距离场的非线性图像插值分割方法,该方法克服了软组织因灰度值分布不均匀难分割的问题,可以快速提取ROI区域.实验表明用该方法进行医学序列切片的分割,可以得到良好的分割结果,并已成功地应用于医学数据的三维重建系统中.
