Abstract: To improve the performance and accuracy of an image segmentation neural network, a cascaded robust learning framework for the segmentation of noisy labeled images includes two stages: a sample selection stage, and a joint optimization stage with label correction. In the first stage, the clean annotated samples are selected for network updating, so that the influence of noisy sample can be interactively eliminated. In the second stage, the label correction module works together with the joint optimization scheme to revise the imperfect labels. Thus, the training of the whole network is supervised by the corrected labels and the original ones.

Abstract: To improve the performance and accuracy of an image segmentation neural network, a cascaded robust learning framework for the segmentation of noisy labeled images includes two stages: a sample selection stage, and a joint optimization stage with label correction. In the first stage, the clean annotated samples are selected for network updating, so that the influence of noisy sample can be interactively eliminated. In the second stage, the label correction module works together with the joint optimization scheme to revise the imperfect labels. Thus, the training of the whole network is supervised by the corrected labels and the original ones.

Abstract: Disclosed are an image processing method, an image processing apparatus, and a readable storage medium. First, an image to be processed is received, and the received image to be processed is divided into regions of interest by region segmentation means. Next, the regions of interest are detected by calling a pre-stored full convolution network structure model, to obtain probability image segments. Finally, the probability image segments are synthesized to generate a target probability image. Wherein, the pre-stored full convolution network structure model includes a full convolution structure. A linear regression layer is replaced by an equivalent convolution layer in the full convolution structure. A blank padding operation layer and an up-sampling layer are removed from the full convolution structure.

Abstract: Disclosed are a method and a device for detecting pulmonary nodule in Computed Tomography (CT) image, as well as a computer-readable storage medium. The method for detecting pulmonary nodule in CT image includes: obtaining a CT image to be detected, performing a pixel segmentation processing on the CT image through a pre-stored three-dimensional convolutional neural pixel segmentation network, to obtain a probability graph corresponding to the CT image, and obtaining a candidate nodule region by marking a connected domain on the probability graph; and predicting the candidate nodule region by various pre-stored prediction models corresponding to different three-dimensional convolutional neural network classifiers, to obtain various probability prediction values of the candidate nodule region, and comprehensively processing the various probability prediction values to obtain a classification result of the candidate nodule region.

Abstract: Disclosed are a method and a device for detecting pulmonary nodule in Computed Tomography (CT) image, as well as a computer-readable storage medium. The method for detecting pulmonary nodule in CT image includes: obtaining a CT image to be detected, performing a pixel segmentation processing on the CT image through a pre-stored three-dimensional convolutional neural pixel segmentation network, to obtain a probability graph corresponding to the CT image, and obtaining a candidate nodule region by marking a connected domain on the probability graph; and predicting the candidate nodule region by various pre-stored prediction models corresponding to different three-dimensional convolutional neural network classifiers, to obtain various probability prediction values of the candidate nodule region, and comprehensively processing the various probability prediction values to obtain a classification result of the candidate nodule region.

Abstract: Disclosed are an image processing method, an image processing apparatus, and a readable storage medium. First, an image to be processed is received, and the received image to be processed is divided into regions of interest by region segmentation means. Next, the regions of interest are detected by calling a pre-stored full convolution network structure model, to obtain probability image segments. Finally, the probability image segments are synthesized to generate a target probability image. Wherein, the pre-stored full convolution network structure model includes a full convolution structure. A linear regression layer is replaced by an equivalent convolution layer in the full convolution structure. A blank padding operation layer and an up-sampling layer are removed from the full convolution structure.