A tumor detection method and device based on fusion of bm3d and dense convolutional network

Abstract

The invention provides a tumor detection method and device based on BM3D and dense convolutional network fusion. The method comprises the steps of marking the similar blocks; randomly discarding and quickly marking; optimizing and training the DenseNet network; and reconstructing according to the spatial information and structural information of the input image and the extracted feature information based on the depth training data; abstracting the input spatial information into the one-dimensional information, so that the irreversible initial feature loss phenomenon is reduced; constructing adense convolutional network fused with BM3D; using a scalable exponential type linear unit activation function for replacing a linear unsaturated unit activation function activation network, introducing the negative part parameters, improving the network optimization degree, enhancing the network robustness, adding a layer of maximum pooling layer abstract image feature behind each dense block, and extracting the tumor core information points. The network tail end adopts a BM3D aggregation method to perform the feature reconstruction, and the gradient and the spatial information are fused to improve the network effect, so that the accuracy of tumor detection is effectively improved.

Description

technical field [0001] The present invention relates to the technical field of medical image processing, in particular to a tumor detection method and device based on fusion of BM3D and dense convolutional network. Background technique [0002] Medical images come from imaging techniques, including computed tomography, magnetic resonance imaging, ultrasound, positron emission tomography, medical ultrasound examination, etc. Medical imaging technology can obtain two-dimensional or three-dimensional images of the corresponding positions of the human body. In a two-dimensional image, the smallest unit element representing specific information is called a pixel; in a three-dimensional image it is called a voxel. Under certain conditions, three-dimensional images can be represented as a series of two-dimensional images, which greatly reduces computational complexity and memory requirements. However, although the medical imaging technology has become more and more mature, the re...

AI Technical Summary

Problems solved by technology

[0005] However, when the traditional CNN or FCN transmits information, there will be more or less problems such as information loss and loss. After the input information or gradient information passes through multiple layers, it is likely to cause gradient disappearance or gradient explosion.

Images