Unlock instant, AI-driven research and patent intelligence for your innovation.

Blood vessel plaque CT image segmentation method based on position convolution attention network

A technology of vascular plaque and CT imaging, applied in the field of medical imaging, to achieve the effect of fine information, rapid screening and labeling

Active Publication Date: 2021-08-03
SHANDONG ARTIFICIAL INTELLIGENCE INST +1
View PDF5 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] In order to overcome the deficiencies of the above technologies, the present invention provides a CT image segmentation method for quickly screening and labeling vascular plaques without manual intervention

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Blood vessel plaque CT image segmentation method based on position convolution attention network
  • Blood vessel plaque CT image segmentation method based on position convolution attention network
  • Blood vessel plaque CT image segmentation method based on position convolution attention network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0046] Step a) includes the following steps:

[0047] a-1) By formula Using the z-score normalization method to calculate the area of ​​vascular plaque in the normalized CT image where x original is the input vascular plaque CT sample, μ is the mean of the batch data, σ is the variance of the batch data, π is a given constant, and the denominator of the formula is 0;

[0048] a-2) In the image obtained by the computed tomography scan, the involuntary movement of the human body will cause artifacts in the detection result, and for this reason, noise reduction processing is performed on the CT image of the vascular plaque. Due to the difference of CT detection equipment, the image size will be different, so the normalized image is set. For m rows and n columns, through the formula convert the image Represented as a two-dimensional array, through the formula F(x,y)=median x,y∈around(x,y) [f(x,y)] applies median filtering to the image Perform noise reduction to obtain...

Embodiment 2

[0051] In step c), the image D is processed by a two-dimensional convolution layer and then processed using batch normalization and a sigmoid activation function.

Embodiment 3

[0053] Step d) through the formula Calculate the feature map D 4 , D 4 ∈R Q×Q , where α is the scaling factor, T is the matrix transpose, is the feature map D 1 The ith pixel in , is the feature map D 2 The jth pixel in , by formula calculate and The degree of location correlation and dependence of , the greater the degree of correlation between them, the more similar the feature representations of the two locations.

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a blood vessel plaque CT image segmentation method based on a position convolution attention network. The method comprises the following steps: sequentially carrying out pre-processing operations of normalization, denoising and enhancement on an originally collected image; secondly, inputting the preprocessed image into a position convolution attention network for segmentation; the position convolution attention network comprises two modules: a first module is a position attention module which captures the position correlation degree and the dependency degree between image pixels by using a self-attention mechanism so as to generate an aggregated image feature map, thereby forming a preliminary segmentation region through extremely high similarity shown by pixel points with high correlation degree; a second module is of a V-Net structure, down-sampling, up-sampling and jump connection structures are utilized to generate a recovery feature graph, and the recovery feature graph fuses more pixel features, so that information such as the edge of a segmented graph is finer; and finally, carrying out pixel point fusion on the aggregated image feature map output by the position convolution attention network and the recovered feature map to obtain a target segmentation image.

Description

technical field [0001] The invention relates to the technical field of medical imaging, in particular to a CT image segmentation method of blood vessel plaques based on a position convolution attention network. Background technique [0002] The current mainstream vascular plaque screening technology is to use CT technology to form images, which are then analyzed manually. But the level of manual analysis is often limited. The previous medical image segmentation is mainly based on the traditional image segmentation method. The method is to simply use the texture, shape, grayscale and other characteristics of the image to segment the image into several disjoint regions, and there is no objective judgment standard for the segmentation performance. . With the development of artificial intelligence technology, the automatic image analysis technology using deep learning can greatly improve the analysis time and efficiency on the basis of replacing manpower. Therefore, the resear...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06T7/00G06T7/12G06T5/00G06T5/50G06N3/04G06N3/08
CPCG06T7/0012G06T7/12G06T5/50G06N3/08G06T2207/10081G06T2207/20032G06T2207/20021G06T2207/20081G06T2207/20084G06T2207/30101G06T2207/20221G06N3/045G06T5/70
Inventor 王英龙徐鹏摇舒明雷周书旺
Owner SHANDONG ARTIFICIAL INTELLIGENCE INST