Method for Rapidly Constructing Cardiac Coronary Vessel Identification Dataset

Abstract

The present invention provides a method for quickly constructing a heart coronary vessel recognition data set, comprising: obtaining an original picture; marking blood vessels in the original picture to form a rough-labeled picture; performing pixel-level labeling on a very small amount of original pictures according to the rough-labeled picture. Form a finely labeled picture; change the finely labeled picture from a three-channel image to a single-channel image; binarize the single-channel image and store it as a binarized picture; use the binarized picture and its corresponding original picture as training data, Train the initial network to obtain the first network; input all the original pictures into the above-mentioned first network to obtain the binarized result map; generate pseudo-fine annotations based on the binarized result map and its corresponding rough-labeled pictures Pictures: Create a data set based on original pictures, finely labeled pictures, and pseudo-finely labeled pictures to train the network. This method greatly reduces the cost of manual labeling while ensuring the quality of the data set, and the training speed can be significantly improved.

Description

technical field [0001] The invention relates to a method for constructing a cardiac vessel segmentation data set, in particular to a method capable of rapidly constructing a cardiac coronary vessel identification data set in a relatively short period of time and on the premise of reducing labor costs as much as possible. Background technique [0002] Artificial neural network is a computing model proposed in the field of artificial intelligence. It abstracts the human brain neuron network from the perspective of information processing, establishes a simple model, and forms different networks according to different connection methods to solve the problem of artificial intelligence. Certain problems in domains such as image recognition. Convolutional Neural Network (CNN) is a kind of artificial neural network. It is a feedforward neural network. Artificial neurons can respond to surrounding units and can perform large-scale image processing, that is, input a picture, and convo...

AI Technical Summary

Problems solved by technology

Segmentation networks trained with general-purpose image datasets cannot be used for identification and segmentation of coronary arteries

[0006] The patent application No. 201810441538.4 provides a method for establishing a convolutional neural network data set for training and identifying cardiovascular types. In addition to requiring doctors with professional knowledge to roughly label the types and positions of blood vessels in pictures, it also People who need to master the labeling technology perform pixel-level labeling of the blood vessel category and position in all pictures, which consumes a high labor cost

Images