Coronary artery stenosis lesion degree identification method based on multi-classifier fusion

Abstract

The invention discloses a coronary artery stenosis lesion degree identification method based on multi-classifier fusion. The method comprises the following steps: firstly, constructing an image sample library; preprocessing a CT original sequence diagram extracted by a heart CTA, and then performing feature extraction to extract three types of features including interested texture features, gray features and geometric features; dividing the samples into a training group and a test group, calculating the correlation between each feature and a prediction result, and removing the features with small correlation; establishing a multi-classifier fusion prediction model, fusing results of the single classifiers to predict the coronary artery stenosis lesion degree, meanwhile, determining the weights of the three single classifiers in the fusion classifier through a weighting method, and when the stenosis degree is lower than 50%, judging that the sample a normal sample, and when the stenosis degree is larger than 50%, judging that the sample is a lesion sample. According to the method, automatic classification and pre-judgment on the aspect of judging the stenosis degree are realized, and the injury to a patient caused by an invasive operation is avoided.

Description

technical field [0001] The invention belongs to the technical field of bioengineering, and in particular relates to a method for identifying the lesion degree of coronary artery stenosis based on multi-classifier fusion. Background technique [0002] In recent years, the incidence and fatality rate of cardiovascular and cerebrovascular diseases have jumped to the top of all kinds of diseases, especially the coronary arteries are located on the surface of the heart and supply blood to the myocardium. Once a disease occurs, more attention should be paid to its examination and treatment. Coronary artery disease can be diagnosed by anatomical parameters (such as diameter stenosis) or functional parameters related to coronary myocardial ischemia. The clinical diagnosis of coronary heart disease and the determination of the treatment plan usually require the patient to undergo a cardiac CTA examination first. The doctor can give a preliminary diagnosis based on the CT image, but o...

AI Technical Summary

Problems solved by technology

[0004] The fluid dynamics simulation modeling process is complicated: it is necessary to establish a patient-specific coronary artery model for hemodynamic simulation, single data and single model, and the human blood flow operation is complex, some factors may not have been considered, such as transient simulation , although the pressure and velocity waveforms can be obtained from the literature, in order to perform more accurate individual case simulations, in vitro or in vivo measurements are required, and the modeling process requires image processing and reconstruction, which requires higher processors and increases The complexity of the operation is increased, and the calculation time is correspondingly extended, ranging from several hours

[0005] The effect of deep learning algorithm is easy to overfit: medical real data is generally difficult to obtain, and usually belongs to the category of small sample learning. Due to the complex model structure of deep neural network, a large number of image samples are required for training

However, this kind of algorithm with strong expressive ability focuses on explaining the training data, and it is easy to sacrifice the ability to explain the future data, that is, the test data. In order to avoid overfitting, more data samples are often required for learning to ensure that it is on the new data set. It can still achieve better results, which is not applicable when the image sample data is small

Images