Cardiovascular vulnerable plaque recognition method and system based on attention model and multi-task neural network

Abstract

The invention discloses a cardiovascular vulnerable plaque identification method and system based on an attention model and a multi-task neural network. The method includes: 1. Eliminating noise in the original polar coordinate image based on a top-down attention model; 2. .Multi-task neural network is used to classify and segment the vulnerable plaque images in the preprocessed images; 3. To perform regional refinement on the classified and segmented vulnerable plaque images; the system includes sequentially connected top-down based The attention model removes the noise in the original polar coordinate image to obtain the sub-system of the pre-processed image, uses the multi-task neural network to classify and segment the vulnerable plaque in the pre-processed image Subsystem for region refinement of vulnerable plaque images. The noise interference of blood vessels on the subsequent identification of vulnerable plaques is eliminated, making the location of vulnerable plaques more accurate.

Description

technical field [0001] The invention belongs to the field of medical image processing, and relates to a cardiovascular vulnerable plaque recognition method and system based on an attention model and a multi-task neural network. Background technique [0002] Vulnerable plaque is the most dangerous plaque in coronary atherosclerotic lesions. Vulnerable plaque is the main cause of thrombosis, acute coronary syndrome, and even sudden death. Therefore, detection and identification of various plaques Vulnerable plaque has a very high value. Cardiovascular optical coherence tomography is an intravascular imaging technology using near-infrared light reflection imaging, which can clearly observe vulnerable plaques. Therefore, the identification of vulnerable plaques based on optical coherence tomography (OCT) has become a important research trends. [0003] Commonly used performance evaluation criteria for OCT vulnerable plaques include: the recall rate R of vulnerable plaque detec...

AI Technical Summary

Problems solved by technology

[0005] definition Set the initial value of nTP, nFP, nFN to 0; sequentially judge the detection area from 1 to M, if any B i , B j All are judged as errors and do not participate in the following calculations; if B i with any A k If there is an intersection, and the DSC value is greater than 0.5, it is considered that the target i is detected correctly, Ntp++, and the DSC value is less than or equal to 0.5, it is considered that the target i is detected incorrectly, nFP++; sequentially judge the real detection target area from 1 to N; if A k with any B i If there is no intersection, it is considered that the target is judged as missed detection, nFN++; recall rate R: R=nTP / (nTP+nFN); accuracy rate P: P=nTP / (nTP+nFP); coincidence degree D is all detected correct The mean value of the DSC value of the area and its corresponding real area, The recall rate, accuracy rate and coincidence degree of the existing OCT image recognition method are not high, and the precision is low, which cannot meet the actual needs

Images