Deep network segmentation method for coronary artery lumen contour under OCT image

Abstract

The invention provides a deep network segmentation method for a coronary artery lumen contour under an OCT image. The segmentation method employs a double-branch deep convolutional network structure, one branch segments a lumen region mask corresponding to the OCT image, the other branch predicts a lumen inner wall contour, the two tasks are subjected to combined learning, and a coupled result is the final segmented contour. According to the invention, the double-branch deep convolutional network is adopted, the correlation between the two tasks is used for joint learning, the results of the two branches are fused to obtain the final segmentation contour, robustness can be kept for the shape change of the inner wall of the lumen, then the contour of the inner wall of the blood vessel is accurately positioned, and accurate segmentation of the contour of the coronary artery lumen under the OCT image is achieved.

Description

technical field [0001] The invention belongs to the technical field of data information processing, and in particular relates to an automatic segmentation method for the contour of a coronary artery lumen under an OCT image. Background technique [0002] With the intensification of the aging process in our country, the incidence of cardiovascular disease is increasing. The current diagnosis and treatment methods can only observe the outline of the lesion through contrast imaging, which is not enough to meet the needs of precise location of the lesion. Optical Coherence Tomography (OCT) is another major technological breakthrough following X-CT and MRI in recent years. The principle of OCT imaging is similar to that of ultrasound, which uses reflected near-infrared rays as the imaging medium to form images. Optical coherence tomography is widely used in high-resolution tomographic imaging of the interior of coronary artery lumens to obtain high-resolution images of the inne...

AI Technical Summary

Problems solved by technology

[0004] Watershed, Level-set, etc. are commonly used image segmentation algorithms, but directly applying these methods to OCT image segmentation problems cannot obtain good segmentation results. Due to the influence of imaging probes, vascular bifurcations, and vascular deformation, such methods It cannot segment the complete contour of the inner wall of the blood vessel, and there are problems such as the contour is not closed after segmentation and the positioning accuracy is low, which cannot meet the clinical high-precision segmentation requirements

Images