Near-infrared subcutaneous vein segmentation method based on multi-feature clustering

Abstract

The present invention designs a subcutaneous vein vessel segmentation method based on near-infrared imaging, which can realize multi-feature extraction and automatic clustering of vein vessels. First, use NiBlack and morphological algorithms to realize skin region segmentation and edge mirror extension; second, through multi-scale IUWT and Hessian matrix analysis to obtain blood vessel similarity images, blood vessel direction maps, blood vessel scale maps, and initially segmented blood vessels; The third step is to extract and repair the centerline of the branch of the vessel by using the initial segmented vessel and the direction map of the vessel, and the position and direction of the branch centerline are corrected by the method of piecewise spline fitting; the fourth step is to calculate the original image to The coordinate mapping relationship of the branch contour image, after mapping the IUWT enhanced image and the blood vessel similarity image to the contour image space respectively, extracts the normalized second-order Gaussian feature and the blood vessel similarity feature; the fifth step is to use the obtained blood vessel feature The K-means algorithm was used to cluster the contour images into three categories: skin, blood vessels and fuzzy regions.

Description

technical field [0001] The invention relates to a subcutaneous vein segmentation method, in particular to a near-infrared subcutaneous vein segmentation method based on multi-feature clustering, which is mainly used in the fields of subcutaneous intravenous injection, vein identification and the like. Background technique [0002] With the continuous research of researchers on invisible spectral imaging technology and spectral imaging characteristics of human tissue structure, infrared spectrum has shown excellent enhancement effect in human tissue imaging, especially in subcutaneous vein imaging. Infrared vein imaging is safer and more convenient than X-ray and ultrasound imaging. The vein enhancement of infrared imaging shows that it is essentially due to the difference in spectral response between blood vessels and skin, which makes it still have stable vein enhancement when it is applied to special populations such as children, the elderly, trauma patients, obese patient...

AI Technical Summary

Problems solved by technology

However, due to the generally poor quality of near-infrared vein images, existing vein segmentation methods often have limitations in the following aspects:

[0005] 1. Due to the unavoidable environmental factors, the resulting non-uniformity of image illumination seriously affects the contrast of veins in different areas, which greatly enhances the difficulty of enhancement, measurement and segmentation of veins in shadow areas;

[0006] 2. Due to the poor imaging quality of near-infrared images, threshold-based segmentation methods based on image grayscale information are often difficult to obtain smooth blood vessel edges. There are also a lot of noise in the segmentation results, and the ability to segment details is poor.

Simply using morphological algorithms to solve the above two types of problems will seriously affect the accuracy of segmentation;

[0007] 3. Also due to the image quality, the connection area in the blood vessel is often difficult to segment completely, and the blood vessel branches are prone to breakage, which also increases the difficulty of noise filtering

Images