CN-122023809-A - Intelligent image recognition system based on digestive system department images

CN122023809ACN 122023809 ACN122023809 ACN 122023809ACN-122023809-A

Abstract

The invention relates to the technical field of image processing, in particular to an intelligent image recognition system based on an internal digestive system image. The technical scheme includes that the method comprises the steps of obtaining a first image of a user intestinal canal in a white light mode under different light source incidence angles and a second image corresponding to a narrow-band imaging mode, identifying a three-dimensional convex-concave structure of the inner wall of the intestinal canal according to gray level differences of the first image under different light source incidence angles to locate a three-dimensional change area, screening out pixel points with blood vessel features in the second image, marking the blood vessel pixel points at corresponding positions of the first image, conducting texture analysis on the blood vessel pixel points of the three-dimensional change area to obtain texture distribution coefficients of each three-dimensional change area, and conducting area division on the first image according to the texture distribution coefficients and optical flow features generated by each three-dimensional change area in adjacent image frames to improve accuracy of intelligent identification of the image area.

Inventors

HUANG YING
LIU CHENYU
YE YUANYUAN
SONG JIAHONG
WANG WEI

Assignees

西安葆康医管数据科技有限公司
民航上海医院

Dates

Publication Date: 20260512
Application Date: 20260408

Claims (8)

1. An image intelligent recognition system based on gastroenterology image, which is characterized in that the system comprises: the acquisition terminal is used for acquiring a first image of the intestinal tract of a user in a white light mode under different light source incidence angles and a second image corresponding to the narrow-band imaging mode, and outputting the first image and the second image to the outside; The processing terminal is connected with the acquisition terminal and is used for obtaining a three-dimensional change area of which the inner wall of the intestinal canal presents a convex-concave structure according to the gray scale difference of the first image at different light source incidence angles; Screening out pixel points with blood vessel characteristics in the second image, and marking the corresponding position of the first image as the blood vessel pixel point; Performing texture feature analysis on the vascular pixel points of the three-dimensional change areas to obtain texture distribution coefficients of each three-dimensional change area; And dividing the first image into areas according to the texture distribution coefficient and the optical flow characteristics generated by each three-dimensional change area in the adjacent image frames.
2. The intelligent image recognition system based on the digestive tract images according to claim 1, wherein the obtaining of the three-dimensional change region of the inner wall of the intestinal tract in the convex-concave structure according to the gray scale difference of the first image at different light source incidence angles comprises: Performing position alignment processing and morphological opening operation on a plurality of first images of the same intestinal region under different light source incidence angles so as to obtain an image region group corresponding to the intestinal region; And carrying out three-dimensional convex-concave characteristic analysis according to the gray level change of the pixel points of each first image in the image region group so as to obtain a three-dimensional change region with the inner wall of the intestinal canal showing a convex-concave structure.
3. The intelligent digestive-system-image-based image recognition system according to claim 2, wherein the performing three-dimensional convex-concave feature analysis according to the pixel gray level change of each first image in the image region group to obtain a three-dimensional change region in which the inner wall of the intestinal tract presents a convex-concave structure comprises: Carrying out gray value difference and absolute value calculation according to pixel points of the same position of each first image in the image area group so as to obtain the difference absolute value of each pixel point; Dividing all absolute values of the difference values by using an Ojin threshold value, and determining pixel points larger than a corresponding threshold value as changed pixel points; Marking the pixels with the pixel difference value larger than a preset value corresponding to each first image of the changed pixels as first type pixels, and marking the pixels with the pixel difference value smaller than or equal to the preset value as second type pixels; and according to the distribution characteristics represented by the first type pixel points and the second type pixel points, obtaining a three-dimensional change area of which the inner wall of the intestinal canal presents a convex-concave structure.
4. The intelligent image recognition system based on the digestive tract images according to claim 3, wherein the obtaining the three-dimensional change area of the inner wall of the intestinal tract in the convex-concave structure according to the distribution characteristics characterized by the first type of pixels and the second type of pixels comprises: Performing morphological open operation and morphological close operation on the first type pixel points and the second type pixel points respectively to obtain a first cluster formed by the first type pixel points and a second cluster formed by the second type pixel points; And comparing the adjacent positions of the first cluster and the second cluster, and if the current first cluster and the current second cluster are the nearest clusters, determining the image area covered by the current first cluster and the current second cluster as the three-dimensional change area.
5. The intelligent image recognition system based on the gastroenterology image according to claim 1, wherein the screening out the pixel points with the blood vessel features in the second image and marking the corresponding position of the first image as the blood vessel pixel point includes: performing position alignment processing on a first image and a second image of the same intestinal tract region, and inputting the second image into an HSV model to obtain hue values of all pixel points in the second image; performing blood vessel feature analysis according to the hue value and the color saturation of each pixel point in the second image to determine a target pixel point which is characterized as a blood vessel feature in the second image; And determining the pixel point of the target pixel point corresponding to the first image as the blood vessel pixel point.
6. The intelligent image recognition system based on the gastroenterology image according to claim 5, wherein the performing the blood vessel feature analysis according to the hue value and the color saturation of each pixel point in the second image to determine the target pixel point characterized as the blood vessel feature in the second image includes: Determining a pixel point with a hue value in a preset interval in the second image as a first screening pixel point, and determining a pixel point with a color saturation greater than a saturation threshold in the second image as a second screening pixel point; and performing morphological closing operation on the pixel points determined by the first screening pixel points and the second screening pixel points in the second image, and determining an operation result as the target pixel point.
7. The intelligent gastroenterology image based image recognition system according to claim 1, wherein the performing texture feature analysis on the vascular pixels of the three-dimensional change region to obtain a texture distribution coefficient of each three-dimensional change region includes: According to the first number of the vascular pixel points and the second number of all the pixel points in each three-dimensional change area, obtaining the vascular density of the corresponding three-dimensional change area, wherein the ratio of the first number to the second number is calculated, and the vascular density is obtained after normalization processing; determining the three-dimensional change area with the blood vessel density larger than a preset density threshold as a target change area; Performing vascular skeleton morphology analysis on the target change region to obtain a curvature average value representing vascular curvature characteristics in the target change region and characteristic values representing gradient change characteristics of each pixel point and peripheral pixel points in the target change region; obtaining texture distribution coefficients of the corresponding three-dimensional change areas according to the curvature average value and all characteristic values of the target change areas; the method for determining the curvature mean value and the characteristic value comprises the following steps: Extracting morphological skeleton characteristics of the target change region to obtain skeleton line segments characterized as vascular extension paths; obtaining the curvature of the corresponding skeleton line segment according to the line segment length of each skeleton line segment and the linear lengths of two endpoints, wherein the line segment length of each line skeleton segment is obtained Obtaining a length straight line of a connecting line between two end points of each line segment The curvature w formula of the skeleton line segment is: ; Obtaining the curvature average value of the corresponding target change area according to the curvature average value of all the skeleton line segments in each target change area; obtaining a result sequence representing binary according to the gradient comparison result of each pixel point and the peripheral pixel points in each target change area; converting the binary result sequence into decimal numbers to obtain characteristic values of each pixel point in the target change area; The method for determining the texture distribution coefficient comprises the following steps: carrying out statistical processing on the arrangement distribution of the characteristic values of each pixel point in the target change area, and determining a statistical result as a texture parameter of the corresponding target change area, wherein the variance of all the characteristic values is calculated as the texture parameter; and determining the texture distribution coefficient of the corresponding three-dimensional change area according to the ratio of the curvature mean value of each target change area to the texture parameter.
8. The intelligent gastroenterology image based image recognition system according to claim 1, wherein the region division of the first image according to the texture distribution coefficient and the optical flow characteristics generated by each three-dimensional change region in adjacent image frames comprises: According to the optical flow characteristics of each three-dimensional change area generated in the adjacent image frames, obtaining the optical flow vector variance of the current three-dimensional change area and the optical flow vector difference value of the current three-dimensional change area and the peripheral three-dimensional change area; determining the regional rigidity of the corresponding three-dimensional change region according to the texture distribution coefficient, the optical flow vector variance and the optical flow vector difference value of each three-dimensional change region; and carrying out the division of the three-dimensional change area by an Ojin method based on the numerical value of the rigidity of the area, and determining different identification areas.

Description

Intelligent image recognition system based on digestive system department images Technical Field The invention relates to the technical field of image processing, in particular to an intelligent image recognition system based on an internal digestive system image. Background The intelligent image recognition based on the digestive system image relates to the combination of medical imaging, computer vision, deep learning and artificial intelligence technology, including X-ray, CT (Computed Tomography ), MRI (Magnetic Resonance Imaging, magnetic resonance imaging), endoscopy (such as gastroscope and enteroscope) and other image technologies. These images can provide detailed information of the digestive system. In recent years, deep learning, particularly convolutional neural networks (CNN, convolutional Neural Network), has made significant progress in medical image analysis. By training a large amount of tagged image data, the deep-learning convolutional neural network model can learn image features and automatically recognize and classify the image features. In order to intelligently identify the regions based on the images of the digestive system, the endoscopic images of the patient need to be acquired to identify different expression regions in the images, but partial bulges and depressions may not be obvious in the images, and normal blood vessels on the surface of the intestinal tract and attachments in the intestinal tract influence the region dividing effect on the images on the surface of the intestinal tract due to complex internal environment of the intestinal tract, so that the reliability of the region dividing of the images is poor. Disclosure of Invention In order to solve the technical problem of poor reliability of image region division in the related art, the invention aims to provide an intelligent image recognition system based on digestive system images, which adopts the following technical scheme: the embodiment of the invention provides an intelligent image recognition system based on an image of a digestive system, which comprises: the acquisition terminal is used for acquiring a first image of the intestinal tract of a user in a white light mode under different light source incidence angles and a second image corresponding to the narrow-band imaging mode, and outputting the first image and the second image to the outside; The processing terminal is connected with the acquisition terminal and is used for obtaining a three-dimensional change area of the inner wall of the intestinal canal in a convex-concave structure according to the gray scale difference of the first image at different light source incidence angles; screening out pixel points with blood vessel characteristics in the second image, and marking the corresponding position of the first image as the blood vessel pixel point; performing texture feature analysis on the vascular pixel points of the three-dimensional change areas to obtain texture distribution coefficients of each three-dimensional change area; and dividing the first image into areas according to the texture distribution coefficient and the optical flow characteristic generated by each three-dimensional change area in the adjacent image frames. In an alternative embodiment, the method for obtaining the three-dimensional change area of the inner wall of the intestinal canal in the convex-concave structure according to the gray scale difference of the first image at different incidence angles of the light source comprises the following steps: Performing position alignment processing and morphological opening operation on a plurality of first images of the same intestinal region under different light source incidence angles so as to obtain an image region group corresponding to the intestinal region; And carrying out three-dimensional convex-concave characteristic analysis according to the gray level change of the pixel points of each first image in the image area group so as to obtain a three-dimensional change area with the inner wall of the intestinal canal showing a convex-concave structure. In an alternative embodiment, the three-dimensional convex-concave characteristic analysis is performed according to the gray scale change of the pixel points of each first image in the image area group, so as to obtain a three-dimensional change area of the inner wall of the intestinal canal, which is in a convex-concave structure, and the method comprises the following steps: carrying out gray value difference and absolute value calculation according to pixel points of the same position of each first image in the image area group so as to obtain the difference absolute value of each pixel point; Dividing all absolute values of the difference values by using an Ojin threshold value, and determining pixel points larger than a corresponding threshold value as changed pixel points; Marking the pixel points with the difference value of the pixel points of each first image co