CN-121982044-A - Classification method for tongue fur segmentation image and digestive diseases

CN121982044ACN 121982044 ACN121982044 ACN 121982044ACN-121982044-A

Abstract

The invention relates to the technical field of image segmentation, in particular to a classification method for tongue fur segmentation images and digestive diseases. According to the method, the positions of the center points of particles are determined by introducing multi-scale LoG filtering, tongue particles and tongue fur particles are distinguished according to the rule of spatial distribution and the density change characteristics, local self-adaptive threshold adjustment is carried out on other pixel points through the change of uneven particle density of different parts, more accurate tongue fur areas and tongue fur areas are obtained through division, and subsequent classification is carried out based on the tongue fur areas and the tongue fur areas. According to the invention, tongue fur areas are more accurately divided through multi-scale Log filtering and space statistics characteristics, a reliable data basis is provided for subsequent auxiliary classification of digestive diseases, and classification accuracy is improved.

Inventors

LIU ZHENGHUI
LI XIN
LI LONGMEI
GAO LEI
SHI YING
FU XIN
Kou Shaojie

Assignees

陕西中医药大学附属医院
西安市中医医院

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (10)

1. A method for classifying tongue coating segmentation images and digestive diseases, the method comprising: analyzing the Log response values of each pixel point position under different scales based on multi-scale Log filter response in the tongue region, and screening out a particle center point; The method comprises the steps of obtaining a tongue coating disorder index of each particle center point according to the distribution rule condition of the particle center points of each particle center point in a local range, wherein the local range is a preset local round window with the particle center point as a center, analyzing the density distribution change degree of the particle center points among square preset area windows which are uniformly divided, and determining a tongue coating density change index of the particle center point; For the non-particle center point pixel point, adjusting the size of a threshold value according to the density proportion condition of tongue coating particle points in all particle center points in a square preset neighborhood region range taking the non-particle center point pixel point as the center, and determining a tongue coating dynamic threshold value of the non-particle center point pixel point; and obtaining a classification result through the tongue fur area and the tongue texture area in the tongue body area.
2. The method for classifying a tongue fur segmentation image and digestive diseases according to claim 1, wherein the method for acquiring the particle center point comprises: For any pixel point, calculating the Log response value of each pixel point under each scale and carrying out normalization processing to obtain a response index; and when the response index of the pixel point under the expected scale is larger than a preset response threshold, taking the pixel point as a particle center point, wherein the preset response threshold is a conventional experience threshold for distinguishing the particle center point from noise interference.
3. The method for classifying a tongue coating segmentation image and a digestive disorder according to claim 1, wherein the method for acquiring the tongue coating disorder index comprises the following steps: for any particle center point, taking the ratio of the number of all particle center points to the area of a preset local round window in a preset local round window with the particle center point as the local density of the particle center point; selecting at least two sub-windows in a preset partial round window, counting the number of particle center points in each sub-window and arranging according to the radial order to obtain a number sequence; and integrating the L function curve, and performing negative correlation mapping on the absolute value after integration to obtain the tongue coating disorder index of the particle center point.
4. The method for classifying a tongue fur segmentation image and digestive diseases according to claim 1, wherein the method for acquiring the tongue fur density change index comprises: Calculating the difference of the density values between each preset area window and each adjacent other preset area window, and taking the maximum difference as the change amplitude value of each preset area window; And taking the change amplitude of the window of the preset area where the center point of each particle is positioned as a tongue fur density change index of the center point of each particle.
5. The method for classifying tongue coating segmentation images and digestive diseases according to claim 1, wherein the step of screening tongue coating particles and tongue texture particles based on the disorder index and the density change index of the tongue coating at the center point of the particles comprises the steps of: Normalizing the product of the tongue coating disorder index and the tongue coating density change index of each particle center point to obtain a tongue coating probability index of the particle center point; And taking the particle center point with the tongue coating probability index larger than a preset tongue coating threshold value as a tongue coating particle point, otherwise taking the particle center point as a tongue quality particle point.
6. The method for classifying a tongue fur segmentation image and digestive diseases according to claim 1, wherein the method for acquiring the tongue fur dynamic threshold value comprises: for any non-particle center point, taking the ratio of the number of tongue coating particle points to the number of all particle center points in the preset neighborhood region range of the non-particle center point as a local tongue coating density index of the non-particle center point; Carrying out maximum and minimum normalization treatment on the local tongue fur density indexes of all non-particle center points to obtain a density adjustment coefficient of each non-particle center point; and taking the sum of a preset minimum threshold value and the threshold value regulating coefficient as the dynamic threshold value of the tongue fur of the non-particle center point.
7. The method for classifying a tongue fur segmentation image and digestive diseases according to claim 1, wherein the pixel-based tongue fur dynamic threshold segmentation method is characterized by comprising the steps of: When the local tongue coating density index of the non-particle center point is larger than the tongue coating dynamic threshold value, marking the non-particle center point as a tongue coating region pixel point, otherwise marking the non-particle center point as a tongue quality region pixel point; And taking the region formed by all the tongue coating region pixel points and the tongue coating particle points as a tongue coating region, and taking the region formed by all the tongue region pixel points and the tongue particle points as a tongue region.
8. The method for classifying a tongue fur segmentation image and digestive diseases according to claim 1, wherein the step of obtaining the classification result through the tongue fur region and the tongue texture region in the tongue body region comprises: And inputting the ratio of tongue fur areas and the ratio of tongue body areas in different areas in the tongue body area into a trained classification model as input, and outputting a classification result.
9. The method of classifying a tongue coating segmentation image and digestive disorder according to claim 7, further comprising, after segmenting the tongue coating region and the tongue texture region: And taking each tongue coating region and each tongue mass region as an analysis region in sequence, removing the marks from the analysis region when the region area of the analysis region is smaller than a preset area threshold value, and smoothly filling the marks in the analysis region through morphological closing operation to obtain updated tongue coating regions and tongue mass regions.
10. A method of classifying a tongue coating segmentation image and digestive disorder according to claim 3, wherein the sub-windows are in accordance with a predetermined partial circular window shape, and the radii increase from a minimum radius by a predetermined step size, each radius corresponding to one sub-window, until the radius is greater than or equal to the predetermined partial circular window.

Description

Classification method for tongue fur segmentation image and digestive diseases Technical Field The invention relates to the technical field of image segmentation, in particular to a classification method for tongue fur segmentation images and digestive diseases. Background The tongue coating is an important basis for traditional Chinese medicine diagnosis, and the thickness, distribution and texture of the tongue coating directly reflect the health condition of the digestive system of a human body. The tongue coating segmentation is performed to accurately separate the tongue from the tongue coating region, so that the coverage range, the distribution mode and other characteristics of the tongue coating are quantitatively analyzed, and a foundation is laid for the subsequent objective and accurate auxiliary analysis based on image processing and machine learning technology. The existing tongue coating segmentation technology is mainly based on threshold segmentation or edge detection of a color space, and the traditional machine learning method mostly depends on macroscopic differences of the tongue coating and the tongue quality in color or texture. However, the dependence of the image acquisition environment is high, the tongue coating and the tongue quality are similar in color, the difference of macroscopic features is not obvious, the segmentation error is easy to generate due to slight change of color, and when the tongue surface microstructure is analyzed, the tongue coating distribution generalization capability facing to complex or uneven is poor, so that the segmentation precision and the robustness are insufficient, and the subsequent classification of the segmentation result is influenced. Disclosure of Invention In order to solve the technical problems in the prior art, the invention aims to provide a classification method for tongue fur segmentation images and digestive diseases, which adopts the following technical scheme: The invention provides a classification method of tongue fur segmentation images and digestive diseases, which comprises the following steps: analyzing the Log response values of each pixel point position under different scales based on multi-scale Log filter response in the tongue region, and screening out a particle center point; The method comprises the steps of obtaining a tongue coating disorder index of each particle center point according to the distribution rule condition of the particle center points of each particle center point in a local range, wherein the local range is a preset local round window with the particle center point as a center, analyzing the density distribution change degree of the particle center points among square preset area windows which are uniformly divided, and determining a tongue coating density change index of the particle center point; For the non-particle center point pixel point, adjusting the size of a threshold value according to the density proportion condition of tongue coating particle points in all particle center points in a square preset neighborhood region range taking the non-particle center point pixel point as the center, and determining a tongue coating dynamic threshold value of the non-particle center point pixel point; and obtaining a classification result through the tongue fur area and the tongue texture area in the tongue body area. Further, the method for acquiring the center point of the particle comprises the following steps: For any pixel point, calculating the Log response value of each pixel point under each scale and carrying out normalization processing to obtain a response index; and when the response index of the pixel point under the expected scale is larger than a preset response threshold, taking the pixel point as a particle center point, wherein the preset response threshold is a conventional experience threshold for distinguishing the particle center point from noise interference. Further, the method for obtaining the tongue coating disorder index comprises the following steps: for any particle center point, taking the ratio of the number of all particle center points to the area of a preset local round window in a preset local round window with the particle center point as the local density of the particle center point; selecting at least two sub-windows in a preset partial round window, counting the number of particle center points in each sub-window and arranging according to the radial order to obtain a number sequence; and integrating the L function curve, and performing negative correlation mapping on the absolute value after integration to obtain the tongue coating disorder index of the particle center point. Further, the method for obtaining the tongue fur density change index comprises the following steps: Calculating the difference of the density values between each preset area window and each adjacent other preset area window, and taking the maximum difference as the change amplitude valu