Quantitative Evaluation of Kidney and Gallbladder Stones by Texture Analysis Using Gray Level Co-Occurrence Matrix Based on Diagnostic Ultrasound Images

Abstract

Background/Objectives: Accurate diagnosis during ultrasound examinations of patients with kidney and gallbladder stones is crucial. Although stone areas typically show posterior acoustic shadowing on ultrasound images, their accurate diagnosis can be challenging if the shaded areas are vague. This study proposes a method to improve the diagnostic accuracy of kidney and gallbladder stones through texture analysis of ultrasound images. Methods: Two doctors and three sonographers evaluated abdominal ultrasound images and categorized kidney and gallbladder stones into groups based on their predicted likelihood of being present: 50-60%, 60-80%, and ≥80%. The texture analysis method for the posterior acoustic shadows generated from ultrasound images of stones was modeled using a gray level co-occurrence matrix (GLCM). Average values and 95% confidence intervals were used to evaluate the method. Results: The three prediction classes were clearly distinguished when GLCM_Contrast was applied to the ultrasound images of patients with kidney and gallbladder stones. However, GLCM_Correlation, GLCM_Energy, and GLCM_Homogeneity were found to be difficult for analyzing the texture of shadowed areas in ultrasound images because they did not clearly or completely distinguish between the three classes. Conclusions: Accurate diagnosis of kidney and gallbladder stones may be possible using the GLCM texture analysis method applied to ultrasound images.

Keywords: diagnostic ultrasound image; gray level co-occurrence matrix; kidney and gallbladder stones; posterior acoustic shadow; texture analysis.

Figure 1

Simplified framework for analyzing the distribution of kidney and gallbladder stones according to the degree of calcification using gray level co-occurrence matrix (GLCM) functions.

Figure 2

Kidney stone ultrasound images acquired for GLCM analysis. Sample ultrasound images of suspected kidney stones, with probabilities of (a) 50–60%, (b) 60–80%, and (c) ≥80%, analyzed by doctors and sonographers.

Figure 3

Analysis of GLCM characteristics with respect to the expected diagnosis rate of kidney stones. (a) ${GLCM}_{Contrast}$, (b) ${GLCM}_{Correlation}$, (c) ${GLCM}_{Energy}$, and (d) ${GLCM}_{Homogeneity}$ analysis graphs for the posterior acoustic shadow region of kidney stones. Black, red, and blue shades in the graph represent the range of confidence interval values for ultrasound images with suspected kidney stones, categorized into probabilities of 50–60%, 60–80%, and ≥80%, respectively.

Figure 4

Gallbladder stone ultrasound images acquired for GLCM analysis. Sample ultrasound images of suspected gallbladder stones, with probabilities of (a) 50–60%, (b) 60–80%, and (c) ≥80%, analyzed by doctors and sonographers.

Figure 5

Analysis of GLCM characteristics with respect to the expected diagnosis rate of gallbladder stones. (a) ${GLCM}_{Contrast}$, (b) ${GLCM}_{Correlation}$, (c) ${GLCM}_{Energy}$, and (d) ${GLCM}_{Homogeneity}$ analysis graphs of the posterior acoustic shadow region of gallbladder stones. Black, red, and blue shades in the graph represent the range of confidence interval values for ultrasound images with suspected gallbladder stones, categorized into probabilities of 50–60%, 60–80%, and ≥80%, respectively.