A Quantitative Analysis of Sonographic Images of the Salivary Gland: A Comparison Between Sonographic and Sialographic Findings

Abstract 

We performed three quantitative analyses (particle analysis, fractional Brownian motion [fBM] model analysis, two-dimensional [2-D] fractal analysis) of the ultrasonographic (US) images of the salivary gland and evaluated whether the obtained indices correlated with the sialographic stage of Rubin-Holt. Our study included 192 patients suspected of having Sjögren's syndrome (SS). In 89 patients, sialography demonstrated abnormal findings. Based on a particle analysis, we calculated both the average size of the particles (avg-area) and the area ratio to evaluate the presence of hypoechoic areas and echogenic lines, which are characteristic of SS. According to the fBM model, we calculated the Hurst index of the original image (Hurst-ori) and the background-subtracted image (Hurst-bs) to evaluate the complexity of the pixel value distribution. We also obtained the 2-D fractal dimension (2-D-FD) to evaluate the complexity of the contour lines. We entered these indices of the parotid glands (PG) into a logistic regression analysis and evaluated which indices were useful predictors for detecting an abnormal sialographic stage. Significant differences were observed between the normal and abnormal groups in all five indices of the PG (Mann-Whitney U test) and all five indices were correlated with the Rubin-Holt stage (Spearman's Rank Correlation Test). As the Rubin-Holt stage became more severe, both the Hurst-ori and 2-D-FD became smaller. Alternatively, the Hurst-bs, avg-area, and area ratio became higher. Three indices (avg-area, area ratio and Hurst-ori) were selected as useful predictors for detecting abnormal sialographic stages. This quantitative analysis system is therefore considered to have potentially useful clinical applications for the detection of abnormal sialographic findings. (Email: chikui@rad.dent.kyushu-u.ac.jp)

Key Words: Sjögren's syndrome, Particle analysis, fBM model analysis, Ultrasonography, Sialography