Texture analysis of sonographic features of the parotid gland in Sjögren's syndrome

OBJECTIVE. The purpose of this study was to develop a method for quantitative analysis of the sonographic features of parotid glands as a noninvasive tool for the diagnosis of Sjögren's syndrome. SUBJECTS AND METHODS. Sonographic texture analyses were performed on the parotid glands of 44 patients with Sjögren's syndrome, 83 healthy volunteers, and 17 patients with chronic parotitis, using a fast Fourier transform program. RESULTS. Texture analysis of sonographic studies of the parotid gland using a Fourier transform showed that the sum of the normalized radial power spectrum in the low-spatial-frequency region (S value) of the parotid gland was significantly higher (p < .0001) in the patients with definite Sjögren's syndrome [6.70 ± 2.13 (×10⁵)] than in the 72 age-matched normal volunteers [3.25 ± 1.08 (×10⁵)]. However, patients with probable Sjögren's syndrome showed S values [3.92 ± 1.88 (×10⁵)] similar to those of the controls. On the other hand, SDs of the echo levels in the parotid gland showed significantly greater (p < .0001) values in patients with definite (4.63 ± 1.07) and probable (4.53 ± 1.47) Sjögren's syndrome than in the normal controls (3.30 ± 0.76). Discriminant analysis showed that a combination of these two distinctive values increased diagnostic accuracy to 96.9%. Furthermore, S values and SDs correlated well with the qualitative grading of sonographic features and with the gradings of sialography. CONCLUSION. The system we describe for texture analysis of sonographic images is useful in the diagnosis of Sjögren's syndrome.