Premium
SU‐E‐I‐115: Wavelet Analysis of Ultrasound Image for the Diagnosis of Sjögren's Syndrome
Author(s) -
Murakami Y,
Shiraishi A,
Sumi T,
Nakamura T,
Ohki M
Publication year - 2012
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4734832
Subject(s) - ultrasound , wavelet , region of interest , feature (linguistics) , parotid gland , artificial intelligence , image processing , medicine , mathematics , pattern recognition (psychology) , radiology , nuclear medicine , image (mathematics) , computer science , pathology , linguistics , philosophy
Purpose: Sjögren's syndrome (SS) is an auto‐immune disease presenting with dry eyes and mouth (keratoconjunctivitis sicca and xerostomia). Ultrasonography is used for the initial and non‐invasive investigation of the parotid gland in the disease. The purpose of this study is to develop an image processing for diagnosis of SS by applying wavelet analysis to ultrasound image. Methods: Ultrasound B‐mode images of the parotid gland were captured and analyzed by a personal computer. A square region of interest (ROI) was set on the image and two‐dimensional discrete wavelet transform was performed within the ROI. As a Result , the image was decomposed into an approximate image and three detailed images in vertical, horizontal and diagonal directions in different scales. A feature quantity for image classification was defined by calculating from the wavelet coefficients of detailed images within selected scales. The ultrasound images of 80 patients who had been referred to Nagasaki University Hospital because of suspicion of SS were analyzed. A total of 37 patients fulfilled the criteria for SS, whereas the remaining 43 patients did not. The severity of SS was graded into four degrees by sialography. The images with each feature quantity were classified by statistical cluster analysis. Results: In this method, the images can be divided into two groups which mainly contained SS and non‐SS. The sensitivity and specificity in the detection of SS was 78% and 95%, respectively. It was also found that the defined feature quantity tended to change with the severity of SS. Conclusions: In ultrasonography, the image analysis based on wavelet transform was useful for the diagnosis of Sjögren's syndrome.