Ultrasound Ovary Image Classification Using Kσ-Classifier

Abstract

Transvaginal Ultra Sound (TVUS) imaging is preferred imaging modality in detection of ovarian abnormalities. The ovarian parameters are measured manually by the expert and the shape of the Ovary is analyzed subjectively. There is a need for computer-assisted diagnostic support system to aid the experts in faster diagnosis as Manual measurement is time consuming. In this paper, we have extracted geometrical and shape features of the Ovary and have used Kσ-classifier to classify the Ovary as normal or abnormal. The proposed method is tested on Transvaginal ultrasound images of ovaries. The obtained experimental results are validated with the manual measurements and inferences by the medical expert and demonstrate the efficacy of the method. The algorithm could achieve a classification rate of 76.67% for Bilinear filtering-Contrast Stretched- Adaptive Thresholding (BCAT) method and 85.8% for Anisotropic filtering-CLAHE-Adaptive Thresholding (ACAT) method.