S-transform Based Detection and Classification of Breast Cancer in Ultrasound Images
No Thumbnail Available
Date
2019-09
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Breast cancer is the second leading cause of death for women all over the world. Since the cause
of breast cancer remains unknown, early detection and diagnosis is the key for control. In that
regard, Breast Ultrasound Imaging (BUS) has become important modality for breast cancer
detection due to its noninvasive, cost effective nature and suitability for screening and diagnosing
in low resource settings.
Ultrasound imaging is one of the most frequently used diagnosis tools to detect and classify
abnormalities in breast. A known drawback of the technology is that it has high amount of speckle
noise which results in poor image quality. This makes it difficult to use the imaging technology
for accurate detection of malignant tumors. The procedure is traditionally carried out by visual
assessment of the images which is often a time taking process prone to observer variability issues.
In this regard, computer aided detection techniques have been developed in various literatures
showing promises with their merits and demerits. Nevertheless, image based accurate detection of
breast cancer is still a topic of interest with many ongoing researches in the area.
In the current work, S-transform based breast cancer detection and classification method is
developed. The proposed system consists of four stages: preprocessing, segmentation, feature
extraction and classification. Image enhancement and speckle noise reduction were implemented
during preprocessing. Region of interest (ROI) is then accurately determined on preprocessed
images by employing canny edge detection. The ultrasound images were then classified based on
different features like mean, variance, standard deviation (STD), entropy and contrast metrics. The
results of the classification stage were compared against available ground truth images acquired
from research image database. Accordingly, the classification procedure implemented using
artificial neural network offered 90% detection accuracy.
Description
Keywords
S-transform, Breast Tumor, Ultrasound, Image Processing, Feature Extraction, Classification