Breast Cancer Detection Using Random Forest Classifier

Breast Cancer Detection Using Random Forest Classifier

Pavithra Suchindran, Vanithamani R., Judith Justin
Copyright: © 2021 |Pages: 14
DOI: 10.4018/978-1-7998-6690-9.ch005
OnDemand:
(Individual Chapters)
Available
$37.50
Current Special Offers
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Breast cancer is the second most prevalent type of cancer among women. Breast ultrasound (BUS) imaging is one of the most frequently used diagnostic tools to detect and classify abnormalities in the breast. To improve the diagnostic accuracy, computer-aided diagnosis (CAD) system is helpful for breast cancer detection and classification. Normally, a CAD system consists of four stages: pre-processing, segmentation, feature extraction, and classification. In this chapter, the pre-processing step includes speckle noise removal using speckle reducing anisotropic diffusion (SRAD) filter. The goal of segmentation is to locate the region of interest (ROI) and active contour-based segmentation and fuzzy C means segmentation (FCM) are used in this work. The texture features are extracted and fed to a classifier to categorize the images as normal, benign, and malignant. In this work, three classifiers, namely k-nearest neighbors (KNN) algorithm, decision tree algorithm, and random forest classifier, are used and the performance is compared based on the accuracy of classification.
Chapter Preview
Top

The major limitations of BUS imaging are low contrast and interference with speckle. The main aim of the pre-processing step is to improve the contrast of the image and reduce the speckle noise. Speckle noise is a random multiplicative noise and it occurs in all coherent imaging such as LASER, SAR and ultrasound imaging. Speckle noise makes the visual observation and analysis challenging. Therefore, removing speckle without destroying the important features for diagnosis is difficult. In this section, some speckle reduction techniques are reviewed.

Many speckle reduction techniques have been proposed and are classified as spatial domain filters, wavelet domain techniques, compounding approaches and non-linear diffusion. Certain speckle reduction techniques enhance the image and remove speckle at the same time. Nonlinear diffusion is such an example. It not only preserves edges but also enhances edges by inhibiting diffusion across edges and allowing diffusion on either side of the edges (Cheng HD et al., 2006). Speckle Reducing Anisotropic Diffusion (SRAD) is proposed particularly for Ultrasound images (Yu & Acton S.T. 2002). Zhenyu Zhou et al. (2015) developed a nonlinear diffusion filter denoising framework for multiplicative noise removal. The authors have presented a doubly degenerate diffusion model and demonstrated that their model outperformed its competitors both visually and quantitatively.

The most important step after preprocessing in CAD is segmentation. Segmentation helps to partition the image into non-overlapping regions. (Horsh et al., 2009) Proposed a method for automatic segmentation of breast lesions from Ultrasound images includes thresholding enhanced mass structures. The active contour model, generally known as snake, is a framework for outlining an object from a 2D image, and has been extensively used for US images. (Abdul Kadir Jumaat et al., 2010) Used active contour to identify the boundaries and adopted mathematical concepts for energy minimization. Balloon Snake algorithm was used for segmenting the masses from BUS images. The accuracy of the Balloon Snake algorithm was calculated by comparing the masses between the radiologist’s observation and it was found to be 95.53%. Neural Network (NN) based methods (Chen. D. R. et al., 2002) are widespread in image segmentation, which convert the segmentation problem into classification decision based on a set of input features.

Complete Chapter List

Search this Book:
Reset