Computer aided diagnosis system for breast lesion classification by fusion of B mode ultrasound and elastography breast image features

Abstract

Breast Cancer (BC) is the second leading cause of death among newlinewomen worldwide. Early detection is crucial for improving survival rates and newlinereduce mortality risks. Visual screening of B-mode ultrasound (US) and newlineelastography images is subjective, and time-consuming, highlighting the newlinenecessity for an automated system to detect BC. This research aims to build newlineComputer Aided Diagnosis (CAD) system to classify breast lesions based on newlineUS and elastography features. newlineThe first phase of the research study involves developing a CAD newlinesystem for diagnosing BC. US and elastography images are separately newlinepreprocessed utilizing circular hybrid median filter and contrast-limited newlineadaptive histogram equalization to suppress noise. A differential-evolution newlinebased pulse-coupled neural network isolates the lesion area from the newlineand mapped on elastography image to get matching Region of Interest (ROI). newlineSeveral features, including intensity, morphology features, and strain-based newlinefeatures are extracted and merged into a single feature matrix. Support Vector newlineMachine (SVM) and Linear Discriminant Analysis (LDA), are utilized to newlineclassify the images based on the fused features. newlineThe second phase of this research work deals with the design of a newlinehybrid system for classifying breast lesions by combining the Grasshopper newlineOptimization Algorithm (GOA) and machine learning classifiers. US and newlineelastography images are independently preprocessed with Enhanced Wavelet newlineThresholding (EWT). newline