Histopathological Image Analysis and Classification Techniques for Breast Cancer Detection
Gaurav Makwana, Ram Narayan Yadav, Lalita Gupta · 2021
Breast malignant growth is one of the most incessant reasons for death in women around the world. The early detection of breast cancer growth can improve the endurance rate. Cancer grading of the histopathological image depends upon the histological structures, like tumors, cell membranes, or nuclei, data size, shape, etc., which are very important prerequisites for disease prediction, analysis, and classification. Histopathological images exhibit a lot of variability in the structure, which creates diagnosis uncertainty and error. The automatic breast cancer detection system can help pathologists in the diagnosis of this problem more efficiently. Image processing algorithms for the automated examination of histopathological images have become progressively famous in the most recent decade with the exceptional development in computational power. The development of high-throughput measuring devices takes into account the computer-aided assessment of microscopic images, bringing about a snappy and unbiased image elucidation that encourages the clinical decision-making process. This chapter will explore different techniques for image acquisition, followed by image pre-processing, image segmentation to find the region of interest (ROI), feature extraction of the segmented image, and classification techniques. The chapter is divided into two sections. The first section is analysis, particularly segmentation, in which test images are partitioned for identification of ROI for feature extraction. This part will investigate various feature extraction methods to classify cells or regions into normal and abnormal categories. The next section covers the image classification process with histopathological image analysis. This research will provide a more versatile classification system with discriminating class-specific dictionaries, which automatically identifies the feature based on which the symptom can be easily classified. The machine learning algorithms are used to decide the presence of abnormalities at the pixel location. Support vector machine (SVM) classifier, minimum distance classifier, and convolutional neural network classifier are also employed for classification. The proposed computer-aided diagnostic system reduces subjectivity and makes an accurate, quick, and more precise diagnostic decision. It will help practitioners in increasing the diagnosis accuracy, sensitivity, and specificity of the pathological test.