Analysis of Different Image Processing Techniques for Classification and Detection of Cancer Cells

Abstract

This chapter gives an analysis of various methodologies for detecting cancer cells through image processing techniques. The challenges during such detections are over-segmentation and computational complexities. Therefore, the algorithms dealing with such problems are analyzed in this chapter. In these algorithms, a watershed and setting up threshold are helpful to overcome segmentation issues. A support vector machine is discussed to detect subtypes of pneumoconiosis for disjointing segments of lungs. For finding lung cancer cells, a segmentation weighted fuzzy probabilistic-based clustering has been used. Multiple variants of thresholding along with classifiers are proposed to detect lungs and liver cancer. Other than that, noise-removal, feature extraction and watershed are used to detect breast cancer. For leukemia, a bimodal thresholding over enhanced images of cytoplasm and nuclei regions has been discussed. kNN classifier, k-mean clustering, and feed-forward neural networks have also been discussed. Results from these techniques vary from 60%-100% depending on the proposed methodology.