Breast Cancer Diagnosis using Multi-Fractal Dimension Spectra
Loay Edwar George, Kamal H. Sager · 2007
The research presented in this paper was aimed to develop a classification system of breast tumors tissues using histopathological images. The paper focuses on using the advantages of fractal geometry texture analysis. The developed approach consists of two main steps: (i) the extraction of the fractal dimension spectra for the regions of interest, and (ii) using a classifier that automatically identifies the kind of the tested breast tumor tissue. The performance of the classifier was enhanced by using k-means clustering algorithm; this algorithm was utilized to define sets of centroids, such that each set various states of a specific kind of breast tumors tissue. The classification powers of several fractal dimension spectra descriptors have been investigated. The proposed classification methods were applied on 24 different histopathological breast tumors images. The reached higher recognition accuracy was (98.8%).