Ensemble of Local Texture Descriptor for Accurate Breast Cancer Detection from Histopathologic Images

Histopathological analysis is important for detection of the breast cancer (BC). Computer-aided diagnosis and systems are developed to assist radiologist in process relieve patient from unnecessary pain. In this study, a computer-aided system early benign malignant proposed. The proposed consists feature extraction, ensemble, classification stages. Various preprocessing steps such as grayscale conversion, noise filtering, image resizing employed on input histopathological images. local texture descriptors namely Local Binary Pattern (LBP), Frequency Decoded LBP (FDLBP), Gabor (BGP), Phase Quantization (LPQ), Binarized Statistical Image Features (BSIF), CENsus TRansform hISTogram (CENTRIST), Pyramid Histogram Oriented Gradients (PHOG) extraction histopathologic obtained features then concatenated construction ensemble features. Three classifiers Support Vector Machines (SVM), K-nearest neighbor (KNN), Neural Networks (NN) used BC accuracy score performance evaluation. A dataset called BreaKHis studies. There 9109 microscopic pictures BreaKHis, with 2480 samples 5429 samples. During collection data, 82 patients' tumor tissues were envisioned using various magnification factors 40X, 100X, 200X, 400X. assess acquired findings. results show that method has potential use accurate detection.