For old woman, ovarian cancer is a severe illness. Based on research, it is seventh major cause for woman death and fifth common disease worldwide. Using Artificial Neural Network (ANN), many researchers performed ovarian cancer classification. For making decision, doctors consider classification accuracy as an efficient factor. For giving proper treatment, doctors consider higher classification accuracy. Early and accurate diagnosis reduces mortality percentage and save lives. This paper proposes the novel annotated ovarian image classification using FR-CNN (fast region-based CNN) on the basis of ROI (region of interest) segmented. Here the input images have been classified into three types namely, epithelial, germ and stroma cells. This image has been preprocessed and segmented. After this annotation process takes places by using FR-CNN. The framework compares the manually annotated feature and trained feature in FRCNN for region based classification. This will help in analyzing the higher accuracy in detection of disease since manual annotation has lower accuracy in existing works so this work will experimentally prove that the machine learning based classification will yield higher accuracy. After the region-based training in FR-CNN, the classification is done by combining SVC-Support vector classifier and Gaussian Naives Bayes classifiers. Due to higher data indexing, the ensemble technique has been used in classification for the features. The simulation gives accurate part of input image to detect ovarian cancer.
Key words: Ovarian cancer, annotated image classification, FR-CNN (Fast Region-based CNN), ROI (Region of Interest), SVM, Gaussian NB, Accuracy.
|
