Figure 1.

General schema of proposed framework. The proposed framework consists of data preprocessing (Processing), feature construction and delineation of malignancy steps. In the processing step, the virtual slides are partitioned into equi-size subimages of 128 × 128. The subimages are labeled to positive (malignant) or negative (healthy) by human pathologist. Then a subset of subimages is selected as training data. Furthermore, a color analysis to categorize colors is performed. In the feature construction step, density-based clustering algorithm is applied to find positive color areas, which builds the features for the classification of subimages. In the step of delineation of malignancy, a classification model is trained. The trained model is then used to classify subimages. The area of malignancy is then delineated.