Pancreatic cancer causes the fourth most cancer-related death in humans worldwide. Early detection of this cancer will improve patient's survival rate considerably. In this paper, we propose an image processing and machine learning system for the exact recognition of pancreatic cancer using PET/CT scan images. The proposed system implicates 5 main elements, i.e., preprocessing, segmentation, feature extraction, feature selection (optimization) and classification. Removal of noises in the image is the major step for the exact identification of tumor, if noises persist; it will provide an in-accurate result. Pre-processing is done as an initial step in removing the noises followed by segmentation in identifying the tumor location; here a novel approach of saliency-based k-means clustering algorithm is utilized to isolate the object from background. Since the features extracted from segmented images consist of irrelevant features, it reduces the classification accuracy in disease recognition. So, efficient feature selection method is introduced in this research work to improve the classification performance. To improve feature selection results, initially, image segmentation is carried out by using saliency-based k-means clustering segmentation, and then feature extraction is done by using First Order and Second Order Statistical features by GLCM and GLRM. Feature selection methods such as PSO and whale optimization methods are utilized. The results obtained by these methods indicate the potential advantages of using feature selection techniques to improve the classification accuracy with a smaller number of feature subset. From the result, one can conclude that the performance of whale is superior to PSO method for classification. Machine Learning Techniques are widely used for the cancer classification. The machine learning classifiers such as DT, KNN, SVM and AdaBoost with ensemble KNN - SVM classifier are utilized to classify the tumor as normal or abnormal. Finally, the proposed framework achieves a classification accuracy of 98.3%.
