CNN Classification of Computed Tomography Images for Pancreatic Tumor Detection

Abstract

The five-year survival rate for pancreatic cancer (PC) is the lowest of any cancer kind, and it is the fourth greatest cause of cancer-related death, with a growing death rate. When it comes to cancer invasion, the most significant risk factors are: smoking; alcohol usage; diabetes; and prior pancreatitis. By using this method, we will be able to detect our PC, which is equipped with picture handling technology. Researchers used CT images as input in this study and preprocessed them to remove any noise in the images that had been learned using an adaptive Weiner filter. Preprocessing is followed by the use of a region grow ideal to segment the noise-free image. Scale Invariant Feature Transform (SIFT) is utilized once more to extract the tumor limits and principal component analysis (PCA) is used to enhance the retrieved structures to improve the types of pancreatic CT images. In order to activate the picture parameters, a convolutional neural network (CNN) classifier is used. In order to categorize an image as nonpancreatic cancer or pancreatic cancer, the test data were compared to the training data and the classified image was compared. MATLAB then initiates the entire process, and the most recent performance estimation approach is utilized, resulting in outstanding accuracy.