Prediction Analysis of Pancreatic Tumour using Transfer Learning

Abstract

Pancreatic cancer is one of the most common types of cancer that affect men. The development of this cancer in the male body impairs a number of health difficulties, both physical and mental. The existence of this illness is associated with a great deal of complexity, mostly linked to image processing and image detecting methods. Different deep learning and transfer learning based approaches and algorithms can be employed to ascertain the same, in addition to image processing techniques. The author of the work typically uses the Kaggle dataset, which has precise information on images of pancreatic cancer that were collected through the data collection process, in order to carry out the proposed research study. 750 photos of cancer and 150 healthy, tumor-undetermined images make up the dataset. Using the notions of a histogram, the normalization technique defines the image in its entirety. In addition, the author suggests using YOLO as a transfer learning-based method and CNN's working theory as a deep learning algorithm. However, the transfer learning-based technique implements the same using a pre-trained model, and in order to get higher accuracy and precision values, further hyper-parameter tuning is also carried out.