CLASSIFICATION OF PANCREATIC CANCER FROM CT SCAN IMAGES USING DEEP LEARNING TECHNIQUE

Abstract

Artificial intelligence is being applied in various computer science domains in order to enhance the performance and decision making capability of manual frameworks, Deep learning accompanied with medical imaging is proven to be really fruitful for the detection and classification of various diseases like brain tumor, breast cancer, head and neck cancer, Hepatic and Pulmonary cancer or classification of leukemia and various diseases that can help medical professionals to diagnose and classify. Cancer is also very hard for doctors or medical professionals to spot in the very first place because major cancer symptoms are really similar to other diseases so finding cancer in a patient is a major objective especially in early stages. Pancreatic cancer, a kind of malignant neoplastic disease where the malignant neoplastic disease/malignant cells develops within the tissues of pancreas, the general survival rate of 5 year of pancreatic cancer is 10% while as the pancreas lies behind the stomach and liver so it requires intensive medical imaging. The proposed methodology presents an extensive solution for this problem, the methodology presents a deep learning based system that is able to classify pancreatic and non-pancreatic cancer through CT scan images of human pancreas. The system has been constructed using a custom VGG16 architecture using a collection of 20014 CT scan images for pancreatic and non-pancreatic cancer. The proposed methodology produces training and validation accuracy of 100%, 100% respectively.