An Efficient Method for Detection and Classification of Pulmonary Neoplasm based on Deep Learning Technique

Abstract

Cancers of the lung and pancreas are two of the most frequent cancers. While lung cancer is the leading cause of cancer-related deaths worldwide, pancreatic cancer has the worst prognosis, with only a 7% 5-year survival rate in the US. In lung and pancreatic cancer screening, accurate nodule detection is critical. Typically, radiologists discover these cancers on medical pictures through a thorough review of CT images, which takes a long time and is inexact. As a result, an unique hybrid model for accurately detecting such malignancies is offered. Because it is a progressive diagnostic tool for medical reasons, the proposed method is accomplished using image processing techniques. There are two stages to the proposed technique. In the first stage, both supervised and unsupervised deep learning algorithms and SVM techniques are used to classify whether lung and pancreatic tumors are malignant or benign, as well as to improve tumor characterization. The tumor component is discovered in the second stage utilizing Convolution Neural Networks and Transfer Learning. The CT image is first pre-processed from the data base for scaling and noise removal. After that, the image is categorized using supervised and unsupervised learning techniques. CNN segments the tumor part if the tumor in the image is assessed as malignant. Finally, several metrics such as accuracy, specificity, and sensitivity are calculated and compared to previous method findings.