Optimized Depth-Wise Separable Convolution Neural Network for Lung and Colon Cancer Detection
摘要
In this study, we develop lung disease data classification using deep learning technique. There are three stages in the proposed method namely, preprocessing, segmentation, and classification. Initially, CT images are collected and preprocessed. A modified median filter is used for preprocessing. Then, the pre-processed images are fed to the segmentation process. In order to segment the image, K-means clustering is used. Then, this segmented region is fed to the input of an optimal depth-wise segmented CNN classifier to classify the image as a normal or cancerous image. To optimize the hyperparameter of the classifier Coati Optimization Algorithm (COA) is used. The suggested approach performs better, with a 95.07% accuracy rate when evaluated using several measures.