Paper Title: PNASFH-Net: Parallel NAS forward harmonic Network for colon cancer detection using CT images
Authors: V T Ram Pavan Kumar M, Chin-Shiuh Shieh , Siva Shankar S, Prasun Chakrabart
Corresponding Author: V T Ram Pavan Kumar M (mrpphd2018@gmail.com ), India
Abstract
Colon cancer is a leading cause of cancer-related deaths worldwide, and early detection is vital to reduce mortality rates. While Deep Learning (DL) models are commonly used for colon cancer detection, they often require large datasets and are time-consuming. To address these challenges, a new model, Parallel Neural Architecture Search Forward Harmonic Network (PNASFH-Net), has been developed. PNASFH-Net begins by preprocessing colon images through Adaptive Median Filtering (AMF) to remove noise. It then segments the affected colon region using Pyramid Non-local U-Net (PNU-Net), optimized by the Remora Shuffled Shepherd Optimization Algorithm (RSSOA)—a hybrid algorithm combining the Remora Optimization Algorithm (ROA) and Shuffled Shepherd Optimization Algorithm (SSOA) for improved segmentation accuracy. Next, features from the segmented images are extracted and analyzed by PNASFH-Net, which combines Harmonic Analysis, Neural Architecture Search Network (NASNet), and Parallel Convolutional Neural Network (PCNN) for accurate detection. Experimental results show that PNASFH-Net achieves 98.345% accuracy, 98.512% specificity, and 98.679% sensitivity, demonstrating its potential for precise and early colon cancer detection.
Keywords
Median filter, Pyramid Non-local U-Net, Parallel convolutional neural network, Harmonic analysis, Neuron attention stage-by-stage net
