Vol. 15 No. 2 (2025): IJCRT, Volume 15, Issue 2, 2025
Articles

Integrating Image Preprocessing and Pretrained CNN Using Transfer Learning for CRC Diagnosis

N. Muhilarasi
PG Scholar., Communication Systems, Department of Electronics and Communication Engineering., Government College of Engineering Salem, Tamil Nadu, India
Dr. P. Indra
Assistant Professor, Department of Electronics and Communication Engineering., Government College of Engineering Salem, Tamil Nadu, India

Published 2025-04-18

Keywords

  • Image Processing,
  • Deep Neural Network (DNN),
  • Deep Learning,
  • CRC Diagnosis

How to Cite

N. Muhilarasi, & Dr. P. Indra. (2025). Integrating Image Preprocessing and Pretrained CNN Using Transfer Learning for CRC Diagnosis. IJCRT Research Journal | UGC Approved and UGC Care Journal | Scopus Indexed Journal Norms, 15(2), 50514–50530. https://doi.org/10.5281/zenodo.15239367

Abstract

Colorectal cancer is one of the leading causes of cancer-related mortality worldwide, where early diagnosis for CRC plays a critical role in improving patient outcomes and survival rates. This study presents a hybrid model that integrates traditional image processing techniques with deep learning using pretrained convolutional neural networks (CNNs) to enhance the detection of colorectal cancer from histopathology images. The Preprocessing phase utilizes a progressive switching median filter to enhance image quality by reducing noise without compromising essential histopathological details. Following denoising, the Superpixel SLIC (Simple Linear Iterative Clustering) algorithm performs precise image segmentation, isolating key regions of interest for improved feature extraction. In the classification phase, three convolutional neural network (CNN) architectures—DenseNet, ResNet50, and EfficientNet—are employed to classify colorectal cancer from histopathology images. Each network’s performance is evaluated based on accuracy in the classification stage, with DenseNet achieving the highest accuracy of 99.32%, followed by EfficientNet with 98.05%, and ResNet50 with 95.99%. The proposed hybrid model demonstrates that integrating sophisticated image processing with pretrained CNN models enhances diagnostic performance, offering a robust and accurate tool for automated colorectal cancer detection. This hybrid approach could support pathologists in identifying cancerous tissues more effectively, potentially improving patient outcomes.