DURUNet: Improving Colorectal Cancer Image Classification with Enhanced Feature Extraction and mRMR Feature Selection
Year: 2026
Type: article
Abstract: Colorectal cancer, a leading global cancer, presents a significant health challenge with elevated mortality rates. In 2020, over 930,000 deaths were attributed to colorectal cancer worldwide, underscoring the critical need for effective diagnostic tools. This study introduces DURUNet, an innovative deep learning-based classification framework specifically designed to enhance colon cancer image classification. DURUNet excels in detailed and coarse feature extraction, utilizing two distinct networks, Father Network and Daughter Network. We enhanced the feature selection strategy by incorporating the mRMR feature selection algorithm to streamline the complex feature set. Subsequently, we applied SVM and KNN classifiers for feature classification. Experimental results demonstrate that the proposed DURUNet achieves an accuracy of 99.80% and an F1-score of 99.70%, surpassing comparable studies in the existing literature.
Author: Ömer Faruk Söylemez
Institution: Dicle University
Language: English
Topic: AI in cancer detection
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7223552944