A COMPARATIVE APPROACH TO PREDICTING COLORECTAL CANCER USING DEEP LEARNING-BASED FRAMEWORKS

Abstract

Colorectal cancer (CRC) is the most serious condition that affects both men and women globally. Recently, growing interest in technological resources has been noticed in diagnosis of cancer of various types. The colorectal cancer is also major cause of death worldwide. High resolution medical images are great source of cancer prediction by using the power of deep learning (DL) algorithms in effective way. Present work focuses on comparative analysis of three deep learning models to predict colorectal cancer using publicly available 7180 images dataset for CRC classification. These frameworks are MobileNetV2, ResNet50 and VGG-16. This can be helpful for practitioners and investigators to understand the challenges in predicting cancer by utilizing new technologies based on deep learning methodologies. Generally, different deep learning methods for colorectal cancer detection are inspected for better understanding of more useful method. The efficiency of deep learning models is evaluated with standard error metrics. The performance metrics of models show that ResNet50 model is more reliable for cancer prediction as compared to MobileNetV2 and VGG-16 applied on selected dataset utilized in this study. The findings highlight the potential of deep learning-based approaches as supportive tools for automated colorectal cancer diagnosis, while also emphasizing the importance of model selection, accuracy, and reliability for future clinical applications.