Deep Learning Based Colon Cancer Detection Using Ensemble Transfer Learning
Sabahat Nurdan Sinan, Selim Akyokuş, Büşra Sinan · 2025
Colon cancer is one of the widespread cancers leading to high mortality rates worldwide. Deep learning and artificial intelligence-based methods have made great progress in early diagnosis of cancer and medical image analysis. In this study, the aim is to reduce the dependency on manual analysis and to provide faster and more accurate diagnoses with the developed DL-based model. Ensemble and transfer learning-based DL techniques are used to improve performance. LC25000 and EBHI-Seg datasets include histopathological images for colon cancer diagnosis. Colon cancer diagnosis was obtained using VGG16, ResNet152V2, InceptionV3 and specially designed CNN models. Ensemble model accuracy reached 100% in LC25000 and 98.63% in EBHI-Seg. The reliability of the model was increased using repeated holdout cross-validation. The findings indicate that the models employed have the potential to be a good decision support tool in clinical practice.