Semi-supervised Polyp Classification in Colonoscopy Images Using GAN

Colorectal cancer (CRC) screening is carried out to appropriately identify and categorize the CRC precursors, called polyps. It is a crucial task for an effective treatment as unnecessary surgeries and ignorance of any potential cancer could be the situations of concern. Thus, classifying any detected lesion as an adenoma (with the potential to become cancerous) or hyperplastic (non-cancerous) is necessary. In recent years, such classification tasks are assisted by automated systems with promising outcomes. However, the extensive data needed for these systems to function satisfactorily have limited the widespread application of deep learning techniques in the medical field. To overcome this issue, in this paper, we employed a semi-supervised learning approach using GAN to classify polyps. Our approach needs limited labeled data while achieving enhanced classification performance. The experimental results show that the suggested semi-GAN approach, when applied to the PolypsSet and SUN datasets, produced a classification accuracy of 72.85% and 74.12% respectively. This represents a significant improvement over existing methods and highlights the potential of GANs in the medical image analysis field. These research findings have implications for improving the efficiency and accuracy of CRC screening and diagnosis, potentially leading to better patient outcomes.