Abstract

This research presents a two-stage computer-aided diagnosis (CAD) system that employs a convolutional neural network (CNN) with self-supervised learning (SSL) to classify polyps as either hyperplastic polyps (HP) or tubular adenomas (TA). The proposed model integrates look-into-object (LIO) and contrastive learning in SimCLR to focus on the comprehensive polyp region, enhancing model performance. The LIO scheme, however, depends on pretraining a model to provide basic representations, which is improved using a warm-up scheme to optimize the loss function. Due to the scarcity of medical images for effective polyp classification training, an alternative approach utilizes natural images for the pretext task instead of polyp images. The research findings demonstrate that the proposed scheme, which leverages polyp object structure information and self-supervised learning, produces a robust model that improves classification accuracy for HP and TA in the prediction head by transferring a backbone. The backbone model effectively employs ResNet-18 to focus on the holistic polyp using a limited number of labeled polyp images. The proposed scheme surpasses existing methods, achieving a 4% increase in accuracy and a 3% improvement in the F1-score.

Bio.

Qi-Xian Huang earned his Ph.D. from National Tsing Hua University, Hsinchu, Taiwan, in 2023. He began his career as an intern at Taiwan Semiconductor Manufacturing Corporation (TSMC) in 2018 and subsequently worked as a Senior Firmware Engineer at Qualcomm Semiconductor Corporation until 2023. Additionally, he served as a researcher at the Artificial Intelligence E-learning Center of National Chengchi University, Taipei, Taiwan. He possesses extensive experience as an IEEE/ACM referee for journals such as Signal Processing Systems, ACM Transactions on Privacy and Security, IEEE Transactions on Systems, Man, and Cybernetics: Systems (T-SMC), IEEE Transactions on Network and Service Management (T-NSM), IFIP International Internet of Things (IoT) Conference, Machine Learning, and Education E-learning. Currently, he is employed as a wireless product developer at MediaTek Corporation and is a member of the Taiwanese Association for Artificial Intelligence (TAAI). His research interests primarily revolve around Deep Learning for Computer Vision Algorithms, focusing on applications in medical imaging, super resolution, large language models (LLM), beyond 5G/6G, Blockchain Network Security Protocols, and Internet of Things Security.

All faculty and students are welcome to join.