School of Engineering Department of Electronic and Computer Engineering 161 Federated Learning with Medical Images Supervisor: LI Xiaomeng / ECE Student: YUAN Mingxuan / ELEC Course: UROP 1100, Fall During this fall term, I participated in the UROP1100 project and conducted research on medical image segmentation under the leadership of Professor Li Xiaomeng and Dr. Wang Haonan. In more than two weeks of study, I have a deeper understanding of medical image segmentation. During this period, I learned how to use and modify the excellent open source code on GitHub and read many papers on SAM. Currently, I am trying to use SAM to train a medical model, and trying to make appropriate modifications to SAM, such as adding an adapter, to make it suitable for the medical model. Learning with Limited Anotated Data for Medical Image Diagnosis Supervisor: LI Xiaomeng / ECE Student: LI Hongrui / COMP Course: UROP 1100, Fall UROP 2100, Spring As explored in previous UROP projects, the field of medical image diagnosis has seen significant progress with the rapid development of deep learning based methods. By improving the performance of various medical image tasks such as image classification and segmentation, these advancements have not only showed great potential in assisting clinical diagnosis, but also given people another perspective to understand some medical issues. In this project of UROP2100, 2 main topics have been explored: (1) continuing the experiments on Mask-guided Brain Tumor Segmentation task begun in UROP1100, and (2) exploring visual stimuli decoding based on fMRI scans. Learning with Limited Anotated Data for Medical Image Diagnosis Supervisor: LI Xiaomeng / ECE Student: LIANG Jialin / ELEC Course: UROP 1100, Fall UROP 2100, Summer Pancreatic cancer, with a 10–12% five-year survival rate, is difficult to detect and progresses rapidly. MRI is vital for staging, treatment planning, and guiding MR-Linac therapies, but visualizing pancreatic tumors is challenging due to complex anatomy and subtle tumor appearance, requiring significant radiologist expertise. The scarcity of public pancreatic cancer MRI datasets hinders AI-driven solution development. Manual segmentation of diagnostic and real-time treatment images is time-consuming, increasing clinical workload. Robust AI models are needed to automate and enhance segmentation, improving treatment planning efficiency and accuracy. We, UROP students including LI Hongrui, are participating in the PANTHER Challenge to address these issues.
RkJQdWJsaXNoZXIy NDk5Njg=