School of Engineering Department of Computer Science and Engineering 101 Deep Learning for Medical Image Analysis Supervisor: CHEN Hao / CSE Student: DENG Yufan / SENG Course: UROP1000, Summer Deep-learning-based algorithms have done well in helping detect diseases from various medical images. However, medical image data to support large-scale model training are scarce and data augmentation methods for normal images fail to preserve medical meaning, which is extremely important. We proposed a synthetic data augmentation method that can preserve meaningful attributes of medical images using disentangled representation. Our method first uses encoders to extract domain-invariant features and domain-specific features between medical images of different qualities. Then, Generative Adversarial Networks (GANs) are used to generate new images that only keep the domain-invariant feature but with attributes from a new domain. Visualization of augmented data shows our method can preserve important feature of original images, and experiments on disease grading show our method have the potential to help improve grading accuracy. In the future, we will keep improving the method and evaluating it with more medical tasks. Deep Learning for Medical Image Analysis Supervisor: CHEN Hao / CSE Student: GUO Meichen / IPO Course: UROP1000, Summer Machine Learning has become a hot topic in high-tech world, with applications in various fields, especially in medical care. Core to many of these applications are visual recognition tasks such as image classification, localization and detection. Through the Undergraduate Research Opportunities Program, under the guidance of Professor Hao CHEN and Doctor Luyang LUO, I went deep into the field of deep learning. Specifically, I first studied some online materials that give a brief introduction to deep learning and machine learning. Then, I was assigned to read two papers on DualConsistency Semi-Supervised Learning and TransUNet, and try to run the relevant codes. I am writing this paper as a reflection.