School of Engineering Department of Computer Science and Engineering 97 The Future of Medical Imaging: Advancements in Analysis through Vision Language and Large Models Supervisor: CHEN Hao / CSE Student: LIN Qianwei / COMP Course: UROP 1000, Summer Our research project focuses on exploring the application of vision-language models and large language models in medical image analysis. My role has been to conduct in-depth research on the paper "Towards Generalist Biomedical AI", which provides insights into similar models developed by our team. Furthermore, I had to annotate our model's answers to medical problems as a non-medical professional. Besides, I had to develop professional annotation interfaces using Label Studio platform and Gradio library and set up a server connection to allow doctors to provide feedback on our model's performance. I also explored the data crawling capabilities of crawl4ai to generate medical datasets, contributing to model improvements. Through this work, I have gained a deeper understanding of the transformative impact of generalist foundation AI models on the medical field. The Future of Medical Imaging: Advancements in Analysis through Vision Language and Large Models Supervisor: CHEN Hao / CSE Student: WEI Yuhan / COMP Course: UROP 1000, Summer The research explores the possibility of replacing real image data with synthetic image data, trying to address the problem of scarcity of medical data and avoid time-consuming and labor-intensive manual annotation. We leverage the Gaussian distribution characteristics of medical images and use mixed Gaussian rendering to generate synthetic data as well as labels. Then the models are trained to decouple Gaussian functions and lay the foundation for analyzing real medical images. Experiment results show promising performance in pretraining for CT image segmentation tasks, and we will continue to test its performance in different datasets, using different data sizes and modalities and so on. Diffusion Model for Classification Tasks Supervisor: CHEN Long / CSE Student: LI Jieru / DSCT Course: UROP 1100, Summer With the development of computer science, how to represent, process and display 3D information effectively in computer becomes an important issue in the study of computer application. Computer graphics then comes into being for the creation, manipulation and edition of virtual images. It is widely used in model construction, computer art and visualization. One of the basic techniques in computer graphic is rendering, which generates photo-realistic image format. This report will discuss two important rendering methods in details, deduce the algorithms of shading and introduce the basic theory of 3D reconstruction, which is the promising application of computer graphic. By explaining this processes, this report aims to help beginners understand how computer graphics works and how to make virtual images seem more realistic.
RkJQdWJsaXNoZXIy NDk5Njg=