Hi all! My name is Jingwen (Lisa) Zhong. I’m a Data Science Specialist and Actuarial Science Major at UofT, graduating in 2026. I’m really happy and honored to have joined Prof. Tyrrell’s lab in the summer of 2024 as an ROP299 student. This was my first research project, and it has truly exercised many of my research and scientific skills, such as literature review, critical thinking, and the ability to get familiar with a brand-new field.
Coming into the lab, I had no research experience and no prior knowledge of medical imaging. As a student just finishing my second year of study, I felt curious about machine learning and artificial intelligence because these topics are so widely discussed. However, I still can’t forget how uneasy I felt during the first few weeks as I tried to think of a research question related to medical images and machine learning. I’m incredibly thankful to Prof. Tyrrell, who ‘relentlessly’ pointed out issues during each lab meeting, and to the lab volunteers, Daniel and Atshuhiro, who were always willing to help and guide me through the process. I couldn’t have gotten my project ready for implementation without their support. After a month of struggle, I finally settled on my research topic: investigating whether LPIPS is a better metric for assessing the similarities of medical images compared to PSNR and SSIM under various degradation conditions.
Having a research question is just the beginning; implementing it is another huge mountain to climb. I remember how excited I was when my research question was finally approved. I worked hard that week to implement almost all the code for my project. If I could go back, I would approach this differently. Instead of diving straight into coding, I would first take the time to design the entire study process—splitting the dataset, testing the code on a smaller dataset, figuring out how to use the GPU, then applying the code to the full dataset, and finally choosing the appropriate statistical analysis. I say this because I stumbled at each of these steps. After completing my code, I found that it ran so slowly that it would take several days to get results. So, I began the process of figuring out how to set up the environment to run on the lab’s GPU. This process took me almost two weeks, but with the help of other ROP students, I finally got the code running on the GPU.
Once the GPU problem was solved, my results came in much faster. However, the next obstacle was interpreting these results. As a Data Science student, it’s hard to admit, but I hadn’t yet learned ANOVA. Initially, I turned to ChatGPT for help, but the results weren’t ideal. Prof. Tyrrell suggested that I use SAS to perform ANOVA, which provided me with ideal and comprehensive results. So, I learned how to use SAS—a very powerful statistical analysis tool compared to Python.
Through this ROP experience, I learned the importance of communication and teamwork. Although we worked on different projects, the weekly lab meetings were incredibly helpful. It was a place where everyone’s intelligence came together, and I always left with new insights and a clear plan in mind.
Overall, this journey has been a steep learning curve but an immensely rewarding one. I am grateful for the opportunity to work with such a supportive team, and I know that the skills and lessons I’ve learned will continue to guide me in my future research endeavors.