Yiran (Lexie) Sun

I am a third-year Ph.D. student at the Department of Electrical and Computer Engineering at Rice Unversity, as well as the Department of Imaging Physics at MD Anderson Cancer Center, where I work on designing generative models in medical imaging analysis. My Ph.D. advisors are Dr. Osama R. Mawlawi and Dr. Guha Balakrishnan.

I am also a member of the Digital Health Initative in Rice ECE.

I have an MS in Data Analytics and Statistics from Washington University in St. Louis, where I was a research assistant for Dr. Ulugbek Kamilov and Dr. Yu Sun in Computational Imaging Group.

For Rice undergraduate students, I am happy to mentor you and I encourage you also to explore available research programs.

Inspired by Dr. Jia-Bin Huang, I set aside some time every week to meet with people (preferrably students from underrepresented groups). Feel free to send me an email to schedule a meeting if you would like to chat with me on any topics.

Email  /  GitHub  /  Google Scholar  /  LinkedIn

Selected Research

I'm interested in developing generative models with applications in 3D tomographic reconstruction. I mostly focus on Computed Tomography (CT), Positron Emission Tomography (PET).

Towards Uncertainty Estimation in PET Partial Volume Correction with a Diffusion Probabilistic Model

We estimate the statistical and systematic uncertainty of resultant images from a diffusion probabilistic model when applied to partial volume correction of brain PET images (DiffPVC).

Can Deep Learning Techniques Bridge the Disparity in Image Quality between Low and High-Performance PET Scanners

We present a diffusion model-based pipeline to improve the quality of PET images from low performing (LP) PET scanners and compare the results to high performing (HP) systems.

DiffPET: A Fine Tuned Sinogram-to-PET Conditional Diffusion Model

We propose a new simple but effective pipeline that utilize a second neural network to finetune the pre-trained cDDPM to generate high fidelity and realistic PET images from sinogram (w/ MRI optionally) inputs.

Partial Volume Correction in Brain PET Imaging Using Swin Transformer

We evaluate the feasibility and performance of a supervised deep learning approach that directly corrects PVE using a Swin Transformer. We also test on out-of-distribution (OOD) datasets to evaluate the robustness of our method.

R2U-DDPM: A Conditional Diffusion Model with R2U-Net Guidance for Realistic PET Image Synthesis

We propose a new robust pipeline (R2U-DDPM) that utilizes a pre-trained attention R2U-Net as an auxiliary prior to inform cDDPM about the coarse structure of reconstructed PET images to increase model reliability and generate realistic results from sinogram inputs.

CT Reconstruction from Few Planar X-Rays with Application Towards Low-Resource Radiotherapy

We propose a deep generative model, building on advances in neural implicit representations to synthesize volumetric CT scans from few input planar X-ray images at different angles.

Async-RED: A Provably Convergent Asynchronous Block Parallel Stochastic Method Using Deep Denoising Priors

We propose a new asynchronous RED (Async-RED) algorithm that enables asynchronous parallel processing of data, making it significantly faster than its serial counterparts for large-scale inverse problems. The computational complexity of Async-RED is further reduced by using a random subset of measurements at every iteration.

Award

IEEE International Conference on Computational Photography (ICCP) Student Travel Grants, 2022

Rice University John Clark Jr. Fellowship Award, 2021

Acknowledgment

Design and source code from Jon Barron and Leonid Keselman's websites.