Talks and presentations

From voxels to patients: A data-centric journey towards consistent translational AI

December 04, 2023

Invited talk, Vanderbilt University seminar, Nashville, USA

Lianrui Zuo
Presented at Vanderbilt University ECE & VISE seminar
Description: In the pursuit of advanced patient care, medical imaging serves as an indispensable tool for understanding individual health and disease. However, the potential of this tool is often compromised by undesired variability due to the heterogeneity of data acquisition, posing significant challenges for medical professionals and translational AI algorithms. In the talk “From Voxels to Patients”, I will explore our approach and commitment to addressing these challenges, aiming to enhance the consistency of imaging data for improved patient care and analysis. I will present how our data-centric approaches can effectively mitigate undesired variations in medical data, thereby enriching both research and clinical applications of medical imaging. Our initiatives will be illustrated through a multi-site clinical trial for multiple sclerosis, where we demonstrate our comprehensive strategies: harmonizing images across diverse platforms, integrating multi-modal imaging, and managing varied data quality and availability. Audience will gain a deeper understanding of how these advanced techniques not only elevate the consistency and reliability of medical image analysis but also offer grand opportunities for improved patient care.
Video: Coming soon

May 31, 2023

Conference talk, Consortium of Multiple Sclerosis Centers, Aurora, Colorado

Authors: Lianrui Zuo, Savannah P. Hays, Blake E. Dewey, Samuel W. Remedios, Yuan Xue, Sandra D. Cassard, Carolyn Koch, Ann Fishman, Aaron Carass, Jerry L. Prince, Ellen M. Mowry, and Scott D. Newsome
Presented at Consortium of Multiple Sclerosis Centers, 2023
Description: Magnetic resonance imaging (MRI) is a powerful and flexible imaging technique in studing multiple sclerosis (MS). However, the flexibility of MRI often comes at a cost: lack of standardization and consistency across imaging studies. Despite the guidelines provided by the MAGNIMS–CMSC–NAIMS, there are still significant variations of MR acquisition in research and clincal settings. The variablity in MR acquisition poses challenges for consistent and accurate study of MS. In our study, we investigate the impact of inconsistent MR acquisition in longitudinal volumetric analysis. We found that inconsistent acquisition has a strong negative impact on the consistency and statistical power of longitudinal analysis, thus hinders our understanding of MS. We then investigate MR harmoniation as a potential solution to mitigate this issue. Our study suggests that image harmonization effectively alleviates this negative impact, yielding nearly identical consistency as that in standardized and consistent acquisitions.
Video:
Poster:

March 01, 2013

Tutorial, UC-Berkeley Institute for Testing Science, Berkeley CA, USA

March 01, 2012

Talk, UC San Francisco, Department of Testing, San Francisco, California

This is a description of your talk, which is a markdown files that can be all markdown-ified like any other post. Yay markdown!