WiM Presents: Building Trustworthiness through Robustness

Wednesday 20th November 2024

Women in MICCAI (WiM) is happy to announce the second invited talk in the WiM "Health Equity in the AI Era" Webinar Series 2024

Building Trustworthiness through Robustness

Speaker: Dr. Maria A. Zuluaga, EURECOM, King's College London

Monday, December 9, 2024
4:00 - 5:00 PM UTC / 8:00 - 9:00 AM PST / 11:00 AM – 12:00 PM EST

Trustworthiness is paramount in the development and deployment of AI systems. A critical aspect of achieving trustworthy AI lies in its robustness—the ability of a model to maintain performance under varying conditions. This talk will delve into the importance of robustness in ensuring trustworthy AI, with a particular focus on two key areas: distribution shifts and input data variations.

Distribution shifts, a common and well-studied challenge in AI, occur when the distribution of test data drifts from that of training. We will discuss techniques to enhance model robustness against such shifts, when dealing with multi-center, multi-modal and multi-organ datasets. Additionally, we will explore the less-studied area of input data variations, such as missing values or modalities. While the standard approach is to impute the missing data, we argue this may be problematic in certain scenarios that we will discuss. Then, we will present our research on developing robust models that can effectively handle missing data without imputation, leading to more reliable, resilient and, overall, more trustworthy AI systems. We will illustrate our work in different medical imaging applications.

Registration (required) is free and open to everyone.

About the Speaker:

Maria A. Zuluaga is a senior lecturer at the Data Science department at EURECOM with an affiliate position within the School of Biomedical Engineering & Imaging Sciences at King’s College London (Visiting Senior Lecturer). She obtained her BSc in Electronics Engineering and MSc in Computer Science degrees at Universidad del Valle and Universidad de los Andes, respectively, in Colombia. She holds a PhD in Signal and Image Processing from Université de Lyon. Her research team focuses on the development of novel machine learning methods from multi-modal data that can be safely used to advance healthcare research and improve clinical practice. From an application standpoint, her research work aims to answer questions coming from cardiovascular and neurovascular imaging, as well as cancer research.