Breakfast & Science Seminar 50

Abstract

Benchmarks in MRI reconstruction are dominated by data-driven methods, yet their clinical adoption remains limited due to trust issues. In this talk, we explore strategies to build confidence in these reconstructions by combining data-driven techniques with interpretable, model-based approaches rooted in MRI physics. We will highlight recent advances in algorithm design that integrate uncertainty quantification, allowing us to assess and communicate the reliability of the images. By merging the strengths of data-driven learning with the transparency of model-based methods, our approach enhances both image quality and diagnostic trust. This presentation will cover key developments, experimental insights, and the potential clinical impact of these techniques.

About the Speaker

Martin Zach is a postdoctoral researcher at CIBM SP EPFL Section under the supervision of Prof. Michael Unser. He specializes in inverse problems in imaging, with a strong focus on biomedical imaging, particularly MRI reconstruction. His research emphasizes a balanced approach to model-based reconstructions, integrating data-driven approaches where appropriate. Martin received his PhD in August 2024 from Graz University of Technology under the supervision of Thomas Pock. His work combines theoretical insights with practical applications, contributing to advancements in biomedical imaging techniques.