Explainable MRI in the Human Brain

Widely used MRI biomarkers for myelin and iron often lack rigorous ground-truth validation. To move beyond mere correlation, we combine quantitative post-mortem MRI of unfixed human brains with multimodal techniques, including mass spectrometry imaging, synchrotron small-angle X-ray scattering (SAXS), and 2D/3D microscopy. This approach allows us to map fundamental MRI parameters, such as relaxometry, susceptibility, and diffusion, directly to tissue microstructure and elemental composition.

The second part of this talk addresses the "black box" problem when deep learning is applied to neuroimaging. While deep neural networks achieve high accuracy in disease classification, they frequently exploit hidden data artifacts rather than identifying true pathology. We will discuss strategies to ensure algorithmic transparency and ensure that models rely on physiologically plausible features.

Neuroimaging Research Unit Medical University of Graz

The lecture will be held in sitem-insel (Room O2.211), Freiburgstr. 3, Bern followed by an Apéro and discussion and boroadcast virtually via Zoom at

https://unibe-ch.zoom.us/j/66256483922?pwd=aE1kTFg5aUZza1JmWEJxZDVsdnV5dz09

Please feel free to spread the word to anybody potentially interested.

For further info please contact: jessica.bastiaansen@unibe.ch

Join the mailing list here: TIC Seminar