BioVis Meetup Speakers
Zoom Logo
Join the meeting using Zoom: link
Meeting ID: 678 6390 4836
Passcode: BioV!s2025

BioVis Meetup #13: Building Explainable AI Tools for Biomedical Applications

Join us on Wednesday, 18th of February at 5pm Berlin / 4 pm London: “Building Explainable AI Tools for Biomedical Applications“ Grace Guo, Harvard University, https://gracegsy.github.io/

Abstract Advances in deep learning have yielded powerful models for diagnosis and prognosis, yet their “black-box” nature continues to stall clinical adoption—especially in biomedicine, where mistakes have life-changing consequences. Many prior studies have demonstrated how visualizations can be used to provide powerful explanations of generic AI models by translating abstract model mechanics into concrete, perceptual cues. However, biomedical data presents two key hurdles for explainable AI (XAI). Firstly, it spans diverse modalities: static and dynamic images, multi-channel time-series, genomic sequences, and free-text—all of which must be reconciled in a single, coherent explanation. Secondly, any explanation must map onto domain concepts that matter to clinicians; generic heat-maps rarely suffice and can even mislead when they are not contextualized in terms of medical expertise or domain knowledge. In this talk, I will present some of my recent work on domain-centered XAI tools for domain experts such as researchers, doctors, and clinicians. I will also discuss some of the key takeaways from these studies, as well as promising directions of future research leveraging visualizations for biomedical XAI.

Recorded talk

The recorded talk will be available. Recordings of previous meetups can be found at the BioVis meetup playlist on youtube

Future and past events