BEGIN:VCALENDAR
CALSCALE:GREGORIAN
VERSION:2.0
X-WR-CALNAME:BioVis meeup\, “Building Explainable AI Tools for Biomedica
 l Applications“\, \nGrace Guo\, Harvard University
METHOD:PUBLISH
PRODID:-//Apple Inc.//macOS 13.3//EN
BEGIN:VTIMEZONE
TZID:Europe/Berlin
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
DTSTART:19810329T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
TZNAME:CEST
TZOFFSETTO:+0200
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
DTSTART:19961027T030000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
TZNAME:CET
TZOFFSETTO:+0100
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
TRANSP:OPAQUE
DTEND;TZID=Europe/Berlin:20260218T180000
UID:1755001478777-64995@ical.marudot.com
DTSTART;TZID=Europe/Berlin:20260218T170000
LAST-MODIFIED:20251009T132615Z
DTSTAMP:20260213T125931Z
LOCATION:Join Zoom Meeting\nhttps://tu-dresden.zoom-x.de/j/67863904836?p
 wd=mVqoyB6DlxAbFAtN7kx0Ka62KHFmUJ.1\n\nMeeting ID: 678 6390 4836\nPassco
 de: BioV!s2025
SUMMARY:BioVis meeup\, “Building Explainable AI Tools for Biomedical App
 lications“\, \nGrace Guo\, Harvard University
SEQUENCE:1
DESCRIPTION:\n\nAbstract\nAdvances in deep learning have yielded powerfu
 l models for diagnosis and prognosis\, yet their “black-box” nature cont
 inues to stall clinical adoption—especially in biomedicine\, where mista
 kes have life-changing consequences. Many prior studies have demonstrate
 d how visualizations can be used to provide powerful explanations of gen
 eric AI models by translating abstract model mechanics into concrete\, p
 erceptual cues. However\, biomedical data presents two key hurdles for e
 xplainable AI (XAI). Firstly\, it spans diverse modalities: static and d
 ynamic 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 t
 o clinicians\; generic heat-maps rarely suffice and can even mislead whe
 n they are not contextualized in terms of medical expertise or domain kn
 owledge. In this talk\, I will present some of my recent work on domain-
 centered XAI tools for domain experts such as researchers\, doctors\, an
 d clinicians. I will also discuss some of the key takeaways from these s
 tudies\, as well as promising directions of future research leveraging v
 isualizations for biomedical XAI.
END:VEVENT
END:VCALENDAR