BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Contributed Talk: Resolution-adaptive networks for uniform perform
 ance across heterogeneous medical image cohorts - Richard McKinley (Insels
 pital\, University Hospital of Bern)
DTSTART:20260211T100000Z
DTEND:20260211T103000Z
UID:TALK242326@talks.cam.ac.uk
DESCRIPTION:In the setting of clinical imaging\, differences in between ve
 ndors\, hospitals and sequences can yield highly inhomogeneous imaging dat
 a. In MRI imaging in particular\, voxel dimension\, slice spacing and acqu
 isition plane can vary dramatically .The usual strategy to deal with heter
 ogeneity of resolution is harmonization by resampling to a common spatial 
 resolution. This can lead to loss of fidelity arising from interpolation a
 rtifacts out-of-plane and downsampling in-plane. We propose a network arch
 itecture designed to be able to learn directly from spatially heterogeneou
 s data\, without resampling: a segmentation network based on the e3nn fram
 ework that leverages a spherical harmonic\, rather than voxel-grid\, param
 eterization of convolutional kernels\, with a fixed physical radius.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR