BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Data-Driven Approaches for Addressing Head Motion Bias in Structur
 al MRI Brain Morphometry - Ruben Klinger\, Technical University of Munich
DTSTART:20260330T115000Z
DTEND:20260330T123000Z
UID:TALK246061@talks.cam.ac.uk
CONTACT:Dace Apšvalka
DESCRIPTION:*Please note a later start! 12:50* (after a project presentati
 on). \n\n*Speaker:* Ruben Klinger\, Technical University of Munich.\n\n*Ti
 tle:* Data-Driven Approaches for Addressing Head Motion Bias in Structural
  MRI Brain Morphometry\n\n*Abstract:* Head motion during structural Magnet
 ic Resonance Imaging (MRI) introduces systematic biases\, such as artifici
 al cortical thinning\, which confound morphometric analyses in clinical an
 d aging studies. This study evaluates two distinct data-driven approaches 
 to mitigate this motion bias: including a data-quality metric as a nuisanc
 e covariate in downstream statistical modeling and retrospective k-space m
 otion correction. First\, for the covariate approach\, a novel deep-learni
 ng-based image motion score is compared against the traditional Euler numb
 er. Results demonstrate that the deep learning metric more effectively cap
 tures motion variance between groups and therefore provides a greater corr
 ective effect when used as a nuisance covariate. However\, because this ap
 proach cannot recover lost anatomical information\, a purely data-driven k
 -space correction method (MotionTTT) is also evaluated. By estimating moti
 on trajectories directly from undersampled data and then correcting for th
 is motion during image reconstruction\, MotionTTT successfully recovered s
 everely corrupted scans that typically fail automated segmentation pipelin
 es and eliminates up to 0.3 mm of artificial cortical thinning bias in mor
 phometric measurements\, returning measurements to levels statistically eq
 uivalent to uncorrupted ground-truth data. In conclusion\, while the stati
 stical covariate approach offers a highly practical solution for mitigatin
 g bias in legacy datasets\, retrospective k-space correction represents th
 e gold standard for prospective studies\, as it prevents the exclusion of 
 patient data and restores true anatomical information. \n\n*Bio:* I am a M
 aster's student in Robotics\, Cognition\, and Intelligence at the Technica
 l University of Munich (TUM)\, currently completing a research internship 
 at the MRC CBU under Dr. Marta Correia. I also hold a B.Sc. in Engineering
  Science from TUM. \n\n*Venue:* MRC CBU West Wing Seminar Room and Zoom ht
 tps://us02web.zoom.us/j/82385113580?pwd=RmxIUmphQW9Ud1JBby9nTDQzR0NRdz09 (
 Meeting ID: 823 8511 3580\; Passcode: 299077)\n
LOCATION: MRC-CBU\, 15 Chaucer Road\, Cambridge
END:VEVENT
END:VCALENDAR