Quality Improvements for MultiModal Neuroimaging Mika Sepp Brain

Quality Improvements for Multi-Modal Neuroimaging Mika Seppä Brain Research Unit / LTL AMI-Centre Helsinki University of Technology

MS/2007

3 central stages: 1. Registration 2. Resampling 3. Visualization MS/2007

1. Registration w Match metric w Single-modal registration – sum of squared differences w Multi-modal registration – mutual information MS/2007

MI artifacts w Registration bias w Local maxima MI w Images aligned w Same pixel/voxel sizes w Continuous sampling – higher-order method offset (pixels) MS/2007

2. Resampling w Very common operation – geom. correction, motion correction, group average etc. . MS/2007 SNR = 10*log 10( s 2/ e 2) 15 x

SNR (d. B) Two-stage resampling MS/2007 Time (ms)

3. Visualization 800 fps 0. 1 -5 fps MS/2007 120 fps

MS/2007

MS/2007

MS/2007

Seppä (2007). On continuous sampling and sub-sample accuracy in mutual-information registration. TKK Report, TKKKYL-017. Seppä (2007). High-quality two-stage resampling for 3 -D volumes in medical imaging. Medical Image Analysis, 11: 346360. Seppä, Hämäläinen (2005). Visualizing human brain surface from T 1 -weighted MR images using texture-mapped triangle meshes. Neuro. Image, 26: 1 -12. Hiltunen, Suortti, Arvela, Seppä, Joensuu, Hari (2005). Diffusion tensor imaging and tractography of distal peripheral nerves at 3 T. Clinical Neurophysiology, 116: 2315 -2323. MS/2007

MS/2007