Blood, brains, (b-movies) and MVPA Alejandro (Sasha) Vicente Grabovetsky.

Blood, brains, (b-movies) and MVPAAlejandro (Sasha) Vicente Grabovetsky

Blood, brains and b-movies

4/10/2010Kamitani & Tong (2005)

Blood, brains and b-movies

4/10/2010Op de Beeck (2009) vs. Kamitani & Sawahata (2010)

Blood, brains and b-movies

4/10/2010

1. hyperacuity vs. coarse scale

2. voxel sampling of BOLD (compact support or spatio-temporal?)

3. relative sensitivity of mechanisms to High Frequency signal

Blood, brains and b-movies

4/10/2010Kriegeskorte et al. (2010)

Compact-kernel, SNR lower at high spatial frequency

A multipronged sensor samples various spatial frequencies in a complex manner

Blood, brains and b-movies

4/10/2010Kriegeskorte et al. (2010)

Blood, brains and b-movies

4/10/2010Kriegeskorte et al. (2010)

Venule samples different regions at different times of the HRFThis gives it a unique spatio-temporal signatureThis may contain high-res information including small imbalances in sampling of neuronal populations

Blood, brains and b-movies

4/10/2010Kriegeskorte et al. (2010)

Back to compact-kernels:•High spatial information could be aliased•But small head motion could completely modify the pattern of activity•Then MVPA should not work for train-test with head motion

Blood, brains and b-movies

4/10/2010Kriegeskorte et al. (2010)

Power of MVPA before and after smoothing if we assume different types of filters:

•point (compact)•box (compact)•gaussian (compact)•complex

Blood, brains and b-movies

4/10/2010Kriegeskorte et al. (2010)

As size decreases, partial volume effects become smaller

Not only for GM, WM, CSF; but also for cortical columns

Potentially, the power across voxels may increase with increased resolution, despite decreases in individual voxel SNR

Blood, brains and b-movies

4/10/2010Kriegeskorte et al. (2010)