Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session B: Wednesday, August 13, 1:00 – 4:00 pm, de Brug & E‑Hall

A BOLD Sampling Scheme to Improve the Estimation of Voxel-wise Encoding Model Parameters

Jochem W. Rieger¹, Funda Yilmaz², Arkan Al-Zubaidi¹; ¹Carl von Ossietzky Universität Oldenburg, ²Donders Institute for Brain, Cognition and Behaviour

Presenter: Jochem W. Rieger

Encoding models are a widely used data driven technique to derive functional models like the sensitivity profiles of voxels to sets of stimulus features. They are typically estimated using linear regression techniques to find a model that uses sets stimulus features as input to predict fMRI BOLD activity of a voxel as output. The regression weights represent the functional model (e.g. a voxel’s receptive field). Sampling rate differences between stimulus and BOLD which can render the estimated model uninterpretable. This is typically counteracted by temporally down sampling the stimulus, which is undesirable as it causes information loss. Here we use simulations to first demonstrate that higher stimulus than fMRI sampling rates combined with regular BOLD sampling make the estimation of encoding models noisy and hard to interpret. We then demonstrate that a novel re-sampling based technique that samples the BOLD response at irregular temporal intervals alleviates these problems and allows to use the stimulus feature space with full temporal resolution for encoding model estimation.

Topic Area: Methods & Computational Tools

Extended Abstract: Full Text PDF