Hanke, M. | Search bioRxiv

bioRxiv2018

The effect of acquisition resolution on orientation decoding from V1: comparison of 3T and 7T

Previously published results indicate that the accuracy of decoding visual orientation from 7T fMRI data of V1 peaks at spatial acquisition resolutions that are routinely accessible with more conventional 3T scanners. This study directly compares the decoding performance between a 3T and a 7T dataset that were acquired using the same stimulation paradigm by applying an identical analysis procedure. The results indicate that decoding models built on 3T data are comparatively impaired. Moreover, we found no evidence for a strong coupling of BOLD signal change magnitude or temporal signal to noise ratio (tSNR) with decoding performance. Direct enhancement of tSNR via multiband fMRI acquisition at the same resolution did not translate into improved decoding performance. Additional voxel selection can boost 3T decoding performance to the 7T level only at a 3mm acquisition resolution. In both datasets the BOLD signal available for orientation decoding is spatially broadband, but, consistent with the size of the BOLD point-spread-function, decoding models at 3T utilize spatially coarser image components.

neuroscience↗

bioRxiv2016

The effect of acquisition resolution on orientation decoding from V1 BOLD fMRI at 7 Tesla

A decade after it was shown that the orientation of visual grating stimuli can be decoded from human visual cortex activity by means of multivariate pattern classification of BOLD fMRI data, numerous studies have investigated which aspects of neuronal activity are reflected in BOLD response patterns and are accessible for decoding. However, it remains inconclusive what the effect of acquisition resolution on BOLD fMRI decoding analyses is. The present study is the first to provide empirical ultra high-field fMRI data recorded at four spatial resolutions (0.8 mm, 1.4 mm, 2 mm, and 3 mm isotropic voxel size) on this topic -- in order to test hypotheses on the strength and spatial scale of orientation discriminating signals. We present detailed analysis, in line with predictions from previous simulation studies, about how the performance of orientation decoding varies with different acquisition resolutions. Moreover, we also examine different spatial filtering procedures and its effects on orientation decoding. Here we show that higher-resolution scans with subsequent down-sampling or low-pass filtering yield no benefit over scans natively recorded in the corresponding lower resolution regarding decoding accuracy. The orientation-related signal in the BOLD fMRI data is spatially broadband in nature, includes both high spatial frequency components, as well as large-scale biases previously proposed in the literature. Moreover, we found above chance-level contribution from large draining veins to orientation decoding. Acquired raw data were publicly released to facilitate further investigation.

neuroscience↗