Limits...
fMRI Activity in Posterior Parietal Cortex Relates to the Perceptual Use of Binocular Disparity for Both Signal-In-Noise and Feature Difference Tasks.

Patten ML, Welchman AE - PLoS ONE (2015)

Bottom Line: Binocular disparity is known to facilitate this process, and it is an open question how activity in different parts of the visual cortex relates to these fundamental visual abilities.To look for similarities between perceptual judgments and brain activity, we constructed 'fMR-metric' functions that described decoding performance as a function of signal magnitude.This highlights common stages of processing during perceptual performance on these tasks.

View Article: PubMed Central - PubMed

Affiliation: School of Psychology, University of Birmingham, Edgbaston, United Kingdom, B15 2TT; School of Psychology, UNSW Australia, Sydney, NSW, Australia.

ABSTRACT
Visually guided action and interaction depends on the brain's ability to (a) extract and (b) discriminate meaningful targets from complex retinal inputs. Binocular disparity is known to facilitate this process, and it is an open question how activity in different parts of the visual cortex relates to these fundamental visual abilities. Here we examined fMRI responses related to performance on two different tasks (signal-in-noise "coarse" and feature difference "fine" tasks) that have been widely used in previous work, and are believed to differentially target the visual processes of signal extraction and feature discrimination. We used multi-voxel pattern analysis to decode depth positions (near vs. far) from the fMRI activity evoked while participants were engaged in these tasks. To look for similarities between perceptual judgments and brain activity, we constructed 'fMR-metric' functions that described decoding performance as a function of signal magnitude. Thereafter we compared fMR-metric and psychometric functions, and report an association between judged depth and fMRI responses in the posterior parietal cortex during performance on both tasks. This highlights common stages of processing during perceptual performance on these tasks.

No MeSH data available.


Prediction accuracies for each region of interest.(A) The mean prediction accuracy of the classifier for the 100% signal condition in the signal-in-noise task. The horizontal red lines mark the baseline of statistical significance generated from permuting the data labels before being fed into the classifier. The location of the line indicates the upper 99.5% centile of the distribution of permuted data. (B) The mean prediction accuracy of the classifier for the 240 arcsec condition in the feature difference task. Again, the dotted horizontal lines indicate the cut off for statistical significance based on a permutation analysis. Error bars depict the SEM.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4631361&req=5

pone.0140696.g004: Prediction accuracies for each region of interest.(A) The mean prediction accuracy of the classifier for the 100% signal condition in the signal-in-noise task. The horizontal red lines mark the baseline of statistical significance generated from permuting the data labels before being fed into the classifier. The location of the line indicates the upper 99.5% centile of the distribution of permuted data. (B) The mean prediction accuracy of the classifier for the 240 arcsec condition in the feature difference task. Again, the dotted horizontal lines indicate the cut off for statistical significance based on a permutation analysis. Error bars depict the SEM.

Mentions: Fig 4A shows the between-subjects mean prediction accuracies obtained for the most discriminable stimulus configurations (100% signal) for each ROI. To establish a baseline for chance performance, and thereby judge responses that were statistically reliable, we ran the classification analysis with randomly permuted fMRI patterns (i.e., we randomized the correspondence between fMRI data and training labels and estimated the classifier prediction for each visual area) over 999 bootstrap iterations for the 100% signal condition. This created a distribution of classification accuracies, and we used the upper 99.5th centile (one-tailed, Bonferroni corrected) as our criterion for statistical significance (Fig 4, dotted lines). For all regions of interest, the median of the shuffled distribution was very close to 0.5 (range, 0.498–0.501) confirming our analysis technique to be unbiased. Considering responses across the sampled regions of interest, we were not able to reliably decode near vs. far depth differences in early visual area V1, ventral regions V3v and V4 or dorsal region hMT+/V5 based on activity measured using the event related fMRI design. However, we found that measured fMRI responses supported classification accuracies that exceeded the criterion for chance decoding in early visual area V2, ventral region LO and for all of the remaining measured dorsal and parietal visual areas.


fMRI Activity in Posterior Parietal Cortex Relates to the Perceptual Use of Binocular Disparity for Both Signal-In-Noise and Feature Difference Tasks.

Patten ML, Welchman AE - PLoS ONE (2015)

Prediction accuracies for each region of interest.(A) The mean prediction accuracy of the classifier for the 100% signal condition in the signal-in-noise task. The horizontal red lines mark the baseline of statistical significance generated from permuting the data labels before being fed into the classifier. The location of the line indicates the upper 99.5% centile of the distribution of permuted data. (B) The mean prediction accuracy of the classifier for the 240 arcsec condition in the feature difference task. Again, the dotted horizontal lines indicate the cut off for statistical significance based on a permutation analysis. Error bars depict the SEM.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4631361&req=5

pone.0140696.g004: Prediction accuracies for each region of interest.(A) The mean prediction accuracy of the classifier for the 100% signal condition in the signal-in-noise task. The horizontal red lines mark the baseline of statistical significance generated from permuting the data labels before being fed into the classifier. The location of the line indicates the upper 99.5% centile of the distribution of permuted data. (B) The mean prediction accuracy of the classifier for the 240 arcsec condition in the feature difference task. Again, the dotted horizontal lines indicate the cut off for statistical significance based on a permutation analysis. Error bars depict the SEM.
Mentions: Fig 4A shows the between-subjects mean prediction accuracies obtained for the most discriminable stimulus configurations (100% signal) for each ROI. To establish a baseline for chance performance, and thereby judge responses that were statistically reliable, we ran the classification analysis with randomly permuted fMRI patterns (i.e., we randomized the correspondence between fMRI data and training labels and estimated the classifier prediction for each visual area) over 999 bootstrap iterations for the 100% signal condition. This created a distribution of classification accuracies, and we used the upper 99.5th centile (one-tailed, Bonferroni corrected) as our criterion for statistical significance (Fig 4, dotted lines). For all regions of interest, the median of the shuffled distribution was very close to 0.5 (range, 0.498–0.501) confirming our analysis technique to be unbiased. Considering responses across the sampled regions of interest, we were not able to reliably decode near vs. far depth differences in early visual area V1, ventral regions V3v and V4 or dorsal region hMT+/V5 based on activity measured using the event related fMRI design. However, we found that measured fMRI responses supported classification accuracies that exceeded the criterion for chance decoding in early visual area V2, ventral region LO and for all of the remaining measured dorsal and parietal visual areas.

Bottom Line: Binocular disparity is known to facilitate this process, and it is an open question how activity in different parts of the visual cortex relates to these fundamental visual abilities.To look for similarities between perceptual judgments and brain activity, we constructed 'fMR-metric' functions that described decoding performance as a function of signal magnitude.This highlights common stages of processing during perceptual performance on these tasks.

View Article: PubMed Central - PubMed

Affiliation: School of Psychology, University of Birmingham, Edgbaston, United Kingdom, B15 2TT; School of Psychology, UNSW Australia, Sydney, NSW, Australia.

ABSTRACT
Visually guided action and interaction depends on the brain's ability to (a) extract and (b) discriminate meaningful targets from complex retinal inputs. Binocular disparity is known to facilitate this process, and it is an open question how activity in different parts of the visual cortex relates to these fundamental visual abilities. Here we examined fMRI responses related to performance on two different tasks (signal-in-noise "coarse" and feature difference "fine" tasks) that have been widely used in previous work, and are believed to differentially target the visual processes of signal extraction and feature discrimination. We used multi-voxel pattern analysis to decode depth positions (near vs. far) from the fMRI activity evoked while participants were engaged in these tasks. To look for similarities between perceptual judgments and brain activity, we constructed 'fMR-metric' functions that described decoding performance as a function of signal magnitude. Thereafter we compared fMR-metric and psychometric functions, and report an association between judged depth and fMRI responses in the posterior parietal cortex during performance on both tasks. This highlights common stages of processing during perceptual performance on these tasks.

No MeSH data available.