Journal: Scientific Reports

Article Title: Multivoxel Pattern of Blood Oxygen Level Dependent Activity can be sensitive to stimulus specific fine scale responses

doi: 10.1038/s41598-020-64044-x

Figure Lengend Snippet: Volumetric and Surface Grid Driven Misalignment. 2-dimensional cartoon representation of the 2 types of artificial misalignments implemented. The curved black lines represent a portion of the cortical sheet. The largest grey curved line, tangential to the cortical sheet, represents a large vessel on the pial surface, while the other curved grey line represents a penetrating draining vessel. The thinner, wavy grey lines represent smaller vessels such as capillaries. The pale red, green and blue colored squares symbolize voxels belonging respectively to inner, mid and outer layers, and the solid green squares represent voxels in the middle layers belonging to the retinotopic representation of the occluded bottom right hand quadrant of the visual field (i.e. the ROI that will be misaligned). Note that while we segmented the cortical sheet into 6 layers, this figure, for simplicity, only shows 3 layers. Panel a) depicts the original, unshifted ROI. Panel b) shows the effect of 1 voxel volumetric misalignment. The arrows on the top right corner represent all possible directions along which misalignment occurs. The yellow arrow indicates the dimension along which misalignment has occurred (in this case, to the right along the x axis). Panel c) is identical to panel b) apart from the dimension along which the training ROI was misaligned (i.e. downwards along the y axis). Panel d) portrays an example of 1 voxel surface grid driven misalignment. In panel d) there is only one yellow arrow on the top right corner because surface grid driven misalignment only occurs along a given layer. Note that: 1. in volumetric misalignment, only the misaligned ROI will include voxels from neighboring layers (panels b and c); 2. during volumetric misalignment, the training ROI can be shifted along large penetrating vessels that are orthogonal to the cortical surface (panel c); and 3. the distance between neighboring voxels within a layer is variable: it is equal to 0.8 mm (i.e. the length side of a voxel) if the voxels lie on the same plane, and 1.13 mm (i.e. the length of the diagonal of a voxel) if they do not. Within the surface grid driven regime, therefore, a 1 voxel shift can lead to a misalignment that is greater than the nominal voxel resolution.

Article Snippet: For each of the 3 classifiers, to test the main effects and interactions between signals (feedforward and feedback), misalignment extents (1 to 6) and cortical depths (1 to 6), we performed a 2 (signals) by 6 (cortical depths) by 6 (voxel shifts) Linear Mixed Model (Matlab, The Mathworks Inc, 2014) with the accuracy scores as a dependent variable, as explained above.

Techniques:

Journal: Scientific Reports

Article Title: Multivoxel Pattern of Blood Oxygen Level Dependent Activity can be sensitive to stimulus specific fine scale responses

doi: 10.1038/s41598-020-64044-x

Figure Lengend Snippet: Volumetric misalignment. Decoding accuracy as a function of voxel misalignment for the feed-forward (left panels) and feed-back (right panels) signals for SVM, LDA and NBC. Different colors represent different cortical depths as indicated in the figure legend. Error-bars represent the hierarchical 95% bootstrapped confidence interval across subjects (see methods).

Article Snippet: For each of the 3 classifiers, to test the main effects and interactions between signals (feedforward and feedback), misalignment extents (1 to 6) and cortical depths (1 to 6), we performed a 2 (signals) by 6 (cortical depths) by 6 (voxel shifts) Linear Mixed Model (Matlab, The Mathworks Inc, 2014) with the accuracy scores as a dependent variable, as explained above.

Techniques:

Journal: Scientific Reports

Article Title: Multivoxel Pattern of Blood Oxygen Level Dependent Activity can be sensitive to stimulus specific fine scale responses

doi: 10.1038/s41598-020-64044-x

Figure Lengend Snippet: Iso-surface grid driven misalignment. Decoding accuracy as a function of voxel misalignment for the feed-forward (left panels) and feed-back (right panels) signals for SVM, LDA and NBC. Different colors represent different cortical depths, as indicated in the figure legend. Error-bars represent the hierarchical 95% bootstrapped confidence interval across subjects (see methods).

Article Snippet: For each of the 3 classifiers, to test the main effects and interactions between signals (feedforward and feedback), misalignment extents (1 to 6) and cortical depths (1 to 6), we performed a 2 (signals) by 6 (cortical depths) by 6 (voxel shifts) Linear Mixed Model (Matlab, The Mathworks Inc, 2014) with the accuracy scores as a dependent variable, as explained above.

Techniques: