t-distributed stochastic neighbor embedding (t-sne) using the jaccard distance algorithm  (MathWorks Inc)

Journal: Cerebral Cortex (New York, NY)

Article Title: To be and not to be: wide-field Ca 2+ imaging reveals neocortical functional segmentation combines stability and flexibility

doi: 10.1093/cercor/bhac523

Figure Lengend Snippet: Segmentation of Ca 2+ imaging data reveals spatially similar ICs across time segments and scales. A) Schematic showing how data is parsed into equal length time segments (set of images show data time segments) for different timescales which are run through the JADE ICA algorithm independently to obtain new ICA solutions (lower images; each color represents an IC). B) Example template map based on an entire mouse’s dataset and time segment ICA map at the day timescale (left; colors show individual ICs) and Jaccard index matrix showing matching (high value) and nonmatching (low value) ICs between a time segment and template ICA map (circles). C) Examples of overlap between matching and nonmatching pairs of template and time segment ICs for circles shown in the Jaccard matrix of B (overlap shown in dark red; individual ICs blue or yellow). D) Three examples of template matching ICs (Jaccard index ≥0.5) for each of the four timescales (columns/green bars). Scale bars: 1 mm.

Article Snippet: Dimensionality reduction of the IC libraries was achieved with t -distributed stochastic neighbor embedding ( t -SNE) using the Jaccard distance algorithm, plotting the position of each IC in this space as a point in a 2D graph (Matlab 2019 tsne function).

Techniques: Imaging

Journal: Cerebral Cortex (New York, NY)

Article Title: To be and not to be: wide-field Ca 2+ imaging reveals neocortical functional segmentation combines stability and flexibility

doi: 10.1093/cercor/bhac523

Figure Lengend Snippet: Spatial ICA of wide-field Ca 2+ imaging data produces spatially independent brain regions. A) Schematic showing the ICA workflow of concatenating data for each animal chronologically across days and trials (days are signified by different colored borders; trials are signified by overlapping images) and sending the combined dataset through the JADE ICA algorithm. B) Example template map (ground-truth to which all other ICA solutions are compared) of spatial ICs produced from running ICA on one mouse’s combined dataset (each different colored region is a single IC; scale bar: 1 mm). White lines denote major regions of the Allen Common Coordinate Framework (CCF; see ). C) Example matrix of Jaccard indices comparing the template map to itself (low off-diagonal Jaccard indices indicate good spatial separation; zero values are shown as white indicating no IC overlap). D) Frequency histograms showing the distribution of off-diagonal Jaccard indices (nonself matches) when comparing the template map to itself for each animal (bin-widths = 0.05).

Article Snippet: Dimensionality reduction of the IC libraries was achieved with t -distributed stochastic neighbor embedding ( t -SNE) using the Jaccard distance algorithm, plotting the position of each IC in this space as a point in a 2D graph (Matlab 2019 tsne function).

Techniques: Imaging, Produced

Journal: Cerebral Cortex (New York, NY)

Article Title: To be and not to be: wide-field Ca 2+ imaging reveals neocortical functional segmentation combines stability and flexibility

doi: 10.1093/cercor/bhac523

Figure Lengend Snippet: Cortex-wide maps cover similar areas across timescales. A) Brain maps from all six experimental subjects showing cumulative cortical coverage of template matching ICs across all time-windows within each of the four timescales examined (color scale shows the number of timescales where an area of cortex was covered by a spatial IC). Scale bar: 1 mm. B) Jaccard index matrices for each of the six experimental subjects showing a high degree of overlapping cortical coverage between timescales (off-diagonal comparisons).

Article Snippet: Dimensionality reduction of the IC libraries was achieved with t -distributed stochastic neighbor embedding ( t -SNE) using the Jaccard distance algorithm, plotting the position of each IC in this space as a point in a 2D graph (Matlab 2019 tsne function).

Techniques: