Parameters¶
| Name | Description |
|---|---|
| Number of PCs | The number of principal components (PCs) kept after performing PCA. It should be more than the Number of ICs. |
| Number of ICs | The number of independent components (ICs) kept after performing ICA. |
| Unmixing type | How to apply the ICA output matrix to the principal components. See ICA Unmixing Dimension for more technical details. |
| ICA Temporal Weight | A weight coefficient for the contribution of the temporal versus spatial information in independent component analysis (ICA). A value of 0 means that only spatial information is used. A value of 1 means that only temporal information is used. See spatiotemporalMixing for more technical details. |
| Maximum Iterations | An upper limit for the number of iterations ICA can perform. |
| Convergence Threshold | The threshold value of the error of the ICA solution that forces the algorithm to stop early. |
| Block Size | The number of frames to store temporarily in memory when performing calculations. Increasing this may reduce run time at the cost of physical memory (RAM) usage. The default value of 1000 should be appropriate in most cases. |
| Average cell diameter | The expected diameter of cells in your data. This parameter can be estimated by drawing a line across a representative cell. Note that the average cell diameter may be different from one dataset to another depending on factors such as the type of GRIN lens used, brain area, cell type, and any spatial downsampling applied to the movie. |
ICA Unmixing Dimension¶
Temporal¶
Applies the ICA matrix to the PCA traces to get the final traces.
Use this option if you want traces with a relatively high signal-to-noise ratio, are mostly
interested in using them for eventDetection, and are not concerned that they are
always uncorrelated.
Spatial¶
Applies the ICA matrix to the PCA images to get the final images. Use this option if you want traces that are not necessarily uncorrelated, but are typically noisier than those obtained using Temporal, and are interested in images with a relatively high signal-to-noise ratio.
Both¶
Applies the ICA matrix to both the PCA traces and PCA images to get the final traces and images. Use this option if you want traces and images with the highest signal-to-noise ratios and are not concerned that the traces will always be uncorrelated.
Warning
This algorithm has many settings and it is often important to explore variations in their values to get satisfactory identification of the image and trace components. This is particularly true for the Number of PCs and the Number of ICs. If the Number of ICs is too low, some ICs will likely correspond to multiple cells. If it is too high, some cells will likely be split into multiple ICs.
A good starting point is to try to estimate the number of neurons in
the movie to set the number of ICs and then add 15-20% to this number to get the number of PCs.
By default, the number of PCs is set to be 20% higher than the number of ICs. See estimateICs to estimate the number of cells using IDPS.
Notes about IC estimation
- This function uses the blob detection algorithm from OpenCV.
- The estimated number of cells is automatically scaled by 20% to obtain a better estimate for the number of ICs to be used in the PCA-ICA algorithm.