Non-parametric mapping |

Notes

My non-parametric mapping software (NPM) is a open source tool for rapidly analyzing brain imaging data. Peer reviewed articles describe the VLSM and VBM functions. This software can use multiple CPUs simultaneously, and in practice is typically much faster than other programs. Furthermore, it offers you the ability to try new statistical tests such as the Brunner Munzel test and Firth's Penalized Logistic Regression.You can download NPM for Windows, Linux or Macintosh OSX,.

NPM is used to analyze brain imaging data. The test you will want to use depends on your images. If your images are binary (each voxel is either block or white, e.g. lesion maps where each voxel is either injured or spared), you will want to choose a command from the 'VLSM' menu. If your images are continuous (levels of gray, such as fractional anisotropy or gray matter probability maps) you will want to use the commands in the 'VBM' menu. I have written to brief tutorials for this software:- VBM anaylsis: This example shows analysis of images with continuous information (e.g. gray matter maps). NPM can perform t-tests, linear regression and Brunner Munzel analysis on this type of dataset. These functions are described in Rorden et al. (2007) Neuroimage, 35, 1531-1537
- VLSM anaylsis: This example shows analysis of images with binary information (e.g. lesion maps). NPM generates Liebermeister measure maps for binary behavioral data, and can generate either t-tests or Brunner-Munzel tests if the behavioral data is continuous. These features are described in Rorden et al. (2007) J Cog Neurosci, 19, 1081-1088

- VLSM : binary deficit - Liebermeister measure.
- VLSM : continuous deficit - t-test and/or the Brunner Munzel test (as specified in the Options/Tests menu).
- VLSM : multiple factors - Firth's Penalized Logisitic Regression.
- VBM : binary groups - t-test and/or the Brunner Munzel test (as specified in the Options/Tests menu).
- VBM : multiple regression - Weighted-least squares multiple linear regression (unthreaded, and therefore slow).
- VBM : paired t-test - repeated measures t-test.
- Options : permutations - decide whether you want to generate permutation thresholds (often less conservative than Bonferroni Correction for multiple comparisons).
- Options : tests - you can select the fast but parametric t-test and/or the the non-parametric Brunner Munzel test.
- Options : thread - choose number of CPU cores dedicated to data processing. For example, on a dual-core computer using two threads will be much faster than a single thread. For more than 2 cores, you may want to choose N-1 (e.g. 3 for a 4 core machine), which is a nice tradeoff between analysis speed and the responsiveness of your computer
- Utilities: variance image - Create a map of the variance in your images
- Utilities: variance image - Create a map of the variance in your images
- Utilities: binarized mean - All non-zero voxels in the input images are considered to have a value of one. The output image is the incidence of the input voxels being one (e.g. output voxels range from 0 if not they are zero in all the input images to 1 if non-zero in all input images).
- Utilities: Mean/StDev image - Create Mean and Standard Deviation maps of selected images
- Utilities: Mean/Single Subject Z-scores - Compares one individual to a group and identifies voxels that are unusually bright and dark.
- Utilities: Physiological Correction.
- Utilities: Count Lesion Overlaps - Randomly select groups of images from your population and determine how many unique tests will be required. This is useful for power analysis.