Synopsis of the WAIR 2.0 Commands

The WAIR 2.0 software is a tool for automated, fast and robust quantitative analysis of various image-registration (warping) techniques applied to a single or multiple data sets. Warp performance is evaluated using the wavelet transforms of the warped-resliced data. Several different schemes for measuring goodness of image registration in "compressed" wavelet space are employed. The main routines in the WAIR 2.0 package are WT_IWT, wave_space_varThresh, CGC_WTC, SGC_CWT and triangleWTC. The first one is used to find the (multi-dimensional) discrete wavelet and the inverse transforms of the resliced data. The second is employed to obtain the thresholded signals, in wavelet space, which are used by the following routines to numerically characterize warp performance based on different classification approaches.

Before invoking the wavelet-based warp classification programs one needs to obtain the WT's of the data (using WT_IWT) and the thresholded WT's (using wave_space_varThresh). This last routine, wave_space_varThresh is used interactively and allows choosing one of three wavelet shrinkage schemes: Uniform (at different levels), Donoho-Johnstone (DJ) and Dinov-Sumners (DS) thresholding. For more advanced users we recommend the use of the automated thresholding routine wave_space_varThresh_inter.
One way to numerically analyze warp performance on a single or multiple data is the "wavelet-space triangle" scheme. It is quite different from the CGC and SGC techniques (see the technical notes) because it uses the original data prior to warping and the target of the warp in determining warp ranking in reduced wavelet space. The "triangle" method is applied by calling the procedure triangleWTC. To quantitatively evaluate warp performance using functional (e.g., PET) data one employs the routine "SGC_CWT". This technique ranks alignment algorithms based on how far apart they warp groups (different activation paradigms) of data (across subjects). To evaluate image registration using the same SGC (Spread Group Classification) scheme without thresholding (equivalent to spatial domain analysis) one omits the wavelet-thresholding step (wave_space_varThresh) of the analysis.

There are several auxiliary routines in WAIR 2.0 that are frequently used for visual interpretation of low-dimensional (1D, 2D) wavelet shrinkage and compression. The program MultiDim_IWT_varThresh, can be employed for inverting the wavelet transform directly (without going through "wave_space_varThresh") using one of three different wavelet-thresholding approaches - Uniform (at various levels), Donoho-Johnstone (DJ) and Dinov-Sumners (DS).

)