Bootstrap Agency Page Template

MSparkles features a freely configurable pre-processing pipeline. Via the pre-processing configuration dialog optimized pre-processing pipelines can be created for each dataset. In addition to that, it is possible to bulk-edit the pre-processing of multiple datasets at once. This is especially helpful, for datasets with similar image characteristics. To do so, simply select the datasets you want to edit from the datamanager and open the pre-processing configuration dialog. To configure a pre-processing pipeline, select one or more algorithms from the dropdown menu and add them to the pipeline. The order of execution is the order in which the algorithms appear in the list of the configuration dialog.
Some datasets require more or other filter operations than others. Also, to our best knowledge, there is no gold standard for pre-processing. However, a best practice approach we are also using in our lab is to first run denoising prior to any other algorithm. Then, if considerable drift is present, we allpy stack registration. Next, we apply a temporal median filter to eliminate small artefacts and perform minimal smooting. Other filters are then applied on demand. Spatial filters automatically adjust to the dimensionality of the dataset (2D or 3D).
By selecting any filter from your configured pipeline, the pre-processing dialog will showa preview, allowing to better estimate the effects of a particular filter.

Available filters

In MSparkles, the filter size is specified via a filter'S kernel half-size parameter. The actual filter size is then computed as 2*HalfSize+1. This ensures, an odd size of the filter which is computationally advantageous, since is ensures the existance of a central element within the filter kernel and avoids additional interpolations. Especially for median filters this is advantageous, since it does not introduce new values and helps preserve sharp edges of structures within an image or signal.

For more details on image filtering, please refer to the MATLAB help.

Filter name	Parameters
Spatial Gauss filter Temporal Gauss filter	Sigma Kernel half-size A Gaussian smoothing filters uses a Gauss-curve as weighting function for the pixels covered by the filter kernel. Typically the kernel half-size is 3*sigma
Spatial median filter Temporal median filter	Kernel halt-size The median filter sorts the values within the filter window in ascending order and chooses the central element (median) as local answer of the filter.
Spatial boxcar filter Temporal boxcar filter	Kernel half-size The boxcar (or mean filter) computed the mean grey value of the pixels within the filter window as to local answer of the filter.
Stack registration	Ref. frame Ref. channel Automated, intensity-based stack registration. Imagees are aligned with respect to a reference frame and a reference channel. After registration the stack is automatically cropped and meta data are updated (if neccessary).
SURE-LET denoising	Automated, per frame denoising, using the SURE-LET algorithm.
Linear unmixing	Linear unmixing of multi-channel images, based on a least-squares algorithm.
Temporal Savitzky-Golay filter	Kernel half-size Order Polynomial smoothing of 1D, time-dependent signals. Order specifies the order of the fitted polynom within the filter window. Order must be less then the length of the filter window (KernelHalfSize*2+1). See MATLAB help for more details.