matitk(operationName,[parameters],[inputArray1],[inputArray2],[seed(s)Array],[Image(s)Spacing])

Legend:

• The first argument to matitk, operationName, specifies the opcode of the implemented ITK method to be invoked.
• The second argument to matitk, parameters, specifies the required parameters of the ITK method to be invoked (specified by operationName).  To find out what parameters are required for a particular method, type matitk(operationName);  
• The third and fourth arguments to matitk, inputArray1 and inputArray2, specify the input image volume.  They must be three dimensional and contain double, float, unsigned char or signed integer data type elements.  In the case where a second image volume is not required for the method being invoked, provide [] as the fourth argument. 
• The fifth argument seedsArray arguments specify the seed points (in MATLAB coordinate system) in the following order: [x₁, y₁, z₁, x₂, y₂, z₂, …, x_n, y_n, z_n].  Because it is three dimensional, the number of elements in seedsArray should be a multiple of three.  In the case where seeding is not required for the method being invoked, provide [] as the fifth argument.
• The last optional argument specifies the spacing of the supplied image volume.  The performance of certain ITK methods may be affected by the spacing.  If this argument is omitted, an isotropic spacing of [1,1,1] is assumed.

Example

 

To demonstrate the functionality of MATITK, we first load the sample built-in 3D brain mri image from MATLAB.  The loaded image will automatically be stored inside the variable D.

 

 

We can use the following commands to visualize an axial brain slice (Figure 1):

Figure 1 : (Left to right) Visualizing an axial, sagittal, and coronal brain slice of the sample image D.

 

The data type of the loaded image is unsigned char.  As such, we would like to use the double data type version of MATITK, and we first convert the input using double(D).  matitk(‘f’) is invoked to show the list of implemented filtering methods and the corresponding opcode. 

 

FCA is the opcode for CurvatureAnisotropicDiffusionImageFilter

The data type of the loaded image is unsigned char.  As such, we would like to use the double data type version of MATITK, and we first convert the input using double(D).  matitk(‘f’) is invoked to show the list of implemented filtering methods and the corresponding opcode. 

 

FCA is the opcode for CurvatureAnisotropicDiffusionImageFilter.  matitk(‘fca’) can be used to list out the arguments required for using

CurvatureAnisotropicDiffusionImageFilter (i.e. numberOfIterations, timeStep and conductance).  For the example, we chose our arguments to be 5, 0.0625 and 3 respectively in this order, and supply the arguments as an array:

 

>> b=matitk('FCA',[5 0.0625 3], double(D)); Image input of type double detected, executing MATITK in double mode   FCA is being executed... FCA has completed.

 

We can use the following commands to visualize the filtering result (Figure 2):

 

 

Figure 2 : Visualizing sample image D after “FCA” operation. 
The operation is performed in double data-type mode.

We apply ConfidenceConnectedImageFilter to the resulting filtered image (Figure 3).  The following example illustrates how the seed point (102, 82, 25) is supplied as an argument:

 

 

Figure 3 : Visualizing sample image D after “FCA” and “SCC” operation.
(Left to right) Axial slice, sagittal, and coronal brain slices.
The operations are performed in double data-type mode.

 

Instead of invoking the “double” version of ConfidenceConnectedImageFilter, we could first cast b into unsigned char first.  The unsigned char version of

ConfidenceConnectedImageFilter will be invoked as a result (Figure 4):

 

 

Figure 4 : Visualizing sample image D after “FCA” and “SCC” operation.
The operations are performed in unsigned char data-type mode.

 

Notice how casting can affect the final result.

 

For another illustration, we apply GradientMagnitudeImageFilter to the original image D of type unsigned char (Figure 5).  Notice the unsigned char version of ITK method will be used:

 

>> G=matitk('FGM',[],D); Image input of type unsigned char detected, executing MATITK in unsigned char mode   FGM is being executed... FGM has completed.

 

 

Figure 5 : Visualizing sample image D after “FGM” operation.
(Left to right) Axial, sagittal, and coronal brain slices.
The operations are performed in double data-type mode.

 

MATITK also supports receiving multiple outputs from ITK methods.  The resulting images of invoking OtsuMultipleThresholdImageFilter will be stored in variables O1, O2, and O3 (Figure 6):

 

 

Figure 6 : Visualizing the three outputs of “FOMT” operation.

Figure 7 : A slice of the cube before (left) and after (right) applying Thin Plate Spline Warping (“RTPS”) operation.

 

To illustrate how warping can be applied using MATITK, consider warping a cube according to the following landmarks:

 

 

The source landmarks correspond to the edges of the cube, and the target landmarks correspond to randomly generated points close to the original edges.

 

 

The result of execution is shown in (Figure 7).

 

Available Opcodes

As of version 2.4.04 released on Aug 24 2006, the following table lists the operations are available.  

 

Opcode	Method
FAAB	AntiAliasBinaryImageFilter
FBB	BinomialBlurImageFilter
FBD	BinaryDilateFilter
FBE	BinaryErodeFilter
FBL	BilateralFilter
FBT	BinaryThresholdImageFilter
FCA	CurvatureAnsioFilter
FCF	CurvatureFlowFilter
FD	DerivativeImageFilter
FDG	DiscreteGaussianImageFilter
FDM	DanielssonDistanceMapImageFilter
FDMV	DanielssonDistanceMapImageFilterGetVoronoiMap
FF	FlipImageFilter
FFFT	FFTImageFilter
FGA	GaussianFilter
FGAD	GradientAnisotropicDiffusionImageFilter
FGM	GradientMagnitudeFilter
FGMRG	GradientMagnitudeRecursiveGaussianImageFilter
FGMS	GradientMagnitudeWithSmoothingFilter
FLS	LaplacianRecursiveGaussianImageFilter
FMEAN	MeanImageFilter
FMEDIAN	MedianImageFilter
FMMCF	MinMaxCurvatureFlowFilter
FOMT	OtsuMultipleThresholdImageFilter
FSN	SigmoidNonlinearMappingFilter
FVBIH	VotingBinaryIterativeHoleFillingImageFilter
FVMI	VesselnessMeasureImageFilter
FRG	RecursiveGaussianImageFilter

 

The following segmentation functions are implemented:

Opcode	Method
SCC	ConfidenceConnectedSegmentation
SCSS	CellularSegmentationSegmentation(Debug)
SCT	ConnectedThresholdSegmentation
SFM	FastMarchSegmentation
SGAC	GeodesicActiveContourLevelSetSegmentation
SIC	IsolatedConnectedSegmentation
SLLS	LaplacianLevelSetLevelSetSegmentation
SNC	NeighbourhoodConnectedSegmentation
SOT	OtsuThresholdSegmentation
SSDLS	ShapeDetectionLevelSetFilter
SWS	WatershedSegmentation

 

The following registration functions are implemented:

Opcode	Method
RD	registerDemon
RTPS	registerThinPlateSpline

 

 

For documentation on each method, simply types its corresponding opcode in MATITK prompt.