From Ohm

Vessel Emphasis with ITK

Emphasis of vessels (vesselness) by two-step filter (itk::HessianRecursiveGaussianImageFilter and itk::Hessian3DToVesselnessMeasureImageFilter):

Basic filter concept: After an eigenvalue decomposition of the Hessian matrix (partial derivatives matrix) bright tubular structures have one eigenvalue close to zero and two negative eigenvalues (positive values correspond to dark tubular structures). The “vesselness” algorithm first produces the Hessian matrices as first filter step, then filters and emphasizes the tubular structures as second step.

without vesselness filter with vesselness filter
// create hessian matrix filter
typedef itk::HessianRecursiveGaussianImageFilter<Image3DDoubleType> HessianImageFilterType;
HessianImageFilterType::Pointer hessianFilter = HessianImageFilterType::New();

// filter settings

// create hessian vesselness filter
typedef itk::Hessian3DToVesselnessMeasureImageFilter<float> VesselnessMeasureFilterType;
VesselnessMeasureFilterType::Pointer vesselnessFilter = VesselnessMeasureFilterType::New();

// filter settings

// connect input of vesselness filter to output of hessian filter

// create rescale filter
typedef itk::Image< float, 3> Image3DFloatType;
typedef itk::RescaleIntensityImageFilter<Image3DFloatType, Image3DType> RescaleImageFilterType;
RescaleImageFilterType::Pointer rescaleFilter = RescaleImageFilterType::New();

// filter settings

// connect input of filter to output of predecessor

... threshold at 1500 ...

... and so on ...
applying a thresholding filter applying a vesselness filter
applying a thresholding filter applying a vesselness filter

Retrieved from

Page last modified on December 19, 2014, at 07:11 PM