bool LFFieldList::IsEqual( const LFLeafStruct& target ) const
/***********************************************************/
{
WCPtrConstSListIter<LFSubField> refIter( _subFieldList );
WCPtrConstSListIter<LFSubField> targetIter( dc(target)._subFieldList );
while ( ++targetIter && ++refIter ) {
if ( ! refIter.current() -> IsEquivalent( *targetIter.current() ) ) {
return false;
}
}
return true;
}
static void OverwriteImage(itk::Image< ipMITKSegmentationTYPE, 2 >::Pointer source, itk::Image< ipMITKSegmentationTYPE, 2 >::Pointer target)
{
typedef itk::Image< ipMITKSegmentationTYPE, 2 > ItkImageType;
typedef itk::ImageRegionIterator<ItkImageType> ImageIteratorType;
ImageIteratorType sourceIter(source, source->GetLargestPossibleRegion());
ImageIteratorType targetIter(target, target->GetLargestPossibleRegion());
while ( ! sourceIter.IsAtEnd())
{
targetIter.Set(sourceIter.Get());
++sourceIter;
++targetIter;
}
}
void mitk::TumorInvasionClassification::SelectTrainingSamples(mitk::DataCollection *collection, unsigned int mode)
{
srand (time(NULL));
MITK_INFO << "LearnProgressionFeatures: Selecting training voxels.";
switch (mode) {
case 0:
{
MITK_INFO << " Selection Mode " << mode << " use all tumor voxels, healthy: 50% vicinity / 50% far away";
CollectionDilation::DilateBinaryByName(collection,m_TargetID,2,0,"EXCLUDE");
CollectionDilation::ErodeBinaryByName(collection,m_TargetID,1,0,"ERODE");
CollectionDilation::DilateBinaryByName(collection,m_TargetID,m_TargetDil2D,m_TargetDil3D,"TRAIN");
DataCollectionImageIterator<unsigned char,3> gtvIter(collection, m_TumorID);
DataCollectionImageIterator<unsigned char,3> brainMaskIter(collection, m_MaskID);
DataCollectionImageIterator<unsigned char,3> targetIter(collection, m_TargetID);
DataCollectionImageIterator<unsigned char,3> targetDil(collection, m_TargetID+"TRAIN");
// Count Healthy/ Tumor voxels
// i.o. to decide how many of each are taken
unsigned int totalTumor = 0;
unsigned int totalHealthyClose = 0;
unsigned int totalHealthy= 0;
while (!brainMaskIter.IsAtEnd())
{
if (brainMaskIter.GetVoxel() != 0)
{
if (targetIter.GetVoxel() == 1 && gtvIter.GetVoxel() == 0 )
++totalTumor;
if (targetIter.GetVoxel() == 0 && targetDil.GetVoxel()==1 && gtvIter.GetVoxel() == 0)
++totalHealthyClose;
if (targetIter.GetVoxel() == 0 && gtvIter.GetVoxel() == 0 && targetDil.GetVoxel()==0)
++totalHealthy; // healthy but not close
}
++brainMaskIter;
++targetIter;
++targetDil;
++gtvIter;
}
brainMaskIter.ToBegin();
targetIter.ToBegin();
targetDil.ToBegin();
gtvIter.ToBegin();
// Total of healthy samples that is to be collected
unsigned int targetHealthy = totalTumor * m_ClassRatio;
// Determines which portion of the healthy samples is drawn from the immediate vicinity of the newly grown tumor
ScalarType ratioClose = .5;
// Compute probabilities thresholds for choosing a close healthy voxel / any healthy voxel
ScalarType thHealthyClose = std::min(1.0 ,(targetHealthy*ratioClose)/totalHealthyClose);
ScalarType thHealthyAny = std::min(1.0 ,(targetHealthy*(1.0-ratioClose))/totalHealthy);
// Some stats
{
MITK_INFO << "Total # Tumor Voxels " << totalTumor;
MITK_INFO << "Total # Healthy Voxels" << totalHealthy;
MITK_INFO << "Target Ratio " << m_ClassRatio;
MITK_INFO << "Ratio for healthy close: " << thHealthyClose;
MITK_INFO << "Ratio for healthy any other: " << thHealthyAny;
}
// DEBUG count occurances to compare
unsigned int tumor = 0;
unsigned int healthy = 0;
while (!brainMaskIter.IsAtEnd())
{
if (brainMaskIter.GetVoxel() != 0)
{
if (targetIter.GetVoxel() == 1)
{ // choose tumor voxels for training
if (gtvIter.GetVoxel() == 0) // tumor always
{
targetIter.SetVoxel(brainMaskIter.GetVoxel()+targetIter.GetVoxel());
++tumor;
}
else
targetIter.SetVoxel(0);
}
else
{ // choose healty tissue voxels for training
ScalarType rndVal = (float)rand()/(float)(RAND_MAX) ; //(0..1)
if (gtvIter.GetVoxel() == 0 && ((targetDil.GetVoxel()==1 && rndVal <= thHealthyClose)))
{
targetIter.SetVoxel(brainMaskIter.GetVoxel()+targetIter.GetVoxel());
++healthy;
}
else if (((targetDil.GetVoxel()==0 && rndVal <= thHealthyAny)))
{
targetIter.SetVoxel(brainMaskIter.GetVoxel()+targetIter.GetVoxel());
++healthy;
}
else
targetIter.SetVoxel(0);
}
}
else
targetIter.SetVoxel(0);
++brainMaskIter;
++targetIter;
++gtvIter;
++targetDil;
}
MITK_INFO << "Training with Samples #Tumor " << tumor << " / healthy # "<< healthy;
}
break;
default:
{
MITK_INFO << " Selection Mode " << mode << " Exclude voxels in border regions, healthy: 50% vicinity / 50% far away";
//weights
ScalarType tumorWeight = 1;
//ScalarType unsureRegion = .25;
ScalarType healthyTissue = 1;
EnsureDataImageInCollection(collection, m_TumorID, "WEIGHTS");
CollectionDilation::DilateBinaryByName(collection,m_TargetID,1,0,"EXCLUDE");
CollectionDilation::ErodeBinaryByName(collection,m_TargetID,1,0,"ERODE");
CollectionDilation::DilateBinaryByName(collection,m_TargetID+"EXCLUDE",m_TargetDil2D,m_TargetDil3D,"TRAIN");
DataCollectionImageIterator<unsigned char,3> gtvIter(collection, m_TumorID);
DataCollectionImageIterator<unsigned char,3> brainMaskIter(collection, m_MaskID);
DataCollectionImageIterator<unsigned char,3> targetIter(collection, m_TargetID);
DataCollectionImageIterator<unsigned char,3> targetDil(collection, m_TargetID+"EXCLUDETRAIN");
DataCollectionImageIterator<unsigned char,3> excludeTumorIter(collection, m_TargetID+"ERODE");
DataCollectionImageIterator<unsigned char,3> excludeHealthyIter(collection, m_TargetID+"EXCLUDE");
DataCollectionImageIterator<double,3> weightsIter(collection, "WEIGHTS");
// Count Healthy/ Tumor voxels
// i.o. to decide how many of each are taken
unsigned int totalTumor = 0;
unsigned int totalHealthyClose = 0;
unsigned int totalHealthyNonClose= 0;
unsigned int totalHealthy= 0;
while (!brainMaskIter.IsAtEnd())
{
if (brainMaskIter.GetVoxel() != 0)
{
if (targetIter.GetVoxel() == 1 && gtvIter.GetVoxel() == 0 && excludeTumorIter.GetVoxel() == 1)
++totalTumor;
if (excludeHealthyIter.GetVoxel() == 0 && targetDil.GetVoxel()==1 && gtvIter.GetVoxel() == 0)
++totalHealthyClose;
if (excludeHealthyIter.GetVoxel() == 0 && gtvIter.GetVoxel() == 0 && targetDil.GetVoxel()==0)
++totalHealthyNonClose; // healthy but not close
if (excludeHealthyIter.GetVoxel() == 0 && gtvIter.GetVoxel() == 0 && targetIter.GetVoxel()==0)
++totalHealthy; // healthy
}
++brainMaskIter;
++targetIter;
++targetDil;
++gtvIter;
++excludeHealthyIter;
++excludeTumorIter;
}
brainMaskIter.ToBegin();
targetIter.ToBegin();
targetDil.ToBegin();
gtvIter.ToBegin();
excludeHealthyIter.ToBegin();
excludeTumorIter.ToBegin();
// Total of healthy samples that is to be collected
unsigned int targetHealthy = (tumorWeight/healthyTissue) * totalTumor * m_ClassRatio;
// Determines which portion of the healthy samples is drawn from the immediate vicinity of the newly grown tumor
ScalarType ratioClose = .5;
// Compute probabilities thresholds for choosing a close healthy voxel / any healthy voxel
ScalarType thHealthyClose = std::min(1.0 ,((double)targetHealthy*ratioClose)/totalHealthyClose);
ScalarType thHealthyAny = std::min(1.0 ,((double)targetHealthy*(1.0-ratioClose))/totalHealthyNonClose);
// Some stats
{
MITK_INFO << "Total Tumor " << totalTumor;
MITK_INFO << "Total healthy " << totalHealthyNonClose;
MITK_INFO << "Total healthy close " << totalHealthyClose;
MITK_INFO << "Total healthy non-close " << totalHealthyNonClose;
MITK_INFO << "Target Ratio " << m_ClassRatio;
MITK_INFO << "Target Healthy " << targetHealthy;
MITK_INFO << "Probabilty close " << thHealthyClose;
MITK_INFO << "Probabilty any other " << thHealthyAny;
}
// DEBUG count occurances to compare
unsigned int potentialClose = 0;
unsigned int selectedClose = 0;
unsigned int tumor = 0;
unsigned int healthy = 0;
while (!brainMaskIter.IsAtEnd())
{
weightsIter.SetVoxel(0);
if (brainMaskIter.GetVoxel() != 0)
{
if (targetIter.GetVoxel() == 1)
{ // choose tumor voxels for training
if (gtvIter.GetVoxel() == 0 && excludeTumorIter.GetVoxel() == 1) // tumor always
{
targetIter.SetVoxel(brainMaskIter.GetVoxel()+targetIter.GetVoxel());
weightsIter.SetVoxel(tumorWeight);
++tumor;
}
else if (gtvIter.GetVoxel() == 0 )
{
weightsIter.SetVoxel(0); //.1);
targetIter.SetVoxel(0); //2);
}
}
else
{ // choose healty tissue voxels for training
if (gtvIter.GetVoxel() == 0 && ((excludeHealthyIter.GetVoxel() == 0 && targetDil.GetVoxel()==1 )))
{
if ((float)rand()/(float)(RAND_MAX) < thHealthyClose)
{
targetIter.SetVoxel(brainMaskIter.GetVoxel()+targetIter.GetVoxel());
weightsIter.SetVoxel(1);
++healthy;
}
}
else if (((targetDil.GetVoxel()==0 && excludeHealthyIter.GetVoxel() == 0 )))
{
if ((float)rand()/(float)(RAND_MAX) < thHealthyAny)
{
targetIter.SetVoxel(brainMaskIter.GetVoxel()+targetIter.GetVoxel());
weightsIter.SetVoxel(1);
++healthy;
}
}
else if ((gtvIter.GetVoxel() == 0 && excludeHealthyIter.GetVoxel() == 1) )
{
targetIter.SetVoxel( 0);//brainMaskIter.GetVoxel()+targetIter.GetVoxel());
weightsIter.SetVoxel(0);//.1);
}
else
{
targetIter.SetVoxel(0);
weightsIter.SetVoxel(0);
}
if (gtvIter.GetVoxel() == 0 && ((excludeHealthyIter.GetVoxel() == 0 && targetDil.GetVoxel()==1 )))
{
potentialClose++;
if ((float)rand()/(float)(RAND_MAX) < thHealthyClose)
selectedClose++;
}
}
}
else
{
targetIter.SetVoxel(0);
weightsIter.SetVoxel(0);
}
if (gtvIter.GetVoxel() != 0)
targetIter.SetVoxel(0);
++brainMaskIter;
++targetIter;
++gtvIter;
++targetDil;
++excludeHealthyIter;
++excludeTumorIter;
++weightsIter;
}
MITK_INFO << "Training with Samples #Tumor " << tumor << " / healthy # "<< healthy;
MITK_INFO << "Potential Close" << potentialClose;
MITK_INFO << "Selected Close" << selectedClose;
}
break;
}
}
