itkAdaptiveThresholdIterator.txx

Go to the documentation of this file.

/*===================================================================

The Medical Imaging Interaction Toolkit (MITK)

Copyright (c) German Cancer Research Center,
Division of Medical and Biological Informatics.
All rights reserved.

This software is distributed WITHOUT ANY WARRANTY; without
even the implied warranty of MERCHANTABILITY or FITNESS FOR
A PARTICULAR PURPOSE.

See LICENSE.txt or http://www.mitk.org for details.

===================================================================*/
#ifndef _itkAdaptiveThresholdIterator_txx
#define _itkAdaptiveThresholdIterator_txx

#include "itkAdaptiveThresholdIterator.h"
#include "mitkProgressBar.h"
#include <math.h>

namespace itk
{
  template <class TImage, class TFunction>
  AdaptiveThresholdIterator<TImage, TFunction>::AdaptiveThresholdIterator(ImageType *imagePtr,
                                                                          FunctionType *fnPtr,
                                                                          IndexType startIndex)
    : m_FineDetectionMode(false), m_DetectionStop(false)
  {
    this->m_OutputImage = imagePtr;
    m_Function = fnPtr;
    m_StartIndices.push_back(startIndex);

    // Set up the temporary image
    this->InitializeIterator();
  }

  template <class TImage, class TFunction>
  AdaptiveThresholdIterator<TImage, TFunction>::AdaptiveThresholdIterator(ImageType *imagePtr,
                                                                          FunctionType *fnPtr,
                                                                          std::vector<IndexType> &startIndex)
    : m_FineDetectionMode(false), m_DetectionStop(false)
  {
    this->m_OutputImage = imagePtr;
    m_Function = fnPtr;
    unsigned int i;
    for (i = 0; i < startIndex.size(); i++)
    {
      m_StartIndices.push_back(startIndex[i]);
    }

    // Set up the temporary image
    this->InitializeIterator();
  }

  template <class TImage, class TFunction>
  AdaptiveThresholdIterator<TImage, TFunction>::AdaptiveThresholdIterator(ImageType *imagePtr, FunctionType *fnPtr)
    : m_FineDetectionMode(false), m_DetectionStop(false)
  {
    this->m_OutputImage = imagePtr; // here we store the image, we have to wite the result to
    m_Function = fnPtr;

    // Set up the temporary image
    this->InitializeIterator();
  }

  template <class TImage, class TFunction>
  void AdaptiveThresholdIterator<TImage, TFunction>::InitializeIterator()
  {
    // Get the origin and spacing from the image in simple arrays
    m_ImageOrigin = this->m_OutputImage->GetOrigin();
    m_ImageSpacing = this->m_OutputImage->GetSpacing();
    m_ImageRegion = this->m_OutputImage->GetBufferedRegion();

    this->InitRegionGrowingState();

    m_VoxelCounter = 0;
    m_LastVoxelNumber = 0;
    m_CurrentLeakageRatio = 0;
    m_DetectedLeakagePoint = 0;
  }

  template <class TImage, class TFunction>
  void AdaptiveThresholdIterator<TImage, TFunction>::SetExpansionDirection(bool upwards)
  {
    m_UpwardsExpansion = upwards;
  }

  template <class TImage, class TFunction>
  void AdaptiveThresholdIterator<TImage, TFunction>::IncrementRegionGrowingState()
  {
    // make the progressbar go one step further
    if (!m_FineDetectionMode)
      mitk::ProgressBar::GetInstance()->Progress();

    // updating the thresholds
    if (m_UpwardsExpansion)
    {
      this->ExpandThresholdUpwards();
    }
    else
    {
      this->ExpandThresholdDownwards();
    }

    // leakage-detection
    int criticalValue = 2000; // calculate a bit more "intelligent"

    if (!m_FineDetectionMode)
    {
      int diff = m_VoxelCounter - m_LastVoxelNumber;
      if (diff > m_CurrentLeakageRatio)
      {
        m_CurrentLeakageRatio = diff;
        m_DetectedLeakagePoint = m_RegionGrowingState;
      }
      m_LastVoxelNumber = m_VoxelCounter;
      m_VoxelCounter = 0;
    }
    else // fine leakage detection
    {
      // counting voxels over interations; if above a critical value (to be extended) then set this to leakage
      int diff = m_VoxelCounter - m_LastVoxelNumber;
      // std::cout<<"diff is: "<<diff<<"\n";
      if (diff <= criticalValue && (!m_DetectionStop))
      {
        // m_CurrentLeakageRatio = diff;
        m_DetectedLeakagePoint = m_RegionGrowingState + 1; // TODO check why this is needed
        // std::cout<<"setting CurrentLeakageRatio to: "<<diff<<" and leakagePoint to: "<<m_DetectedLeakagePoint<<"\n";
      }
      else
      {
        m_DetectionStop = true;
        // std::cout<<"\n\n[ITERATOR] detection stop!!!\n";
      }
      m_LastVoxelNumber = m_VoxelCounter;
      m_VoxelCounter = 0;
    }
    // increment the counter
    m_RegionGrowingState++;
  }

  template <class TImage, class TFunction>
  void AdaptiveThresholdIterator<TImage, TFunction>::ExpandThresholdUpwards()
  {
    int upper = (int)m_Function->GetUpper();
    upper++;
    m_Function->ThresholdBetween(m_MinTH, upper);
  }

  template <class TImage, class TFunction>
  void AdaptiveThresholdIterator<TImage, TFunction>::ExpandThresholdDownwards()
  {
    int lower = (int)m_Function->GetLower();
    lower--;
    m_Function->ThresholdBetween(lower, m_MaxTH);
  }

  template <class TImage, class TFunction>
  void AdaptiveThresholdIterator<TImage, TFunction>::InitRegionGrowingState()
  {
    this->m_RegionGrowingState = 1;
  }

  template <class TImage, class TFunction>
  int AdaptiveThresholdIterator<TImage, TFunction>::EstimateDistance(IndexType tempIndex)
  {
    PixelType value = this->m_Function->GetInputImage()->GetPixel(tempIndex);
    PixelType minPixel = PixelType(m_MinTH);
    PixelType maxPixel = PixelType(m_MaxTH);

    if (value > maxPixel || value < minPixel)
    {
      return 0;
    }
    if (m_UpwardsExpansion)
    {
      return (int)(value - m_SeedPointValue);
    }
    else
    {
      return (int)(m_SeedPointValue - value);
    }
  }

  template <class TImage, class TFunction>
  void AdaptiveThresholdIterator<TImage, TFunction>::SetMinTH(int min)
  {
    m_MinTH = min;
  }

  template <class TImage, class TFunction>
  void AdaptiveThresholdIterator<TImage, TFunction>::SetMaxTH(int max)
  {
    m_MaxTH = max;
  }

  template <class TImage, class TFunction>
  int AdaptiveThresholdIterator<TImage, TFunction>::GetSeedPointValue()
  {
    return this->m_SeedPointValue;
  }

  template <class TImage, class TFunction>
  void AdaptiveThresholdIterator<TImage, TFunction>::GoToBegin()
  {
    m_QueueMap.clear();

    m_SeedPointValue = 0;

    IndexType seedIndex = m_StartIndices[0];

    bool doAverage = false; // enable or disable manually!
    if (doAverage)
    {
      // loops for creating the sum of the N27-neighborhood around the seedpoint
      for (int i = -1; i <= 1; i++)
      {
        for (int j = -1; j <= 1; j++)
        {
          for (int k = -1; k <= 1; k++)
          {
            seedIndex[0] = seedIndex[0] + i;
            seedIndex[1] = seedIndex[1] + j;
            seedIndex[2] = seedIndex[2] + k;

            m_SeedPointValue += (int)m_Function->GetInputImage()->GetPixel(seedIndex);
          }
        }
      }

      // value of seedpoint computed from mean of N27-neighborhood
      m_SeedPointValue = m_SeedPointValue / 27;
    }
    else
    {
      m_SeedPointValue = (int)m_Function->GetInputImage()->GetPixel(seedIndex);
    }

    this->CheckSeedPointValue();

    m_InitializeValue = (this->CalculateMaxRGS() + 1);
    if (!m_FineDetectionMode)
      mitk::ProgressBar::GetInstance()->AddStepsToDo(m_InitializeValue - 1);

    // only initialize with zeros for the first segmention (raw segmentation mode)
    if (!m_FineDetectionMode)
    {
      this->m_OutputImage->FillBuffer((PixelType)0);
    }
    if (m_UpwardsExpansion)
    {
      m_Function->ThresholdBetween(m_MinTH, m_SeedPointValue);
    }
    else
    {
      m_Function->ThresholdBetween(m_SeedPointValue, m_MaxTH);
    }

    this->m_IsAtEnd = true;

    seedIndex = m_StartIndices[0]; // warum noch mal? Steht doch schon in Zeile 224

    if (this->m_OutputImage->GetBufferedRegion().IsInside(seedIndex) && this->m_SeedPointValue > this->m_MinTH &&
        this->m_SeedPointValue < this->m_MaxTH)
    {
      // Push the seed onto the queue
      this->InsertIndexTypeIntoQueueMap(m_RegionGrowingState, seedIndex);

      // Obviously, we're at the beginning
      this->m_IsAtEnd = false;

      this->m_OutputImage->SetPixel(seedIndex, (m_InitializeValue - m_RegionGrowingState));
    }
  }

  template <class TImage, class TFunction>
  void AdaptiveThresholdIterator<TImage, TFunction>::CheckSeedPointValue()
  {
    // checks, if the value, that has been averaged over the N-27 neighborhood aorund the seedpoint is still inside
    // the thresholds-ranges. if not, the actual value of the seedpoint (not averaged) is used
    if (m_SeedPointValue < m_MinTH || m_SeedPointValue > m_MaxTH)
    {
      m_SeedPointValue = (int)m_Function->GetInputImage()->GetPixel(m_StartIndices[0]);
    }
  }

  template <class TImage, class TFunction>
  unsigned int AdaptiveThresholdIterator<TImage, TFunction>::CalculateMaxRGS()
  {
    if (m_UpwardsExpansion)
    {
      return (m_MaxTH - m_SeedPointValue);
    }
    else
    {
      return (m_SeedPointValue - m_MinTH);
    }
  }

  template <class TImage, class TFunction>
  bool AdaptiveThresholdIterator<TImage, TFunction>::IsPixelIncluded(const IndexType &index) const
  {
    // checks, if grayvalue of current voxel is inside the currently used thresholds
    return this->m_Function->EvaluateAtIndex(index);
  }

  template <class TImage, class TFunction>
  void AdaptiveThresholdIterator<TImage, TFunction>::InsertIndexTypeIntoQueueMap(unsigned int key, IndexType index)
  {
    // first check if the key-specific queue already exists
    if (m_QueueMap.count(key) == 0)
    {
      // if queue doesn´t exist, create it, push the IndexType onto it
      // and insert it into the map

      IndexQueueType newQueue;
      newQueue.push(index);

      typedef typename QueueMapType::value_type KeyIndexQueue;

      m_QueueMap.insert(KeyIndexQueue(key, newQueue));
    }
    else
    {
      // if queue already exists in the map, push IndexType onto its specific queue
      (*(m_QueueMap.find(key))).second.push(index);
    }
  }

  template <class TImage, class TFunction>
  void AdaptiveThresholdIterator<TImage, TFunction>::DoExtendedFloodStep()
  {
    // The index in the front of the queue should always be
    // valid and be inside since this is what the iterator
    // uses in the Set/Get methods. This is ensured by the
    // GoToBegin() method.

    typename QueueMapType::iterator currentIt = m_QueueMap.find(m_RegionGrowingState);

    if (currentIt == m_QueueMap.end())
    {
      this->IncrementRegionGrowingState();
    }
    else
    {
      IndexQueueType *currentQueue = &(*currentIt).second;

      // Take the index in the front of the queue
      const IndexType &topIndex = currentQueue->front();

      // Iterate through all possible dimensions
      // NOTE: Replace this with a ShapeNeighborhoodIterator
      for (unsigned int i = 0; i < NDimensions; i++)
      {
        // The j loop establishes either left or right neighbor (+-1)
        for (int j = -1; j <= 1; j += 2)
        {
          IndexType tempIndex;

          // build the index of a neighbor
          for (unsigned int k = 0; k < NDimensions; k++)
          {
            if (i != k)
            {
              tempIndex.m_Index[k] = topIndex[k];
            }
            else
            {
              tempIndex.m_Index[k] = topIndex[k] + j;
            }
          } // end build the index of a neighbor

          // If this is a valid index and have not been tested,
          // then test it.
          if (m_ImageRegion.IsInside(tempIndex))
          {
            // check if voxel hasn´t already been processed
            if (this->m_OutputImage->GetPixel(tempIndex) == 0)
            {
              // if it is inside, push it into the queue
              if (this->IsPixelIncluded(tempIndex))
              {
                // hier wird Voxel in momentan aktiven Stack und ins OutputImage geschrieben
                this->InsertIndexTypeIntoQueueMap((m_RegionGrowingState), tempIndex);
                this->m_OutputImage->SetPixel(tempIndex, (m_InitializeValue - m_RegionGrowingState));
              }
              else // If the pixel is not inside the current threshold
              {
                int distance = this->EstimateDistance(
                  tempIndex); // [!] sollte nicht estimateDistance sondern calculateDistance() heißen!
                if (distance != 0)
                {
                  // hier wird Voxel in entsprechenden Stack und ins OutputImage geschrieben
                  this->InsertIndexTypeIntoQueueMap((distance), tempIndex);
                  this->m_OutputImage->SetPixel(tempIndex, (m_InitializeValue - distance));
                }
              }
            }
          }
        } // end left/right neighbor loop
      }   // end check all neighbors

      // Now that all the potential neighbors have been
      // inserted we can get rid of the pixel in the front
      currentQueue->pop();
      m_VoxelCounter++;

      if (currentQueue->empty())
      {
        // if currently used queue is empty
        this->IncrementRegionGrowingState();
      }
    }
    if (m_RegionGrowingState >= (m_InitializeValue) || m_DetectionStop)
    {
      this->m_IsAtEnd = true;
      // std::cout << "RegionGrowing finished !" << std::endl;
      // std::cout << "Detected point of leakage: " << m_DetectedLeakagePoint << std::endl;
    }
  }

} // end namespace itk

#endif