diff --git a/Modules/FiberDissection/MachineLearning/mitkStreamlineFeatureExtractor.cpp b/Modules/FiberDissection/MachineLearning/mitkStreamlineFeatureExtractor.cpp
index c92d6c3..8cd76ba 100644
--- a/Modules/FiberDissection/MachineLearning/mitkStreamlineFeatureExtractor.cpp
+++ b/Modules/FiberDissection/MachineLearning/mitkStreamlineFeatureExtractor.cpp
@@ -1,920 +1,935 @@
 /*===================================================================
 
 The Medical Imaging Interaction Toolkit (MITK)
 
 Copyright (c) German Cancer Research Center.
 
 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.
 
 ===================================================================*/
 #include "mitkStreamlineFeatureExtractor.h"
 
 #define _USE_MATH_DEFINES
 #include <math.h>
 #include <boost/progress.hpp>
 #include <vnl/vnl_sparse_matrix.h>
 #include <mitkIOUtil.h>
 
 namespace mitk{
 
 StreamlineFeatureExtractor::StreamlineFeatureExtractor()
   : m_NumPoints(40)
 {
 
 }
 
 StreamlineFeatureExtractor::~StreamlineFeatureExtractor()
 {
 
 }
 
 void StreamlineFeatureExtractor::SetTractogramPrediction(const mitk::FiberBundle::Pointer &TractogramPrediction)
 {
   m_TractogramPrediction= TractogramPrediction;
 }
 
 void StreamlineFeatureExtractor::SetTractogramGroundtruth(const mitk::FiberBundle::Pointer &TractogramGroundtruth)
 {
   m_TractogramGroundtruth= TractogramGroundtruth;
 }
 
 void StreamlineFeatureExtractor::SetTractogramPlus(const mitk::FiberBundle::Pointer &TractogramPlus)
 {
   m_TractogramPlus = TractogramPlus;
 }
 
 void StreamlineFeatureExtractor::SetTractogramMinus(const mitk::FiberBundle::Pointer &TractogramMinus)
 {
   m_TractogramMinus = TractogramMinus;
 }
 
 void StreamlineFeatureExtractor::SetTractogramPrototypes(const mitk::FiberBundle::Pointer &TractogramPrototypes, bool standard)
 {
   if (standard)
   {
       MITK_INFO << "Use Standard Prototypes...";
     m_inputPrototypes = mitk::IOUtil::Load<mitk::FiberBundle>("/home/r948e/E132-Projekte/Projects/2022_Peretzke_Interactive_Fiber_Dissection/mitk_diff/prototypes_599671_40.trk");
   }
   else {
       MITK_INFO << "Use individual Prototypes...";
       m_inputPrototypes = TractogramPrototypes;
   }
 
 
 }
 
 void StreamlineFeatureExtractor::SetActiveCycle(int &activeCycle)
 {
   m_activeCycle= activeCycle;
 }
 
 
 void StreamlineFeatureExtractor::SetTractogramTest(const mitk::FiberBundle::Pointer &TractogramTest, std::string TractogramTestName)
 {
     MITK_INFO << TractogramTestName;
     m_TractogramTest= TractogramTest;
 }
 
 std::vector<vnl_matrix<float> > StreamlineFeatureExtractor::ResampleFibers(mitk::FiberBundle::Pointer tractogram)
 {
   MITK_INFO << "Infunction";
 //  mitk::FiberBundle::Pointer temp_fib = tractogram->GetDeepCopy();
 //  temp_fib->ResampleToNumPoints(m_NumPoints);
   MITK_INFO << "Resampling Done";
 
   std::vector< vnl_matrix<float> > out_fib(tractogram->GetFiberPolyData()->GetNumberOfCells());
 //#pragma omp parallel for
   for (int i=0; i<tractogram->GetFiberPolyData()->GetNumberOfCells(); i++)
   {
     vtkCell* cell = tractogram->GetFiberPolyData()->GetCell(i);
     int numPoints = cell->GetNumberOfPoints();
     vtkPoints* points = cell->GetPoints();
 
     vnl_matrix<float> streamline;
     streamline.set_size(3, m_NumPoints);
     streamline.fill(0.0);
 
     for (int j=0; j<numPoints; j++)
     {
       double cand[3];
       points->GetPoint(j, cand);
 
       vnl_vector_fixed< float, 3 > candV;
       candV[0]=cand[0]; candV[1]=cand[1]; candV[2]=cand[2];
       streamline.set_column(j, candV);
     }
 
     out_fib.at(i)=streamline;
   }
 
 
 
   return out_fib;
 }
 
 std::vector<vnl_matrix<float> > StreamlineFeatureExtractor::CalculateDmdf(std::vector<vnl_matrix<float> > tractogram,
                                                                           std::vector<vnl_matrix<float> > prototypes)
 {
+// Initialize a vector to store distance matrices for each tractogram item
+std::vector<vnl_matrix<float>> dist_vec(tractogram.size());
 
-    std::vector< vnl_matrix<float> >  dist_vec(tractogram.size());//
-    MITK_INFO << "Start Calculating Dmdf";
-
+// Loop over each tractogram item
 #pragma omp parallel for
-    for (unsigned int i=0; i<tractogram.size(); i++)
+for (std::size_t i = 0; i < tractogram.size(); ++i)
+{
+    // Initialize a distance matrix for the current tractogram item
+    vnl_matrix<float> distances(1, 2 * prototypes.size(), 0.0);
+
+    // Loop over each prototype
+    for (std::size_t j = 0; j < prototypes.size(); ++j)
     {
-        vnl_matrix<float> distances;
-        distances.set_size(1, prototypes.size());
-        distances.fill(0.0);
-        for (unsigned int j=0; j<prototypes.size(); j++)
+        // Initialize matrices to store distances between the current tractogram item and the current prototype
+        vnl_matrix<float> single_distances(1, tractogram[0].cols(), 0.0);
+        vnl_matrix<float> single_distances_flip(1, tractogram[0].cols(), 0.0);
+        vnl_matrix<float> single_end_distances(1, 2, 0.0);
+        vnl_matrix<float> single_end_distances_flip(1, 2, 0.0);
+
+        // Loop over each point in the current tractogram item
+        for (std::size_t ik = 0; ik < tractogram[0].cols(); ++ik)
         {
-            vnl_matrix<float> single_distances;
-            single_distances.set_size(1, tractogram.at(0).cols());
-            single_distances.fill(0.0);
-            vnl_matrix<float> single_distances_flip;
-            single_distances_flip.set_size(1, tractogram.at(0).cols());
-            single_distances_flip.fill(0.0);
-            for (unsigned int ik=0; ik<tractogram.at(0).cols(); ik++)
+            // Calculate the Euclidean distance between the current point in the current tractogram item and the current point in the current prototype
+            const double cur_dist = std::sqrt(std::pow(tractogram[i](0, ik) - prototypes[j](0, ik), 2.0)
+                                              + std::pow(tractogram[i](1, ik) - prototypes[j](1, ik), 2.0)
+                                              + std::pow(tractogram[i](2, ik) - prototypes[j](2, ik), 2.0));
+            single_distances(0, ik) = cur_dist;
+
+            // Calculate the Euclidean distance between the current point in the current tractogram item and the mirrored current point in the current prototype
+            const double cur_dist_flip = std::sqrt(std::pow(tractogram[i](0, ik) - prototypes[j](0, prototypes[0].cols() - (ik + 1)), 2.0)
+                                                   + std::pow(tractogram[i](1, ik) - prototypes[j](1, prototypes[0].cols() - (ik + 1)), 2.0)
+                                                   + std::pow(tractogram[i](2, ik) - prototypes[j](2, prototypes[0].cols() - (ik + 1)), 2.0));
+            single_distances_flip(0, ik) = cur_dist_flip;
+
+            // Store the distances between the first and last points of the current tractogram item and the current prototype
+            if (ik == 0)
             {
-                double cur_dist;
-                double cur_dist_flip;
-
-                cur_dist = sqrt(pow(tractogram.at(i).get(0,ik) - prototypes.at(j).get(0,ik), 2.0) +
-                                       pow(tractogram.at(i).get(1,ik) - prototypes.at(j).get(1,ik), 2.0) +
-                                       pow(tractogram.at(i).get(2,ik) - prototypes.at(j).get(2,ik), 2.0));
-
-                cur_dist_flip = sqrt(pow(tractogram.at(i).get(0,ik) - prototypes.at(j).get(0,prototypes.at(0).cols()-(ik+1)), 2.0) +
-                                       pow(tractogram.at(i).get(1,ik) - prototypes.at(j).get(1,prototypes.at(0).cols()-(ik+1)), 2.0) +
-                                       pow(tractogram.at(i).get(2,ik) - prototypes.at(j).get(2,prototypes.at(0).cols()-(ik+1)), 2.0));
-
-                single_distances.put(0,ik, cur_dist);
-                single_distances_flip.put(0,ik, cur_dist_flip);
-
+                single_end_distances(0, 0) = cur_dist;
+                single_end_distances_flip(0, 0) = cur_dist_flip;
             }
-
-            if (single_distances_flip.mean()> single_distances.mean())
+            else if (ik == tractogram[0].cols() - 1)
             {
-                distances.put(0,j, single_distances.mean());
+                single_end_distances(0, 1) = cur_dist;
+                single_end_distances_flip(0, 1) = cur_dist_flip;
             }
-            else {
-                distances.put(0,j, single_distances_flip.mean());
-            }
-          }
-        dist_vec.at(i) = distances;
-       }
+        }
+
+        // Calculate the mean distance between the current tractogram item and the current prototype
+        const double mean_single_distances = single_distances.mean();
+        const double mean_single_distances_flip = single_distances_flip.mean();
+        distances(0, j) = mean_single_distances_flip > mean_single_distances ? mean_single_distances : mean_single_distances_flip;
+
+        // // Calculate the mean END distance between the current tractogram item and the current prototype
+        const double mean_single_end_distances = single_end_distances.mean();
+        const double mean_single_end_distances_flip = single_end_distances_flip.mean();
+        const std::size_t end_distances_index = prototypes.size() + j;
+        distances(0, end_distances_index) = mean_single_end_distances_flip > mean_single_end_distances ? mean_single_distances : mean_single_end_distances_flip;
+    }
+
+        dist_vec[i] = distances;
+    }
 
-    MITK_INFO << "Done Calculation";
-    MITK_INFO << dist_vec.at(0).size();
+    MITK_INFO << "The size of the distances is " <<dist_vec[0].cols();
+    MITK_INFO << "Done calculating Dmdf";
 
     return dist_vec;
 }
 
 std::vector<vnl_matrix<float> > StreamlineFeatureExtractor::MergeTractogram(std::vector<vnl_matrix<float> > prototypes,
                                                                            std::vector<vnl_matrix<float> > positive_local_prototypes,
                                                                            std::vector<vnl_matrix<float> > negative_local_prototypes)
 {
     unsigned int pos_locals;
     unsigned int neg_locals;
 
     if (positive_local_prototypes.size() >= 50)
     {
         pos_locals= 50;
     }
     else {
         pos_locals= positive_local_prototypes.size();
     }
 
     if (negative_local_prototypes.size() >= 50)
     {
         neg_locals = 50;
     }
     else {
         neg_locals= negative_local_prototypes.size();
     }
 
 
 
     std::vector< vnl_matrix<float> > merged_prototypes;
 
     for (unsigned int k=0; k<prototypes.size(); k++)
     {
         merged_prototypes.push_back(prototypes.at(k));
     }
 
     for (unsigned int k=0; k<positive_local_prototypes.size(); k++)
     {
         merged_prototypes.push_back(positive_local_prototypes.at(k));
     }
 
     for (unsigned int k=0; k<negative_local_prototypes.size(); k++)
     {
         merged_prototypes.push_back(negative_local_prototypes.at(k));
     }
 
     MITK_INFO << "Number of prototypes:";
     MITK_INFO << prototypes.size();
     MITK_INFO << "Number of positive local prototypes:";
     MITK_INFO << pos_locals;
     MITK_INFO << "Number of negative local prototypes:";
     MITK_INFO << neg_locals;
 
     return merged_prototypes;
 
 }
 
 
 std::vector<unsigned int> StreamlineFeatureExtractor::Sort(std::vector<float> sortingVector, int lengths, int start)
 {
     std::vector<unsigned int> index;
     std::priority_queue<std::pair<float, int>> q;
 
     for (unsigned int i = 0; i < sortingVector.size(); ++i)
     {
         q.push(std::pair<float, int>(sortingVector[i], i));
     }
 
 
     for (int i = 0; i < lengths; ++i)
     {
         int ki = q.top().second;
         if (i>=start)
         {
         index.push_back(ki);
         }
         q.pop();
     }
     return index;
 }
 
 
 std::vector<std::vector<unsigned int>>  StreamlineFeatureExtractor::GetData()
 {
     MITK_INFO << "Start Function Get Data";
 
     /*Vector which saves Prediction and Fibers to label based on uncertainty*/
     std::vector<std::vector<unsigned int>> index_vec;
 
     cv::Mat data;
     cv::Mat labels_arr_vec;
 
     int size_plus = 0;
 
 
     /*Create Trainingdata: Go through positive and negative Bundle and save distances as cv::Mat and create vector with labels*/
     for ( unsigned int i=0; i<m_DistancesPlus.size(); i++)
     {
         float data_arr [m_DistancesPlus.at(0).size()];
 
         labels_arr_vec.push_back(1);
 
         for ( unsigned int j=0; j<m_DistancesPlus.at(0).cols(); j++)
         {
             data_arr[j] = m_DistancesPlus.at(i).get(0,j);
         }
         cv::Mat curdata(1, m_DistancesPlus.at(0).size(), CV_32F, data_arr);
         data.push_back(curdata);
         size_plus++;
 
     }
 
     for ( unsigned int i=m_DistancesPlus.size(); i<m_DistancesPlus.size()+m_DistancesMinus.size(); i++)
     {
         int it = i - size_plus;
         float data_arr [m_DistancesMinus.at(0).size()];
 
         labels_arr_vec.push_back(0);
 
          for ( unsigned int j=0; j<m_DistancesMinus.at(0).cols(); j++)
         {
             data_arr[j] = m_DistancesMinus.at(it).get(0,j);
         }
         cv::Mat curdata(1, m_DistancesPlus.at(0).size(), CV_32F, data_arr);
         data.push_back(curdata);
 
     }
     MITK_INFO << data.cols;
     MITK_INFO << data.rows;
 
     /*Calculate weights*/
     int zerosgt = labels_arr_vec.rows - cv::countNonZero(labels_arr_vec);
     int onesgt = cv::countNonZero(labels_arr_vec);
     float plusval = labels_arr_vec.rows / (2.0 * onesgt ) ;
     float minusval = labels_arr_vec.rows / (2.0 * zerosgt );
     float w[2] = {minusval, plusval};
     cv::Mat newweight = cv::Mat(1,2, CV_32F, w);
     MITK_INFO << "Weights";
     MITK_INFO << newweight;
 
 
     /*Shuffle Data*/
     std::vector <int> seeds;
     for (int cont = 0; cont < labels_arr_vec.rows; cont++)
     {
         seeds.push_back(cont);
     }
 
     cv::randShuffle(seeds);
     cv::Mat labels_shuffled;
     cv::Mat samples_shuffled;
 
 
 
 
     for (int cont = 0; cont < labels_arr_vec.rows; cont++)
     {
         labels_shuffled.push_back(labels_arr_vec.row(seeds[cont]));
     }
 
     for (int cont = 0; cont < labels_arr_vec.rows; cont++)
     {
         samples_shuffled.push_back(data.row(seeds[cont]));
     }
 
     /*Create Dataset and initialize Classifier*/
     cv::Ptr<cv::ml::TrainData> m_traindata = cv::ml::TrainData::create(samples_shuffled, cv::ml::ROW_SAMPLE, labels_shuffled);
 
 
 
 
     statistic_model= cv::ml::RTrees::create();
     auto criteria = cv::TermCriteria();
     criteria.type = cv::TermCriteria::MAX_ITER;
 //    criteria.epsilon = 1e-8;
     criteria.maxCount = 300;
 
     statistic_model->setMaxDepth(10); //set to three
 //    statistic_model->setMinSampleCount(m_traindata->getNTrainSamples()*0.01);
     statistic_model->setMinSampleCount(2);
     statistic_model->setTruncatePrunedTree(false);
     statistic_model->setUse1SERule(false);
     statistic_model->setUseSurrogates(false);
     statistic_model->setTermCriteria(criteria);
     statistic_model->setCVFolds(1);
     statistic_model->setPriors(newweight);
 
 
     /*Train Classifier*/
     MITK_INFO << "Start Training";
     statistic_model->train(m_traindata);
 
 
     /*Predict on Test Data*/
     MITK_INFO << "Predicting";
 
     /*Create Dataset as cv::Mat*/
     cv::Mat dataTest;
     for ( unsigned int i=0; i<m_DistancesTest.size(); i++)
     {
         float data_arr [m_DistancesTest.at(0).size()];
 
         for ( unsigned int j=0; j<m_DistancesTest.at(0).cols(); j++)
         {
             data_arr[j] = m_DistancesTest.at(i).get(0,j);
         }
         cv::Mat curdata(1, m_DistancesTest.at(0).size(), CV_32F, data_arr);
         dataTest.push_back(curdata);
     }
 
 
     std::vector<unsigned int> indexPrediction;
     std::vector<float> e(m_DistancesTest.size());
     std::vector<int> pred(m_DistancesTest.size());
 
 
     /*For every Sample/Streamline get Prediction and entropy (=based on counts of Random Forest)*/
     MITK_INFO << "Predicting on all cores";
 #pragma omp parallel for
     for (unsigned int i=0; i<m_DistancesTest.size(); i++)
     {
 
 
         int val = statistic_model->predict(dataTest.row(i));
         pred.at(i)=val;
 
         #pragma omp critical
         if (val==1)
         {
            indexPrediction.push_back(i);
         }
 
 
         cv::Mat vote;
         statistic_model->getVotes(dataTest.row(i), vote, 0);
         e.at(i) = ( -(vote.at<int>(1,0)*1.0)/ (vote.at<int>(1,0)+vote.at<int>(1,1)) * log2((vote.at<int>(1,0)*1.0)/ (vote.at<int>(1,0)+vote.at<int>(1,1))) -
                     (vote.at<int>(1,1)*1.0)/ (vote.at<int>(1,0)+vote.at<int>(1,1))* log2((vote.at<int>(1,1)*1.0)/ (vote.at<int>(1,0)+vote.at<int>(1,1))));
 
         if (isnan(e.at(i)))
         {
             e.at(i)=0;
         }
 
     }
     MITK_INFO << "Done";
 
     MITK_INFO << "--------------";
     MITK_INFO << "Prediction vector size:";
     MITK_INFO << indexPrediction.size();
     MITK_INFO << "Entropy vector size:";
     entropy_vector = e;
     MITK_INFO << e.size();
 
     MITK_INFO << "--------------";
 
     /*Get index of most unertain data (lengths defines how many data is saved)*/
 //    int lengths=500;
   int lengths = std::count_if(e.begin(), e.end(),[&](auto const& val){ return val >= 0.95; });
   if (lengths>1500)
   {
       lengths=1500;
   }
 
 
 
     int lengthsCertain = std::count_if(e.begin(), e.end(),[&](auto const& val){ return val < 0.1; });
 
     std::vector<unsigned int> indexUnc = Sort(e, lengths, 0);
 
     std::vector<unsigned int> indexCertain = Sort(e, e.size() , e.size()-lengthsCertain );
 
 //    std::vector<unsigned int> indexCertainBetween = Sort(e, e.size()-lengthsCertain , lengths);
 
     MITK_INFO << "Index Certainty Vector size";
     MITK_INFO << indexCertain.size();
 
     std::vector<unsigned int> indexCertainNeg;
     std::vector<unsigned int> indexCertainPos;
 
     for (unsigned int i=0; i<indexCertain.size(); i++)
     {
         if(pred.at(indexCertain.at(i))==0 )
         {
             indexCertainNeg.push_back(indexCertain.at(i));
         }
         else {
             indexCertainPos.push_back(indexCertain.at(i));
         }
 
     }
 
 
     //    for (unsigned int i=indexCertain.size(); i>=0; --i)
 //    std::vector<unsigned int> indexCertainBetweenNeg;
 //    std::vector<unsigned int> indexCertainBetweenPos;
 
 //    for (unsigned int i=0; i<indexCertainBetween.size(); i++)
 //    {
 //        if(pred.at(indexCertainBetween.at(i))==0 )
 //        {
 //            indexCertainBetweenNeg.push_back(indexCertainBetween.at(i));
 //        }
 //        else {
 //            indexCertainBetweenPos.push_back(indexCertainBetween.at(i));
 //        }
 
 //    }
 
     MITK_INFO << "Index Uncertainty Vector size";
     MITK_INFO << indexUnc.size();
     MITK_INFO << "Index Certainty Vector size";
     MITK_INFO << indexCertainPos.size();
     MITK_INFO << indexCertainNeg.size();
     MITK_INFO << indexCertainNeg.size() +indexCertainPos.size();
     MITK_INFO << "Index Certainty between Vector size";
 //    MITK_INFO << indexCertainBetweenPos.size();
 //    MITK_INFO << indexCertainBetweenNeg.size();
 //    MITK_INFO << indexCertainBetweenNeg.size() +indexCertainBetweenPos.size();
 
 
 //    vnl_matrix<float> distances_matrix;
 
 //    distances_matrix.set_size(lengths, lengths);
 //    distances_matrix.fill(0.0);
 
 //    std::vector<float> distances_matrix_mean;
 
 
 //    for (int i=0; i<lengths; i++)
 //    {
 //        for (int k=0; k<lengths; k++)
 //        {
 //            /*From the length.size() Samples with the highest Entropey calculate the differen between the Features*/
 //            vnl_matrix<float> diff =  m_DistancesTest.at(indexUnc.at(i)) - m_DistancesTest.at(indexUnc.at(k));
 
 //            /*Into the eucledean difference matrix, put the distance in Feature Space between every sample pair*/
 //            distances_matrix.put(i,k,diff.absolute_value_sum()/m_DistancesTest.at(0).size());
 
 //        }
 //        /*For every Sample/Streamline get the mean eucledean distance to all other Samples => one value for every Sample*/
 ////        distances_matrix_mean.push_back(distances_matrix.get_row(i).mean());
 ////        MITK_INFO << meanval.at(i);
 
 //    }
 
 
 
 //    /*Index to find values in distancematrix*/
 //    std::vector<unsigned int> myidx;
 //    /*Index to find actual streamlines using indexUnc*/
 //    std::vector<unsigned int> indexUncDist;
 //    /*Start with the Streamline of the highest entropy, which is in distance_matrix at idx 0*/
 //    myidx.push_back(0);
 //    indexUncDist.push_back(indexUnc.at(myidx.at(0)));
 
 //    /*Vecotr that stores minvalues of current iteration*/
 //    vnl_matrix<float> cur_vec;
 //    cur_vec.set_size(1,lengths);
 //    cur_vec.fill(0.0);
 //    for (int i=0; i<lengths; i++)
 //    {
 
 ////        unsigned int cur_i = indexUnc.at(myidx.at(i));
 
 //        /*Save mean distance of all used Samples*/
 //        vnl_matrix<float> sum_matrix;
 //        sum_matrix.set_size(myidx.size(), lengths);
 //        sum_matrix.fill(0);
 //        for (unsigned int ii=0; ii<myidx.size(); ii++)
 //        {
 
 //            sum_matrix.set_row(ii, distances_matrix.get_column(myidx.at(ii)));
 //        }
 
 //        for (unsigned int k=0; k<sum_matrix.columns(); k++)
 //        {
 //            cur_vec.put(0,k, sum_matrix.get_column(k).min_value());
 //        }
 //        myidx.push_back(cur_vec.arg_max());
 
 
 //        indexUncDist.push_back(indexUnc.at(myidx.at(i+1)));
 //        sum_matrix.clear();
 
 //    }
 
 //    MITK_INFO << "Dist_stop";
 
 
     /*Save Prediction*/
     index_vec.push_back(indexPrediction);
     /*Save index of uncertainty measures*/
     index_vec.push_back(indexUnc);
     /*Save index of uncertainty measures influenced by distance*/
 //    index_vec.push_back(indexUncDist);
     /*Save index of certain measures*/
     index_vec.push_back(indexCertainNeg);
     /*Save index of certain measures*/
 //    index_vec.push_back(indexCertainPos);
     /*Save index of certain measures*/
 //    index_vec.push_back(indexCertainBetweenNeg);
     /*Save index of certain measures*/
 //    index_vec.push_back(indexCertainBetweenPos);
 
     return index_vec;
 
 }
 
 std::vector<std::vector<unsigned int>>  StreamlineFeatureExtractor::GetDistanceData(float &value)
 {
     /*Vector which saves Fibers to be labeled based on fft subset uncertainty*/
     std::vector<std::vector<unsigned int>> index_vec;
 
     /*Get index of most unertain data (lengths defines how many data is saved)*/
 //    int lengths=500;
     MITK_INFO << entropy_vector.size();
   int lengths = std::count_if(entropy_vector.begin(), entropy_vector.end(),[&](auto const& val){ return val >= value; });
   if (lengths>1500)
   {
       lengths=1500;
   }
   MITK_INFO << lengths;
     /*Maybe shuffling of length so not the most uncertain values are chosen*/
     std::vector<unsigned int> indexUnc = Sort(entropy_vector, lengths, 0);
 
     vnl_matrix<float> distances_matrix;
 
     distances_matrix.set_size(lengths, lengths);
     distances_matrix.fill(0.0);
 
     std::vector<float> distances_matrix_mean;
 
 
     for (int i=0; i<lengths; i++)
     {
         for (int k=0; k<lengths; k++)
         {
             /*From the length.size() Samples with the highest Entropey calculate the differen between the Features*/
             vnl_matrix<float> diff =  m_DistancesTest.at(indexUnc.at(i)) - m_DistancesTest.at(indexUnc.at(k));
 
             /*Into the eucledean difference matrix, put the distance in Feature Space between every sample pair*/
             distances_matrix.put(i,k,diff.absolute_value_sum()/m_DistancesTest.at(0).size());
 
         }
         /*For every Sample/Streamline get the mean eucledean distance to all other Samples => one value for every Sample*/
 //        distances_matrix_mean.push_back(distances_matrix.get_row(i).mean());
 //        MITK_INFO << meanval.at(i);
 
     }
 
     MITK_INFO << "Distance Matrix Calculated";
 
     /*Index to find values in distancematrix*/
     std::vector<unsigned int> myidx;
     /*Index to find actual streamlines using indexUnc*/
     std::vector<unsigned int> indexUncDist;
     /*Start with the Streamline of the highest entropy, which is in distance_matrix at idx 0*/
     myidx.push_back(0);
     indexUncDist.push_back(indexUnc.at(myidx.at(0)));
 
     /*Vecotr that stores minvalues of current iteration*/
     vnl_matrix<float> cur_vec;
     cur_vec.set_size(1,lengths);
     cur_vec.fill(0.0);
     for (int i=0; i<lengths; i++)
     {
 
 //        unsigned int cur_i = indexUnc.at(myidx.at(i));
 
         /*Save mean distance of all used Samples*/
         vnl_matrix<float> sum_matrix;
         sum_matrix.set_size(myidx.size(), lengths);
         sum_matrix.fill(0);
         for (unsigned int ii=0; ii<myidx.size(); ii++)
         {
 
             sum_matrix.set_row(ii, distances_matrix.get_column(myidx.at(ii)));
         }
 
         for (unsigned int k=0; k<sum_matrix.columns(); k++)
         {
             cur_vec.put(0,k, sum_matrix.get_column(k).min_value());
         }
         myidx.push_back(cur_vec.arg_max());
 
 
         indexUncDist.push_back(indexUnc.at(myidx.at(i+1)));
         sum_matrix.clear();
 
     }
 
     MITK_INFO << "Dist_stop";
 
     index_vec.push_back(indexUncDist);
 
     return index_vec;
 }
 
 mitk::FiberBundle::Pointer StreamlineFeatureExtractor::CreatePrediction(std::vector<unsigned int> &index, bool removefrompool)
 {
     mitk::FiberBundle::Pointer Prediction;
     MITK_INFO << "Create Bundle";
 
     vtkSmartPointer<vtkPolyData> FibersData;
     FibersData = vtkSmartPointer<vtkPolyData>::New();
     FibersData->SetPoints(vtkSmartPointer<vtkPoints>::New());
     FibersData->SetLines(vtkSmartPointer<vtkCellArray>::New());
 
     vtkSmartPointer<vtkPolyData> vNewPolyData = vtkSmartPointer<vtkPolyData>::New();
     vtkSmartPointer<vtkCellArray> vNewLines = vtkSmartPointer<vtkCellArray>::New();
     vtkSmartPointer<vtkPoints> vNewPoints = vtkSmartPointer<vtkPoints>::New();
 
 //    vtkSmartPointer<vtkFloatArray> weights = vtkSmartPointer<vtkFloatArray>::New();
 //    weights->SetNumberOfValues(this->GetNumFibers()+fib->GetNumFibers());
 
     unsigned int indexSize = index.size();
     unsigned int counter = 0;
     MITK_INFO << "Start Loop";
     for (unsigned int i=0; i<indexSize; i++)
     {
 
         vtkCell* cell = m_TractogramTest->GetFiberPolyData()->GetCell(index[i]);
         auto numPoints = cell->GetNumberOfPoints();
         vtkPoints* points = cell->GetPoints();
 
         vtkSmartPointer<vtkPolyLine> container = vtkSmartPointer<vtkPolyLine>::New();
         for (unsigned int j=0; j<numPoints; j++)
         {
             double p[3];
             points->GetPoint(j, p);
 
             vtkIdType id = vNewPoints->InsertNextPoint(p);
             container->GetPointIds()->InsertNextId(id);
         }
 //            weights->InsertValue(counter, fib->GetFiberWeight(i));
         vNewLines->InsertNextCell(container);
         counter++;
 
     }
     if (removefrompool)
     {
         for (unsigned int i=0; i<indexSize; i++)
         {
             m_TractogramTest->GetFiberPolyData()->DeleteCell(index[i]);
         }
     }
 
 
 
 
     MITK_INFO << "Counter";
     MITK_INFO << counter;
 
 
      vNewPolyData->SetLines(vNewLines);
      vNewPolyData->SetPoints(vNewPoints);
 
      FibersData = vtkSmartPointer<vtkPolyData>::New();
      FibersData->SetPoints(vtkSmartPointer<vtkPoints>::New());
      FibersData->SetLines(vtkSmartPointer<vtkCellArray>::New());
      FibersData->SetPoints(vNewPoints);
      FibersData->SetLines(vNewLines);
 
      Prediction = mitk::FiberBundle::New(vNewPolyData);
 
 //     Bundle->SetFiberColors(255, 255, 255);
     MITK_INFO << "Cells Prediciton";
     MITK_INFO << Prediction->GetFiberPolyData()->GetNumberOfCells();
     MITK_INFO << "Cells Tractorgram";
     MITK_INFO << m_TractogramTest->GetFiberPolyData()->GetNumberOfCells();
 
     return Prediction;
 }
 
 std::vector<int> StreamlineFeatureExtractor::CreateLabels(std::vector<vnl_matrix<float> > Testdata,
                                                               std::vector<vnl_matrix<float> > Prediction)
 {
 //    vnl_vector<int> labels;
 //    vnl_vector.set_size(Testdata.size());
 //    vnl_vector.fill(0);
     std::vector<int>  labels(Testdata.size(), 0);
 
 
 #pragma omp parallel for
     for (unsigned int i=0; i<Prediction.size(); i++)
     {
         for (unsigned int k=0; k<Testdata.size(); k++)
         {
             if (Testdata.at(k)==Prediction.at(i))
             {
                 labels.at(k)=1;
             }
         }
     }
 
 
     return labels;
 
 
 
 
 }
 
 void StreamlineFeatureExtractor::GenerateData()
 {
     MITK_INFO << "Update";
 
     std::vector<vnl_matrix<float> >             T_Prototypes;
     std::vector<vnl_matrix<float> >             T_TractogramPlus;
     std::vector<vnl_matrix<float> >             T_TractogramMinus;
     std::vector<vnl_matrix<float> >             T_TractogramTest;
     std::vector<vnl_matrix<float> >             T_mergedPrototypes;
 
 
     MITK_INFO << "Resample Input Prototypes";
     T_Prototypes = ResampleFibers(m_inputPrototypes);
     MITK_INFO << "Resample Input Tractogram Minus";
     T_TractogramMinus= ResampleFibers(m_TractogramMinus);
     MITK_INFO << "Resample Input Tractogram Plus";
     T_TractogramPlus= ResampleFibers(m_TractogramPlus);
 
     /* Merge T_Prototypes, T_TractogramMinus and T_TractogramPlus for extra Features*/
     MITK_INFO << "Merging Prototypes";
     T_mergedPrototypes = MergeTractogram(T_Prototypes, T_TractogramPlus, T_TractogramMinus);
 
 
 
 
     MITK_INFO << "Calculate Features";
     MITK_INFO << "Calculate Minus Features";
     m_DistancesMinus = CalculateDmdf(T_TractogramMinus, T_mergedPrototypes);
     MITK_INFO << "Calculate Plus Features";
     m_DistancesPlus = CalculateDmdf(T_TractogramPlus, T_mergedPrototypes);
 
     MITK_INFO << "Resample Test Data";
     T_TractogramTest= ResampleFibers(m_TractogramTest);
     MITK_INFO << "Calculate Features of Test Data";
     m_DistancesTest= CalculateDmdf(T_TractogramTest, T_mergedPrototypes);
 
 
 
 
     MITK_INFO << "Done with Datacreation";
     m_index =GetData();
 
 }
 
 vnl_vector<float> StreamlineFeatureExtractor::ValidationPipe()
 {
     std::vector<vnl_matrix<float> >             T_Prototypes;
     std::vector<vnl_matrix<float> >             T_TractogramPrediction;
     std::vector<vnl_matrix<float> >             T_TractogramGroundtruth;
     std::vector<vnl_matrix<float> >             T_TractogramTest;
     std::vector<vnl_matrix<float> >             DistancesPrediction;
     std::vector<vnl_matrix<float> >             DistancesGroundtruth;
     std::vector<vnl_matrix<float> >             DistancesTest;
     std::vector<int>                            LabelsPrediction;
     std::vector<int>                            LabelsGroundtruth;
 
     MITK_INFO << "Start Resampling";
     T_Prototypes = ResampleFibers(m_inputPrototypes);
     T_TractogramPrediction= ResampleFibers(m_TractogramPrediction);
     T_TractogramGroundtruth= ResampleFibers(m_TractogramGroundtruth);
     T_TractogramTest= ResampleFibers(m_TractogramTest);
 
 
 
     MITK_INFO << "Calculate Features";
     DistancesPrediction = CalculateDmdf(T_TractogramPrediction, T_Prototypes);
     DistancesGroundtruth = CalculateDmdf(T_TractogramGroundtruth, T_Prototypes);
     DistancesTest = CalculateDmdf(T_TractogramTest, T_Prototypes);
 
     LabelsGroundtruth = CreateLabels(DistancesTest, DistancesGroundtruth);
     LabelsPrediction = CreateLabels(DistancesTest, DistancesPrediction);
 
     float FP = 0.0;
     float FN = 0.0;
     float TP = 0.0;
     float TN = 0.0;
     std::vector<int> indexfp;
 //#pragma omp parallel for
     for (unsigned int i=0; i<LabelsGroundtruth.size(); i++)
     {
         if (LabelsGroundtruth.at(i)==1 && LabelsPrediction.at(i)==1)
         {
             TP +=1;
         }
         else if (LabelsGroundtruth.at(i)==1 && LabelsPrediction.at(i)==0)
         {
             FN +=1;
         }
         else if (LabelsGroundtruth.at(i)==0 && LabelsPrediction.at(i)==1)
         {
             FP +=1;
             indexfp.push_back(i);
         }
         else if (LabelsGroundtruth.at(i)==0 && LabelsPrediction.at(i)==0)
         {
             TN +=1;
         }
     }
 
     float Precision;
     float Recall;
     float F1_score;
 
     MITK_INFO << "TP";
     MITK_INFO << TP;
     MITK_INFO << "FP";
     MITK_INFO << FP;
     MITK_INFO << "TN";
     MITK_INFO << TN;
     MITK_INFO << "FN";
     MITK_INFO << FN;
     Precision = TP/(TP + FP);
     MITK_INFO << "Precision";
     MITK_INFO << Precision;
     Recall = TP/(TP + FN);
     MITK_INFO << "Recall";
     MITK_INFO << Recall;
     F1_score = (2*Precision*Recall)/(Precision+Recall);
     MITK_INFO << "F1_score";
     MITK_INFO << F1_score;
 
 
 
     vnl_vector<float> metrics(7);
 
     metrics.put(0, TP);
     metrics.put(1, FP);
     metrics.put(2, TN);
     metrics.put(3, FN);
     metrics.put(4, Precision);
     metrics.put(5, Recall);
     metrics.put(6, F1_score);
 
 
 
 
     return metrics;
 
 }
 
 
 }