diff --git a/Modules/Classification/CLUtilities/include/mitkGIFFirstOrderNumericStatistics.h b/Modules/Classification/CLUtilities/include/mitkGIFFirstOrderNumericStatistics.h index 0a85b0e04b..9f28cde2d6 100644 --- a/Modules/Classification/CLUtilities/include/mitkGIFFirstOrderNumericStatistics.h +++ b/Modules/Classification/CLUtilities/include/mitkGIFFirstOrderNumericStatistics.h @@ -1,164 +1,156 @@ /*=================================================================== 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 mitkGIFFirstOrderStatistics_h #define mitkGIFFirstOrderStatistics_h #include #include #include namespace mitk { class MITKCLUTILITIES_EXPORT GIFFirstOrderNumericStatistics : public AbstractGlobalImageFeature { public: /** * \brief Calculates first order statistics of the given image. * * The first order statistics for the intensity distribution within a given Region of Interest (ROI) * is caluclated. The ROI is defined using a mask. * * The features are calculated on a quantified image. If the bin-size is too big, the obtained values * can be errornous and missleading. It is therefore important to use enough bins. The binned approach is * used in order to avoid floating-point related errors. * * This feature calculator is activated by the option -first-order or -fo. * * The connected areas are based on the binned image, the binning parameters can be set via the default * parameters as described in AbstractGlobalImageFeature. It is also possible to determine the * dimensionality of the neighbourhood using direction-related commands as described in AbstractGlobalImageFeature. * No other options are possible beside these two options. * * The features are calculated based on a mask. It is assumed that the mask is * of the type of an unsigned short image. All voxels with the value 1 are treated as masked. * * The following features are then defined using the (binned) voxel intensity \f$ x_i \f$ of each voxel, the probability * an intensity \f$ p_x \f$, and the overall number of voxels within the mask \f$ N_v \f$: * - First Order::Mean: The mean intensity within the ROI * \f[ \textup{Mean}= \mu = \frac{1}{N_v} \sum x_i \f] * - First Order::Unbiased Variance: An unbiased estimation of the variance: * \f[ \textup{Unbiased Variance} = \frac{1}{N_v - 1} \sum \left( x_i - \mu \right)^2 \f] * - First Order::Biased Variance: An biased estimation of the variance. If not specified otherwise, this is * used as the variance: * \f[ \textup{Biased Variance} = \sigma^2 = \frac{1}{N_v} \sum \left( x_i - \mu \right)^2 \f] * - First Order::Unbiased Standard Deviation: Estimation of diversity within the intensity values * \f[ \textup{Unbiased Standard Deviation} = \sqrt{\frac{1}{N_v-1} \sum \left( x_i - \mu \right)^2} \f] * - First Order::Biased Standard Deviation: Estimation of diversity within the intensity values * \f[ \textup{Biased Standard Deviation} = \sigma = \sqrt{\frac{1}{N_v} \sum \left( x_i - \mu \right)^2} \f] * - First Order::Skewness: * \f[ \textup{Skewness} = \frac{\frac{1}{N_v} \sum \left( x_i - \mu \right)^3}{\sigma^3} \f] * - First Order::Kurtosis: The kurtosis is a measurement of the peakness of the given * distirbution: * \f[ \textup{Kurtosis} = \frac{\frac{1}{N_v} \sum \left( x_i - \mu \right)^4}{\sigma^4} \f] * - First Order::Excess Kurtosis: The kurtosis is a measurement of the peakness of the given * distirbution. The excess kurtosis is similar to the kurtosis, but is corrected by a fisher correction, * ensuring that a gaussian distribution has an excess kurtosis of 0. * \f[ \textup{Excess Kurtosis} = \frac{\frac{1}{N_v} \sum \left( x_i - \mu \right)^4}{\sigma^4} - 3 \f] * - First Order::Median: The median is defined as the median of the all intensities in the ROI. * - First Order::Minimum: The minimum is defined as the minimum of the all intensities in the ROI. * - First Order::05th Percentile: \f$ P_{5\%} \f$ The 5% percentile. 5% of all voxel do have this or a lower intensity. * - First Order::10th Percentile: \f$ P_{10\%} \f$ The 10% percentile. 10% of all voxel do have this or a lower intensity. * - First Order::15th Percentile: \f$ P_{15\%} \f$ The 15% percentile. 15% of all voxel do have this or a lower intensity. * - First Order::20th Percentile: \f$ P_{20\%} \f$ The 20% percentile. 20% of all voxel do have this or a lower intensity. * - First Order::25th Percentile: \f$ P_{25\%} \f$ The 25% percentile. 25% of all voxel do have this or a lower intensity. * - First Order::30th Percentile: \f$ P_{30\%} \f$ The 30% percentile. 30% of all voxel do have this or a lower intensity. * - First Order::35th Percentile: \f$ P_{35\%} \f$ The 35% percentile. 35% of all voxel do have this or a lower intensity. * - First Order::40th Percentile: \f$ P_{40\%} \f$ The 40% percentile. 40% of all voxel do have this or a lower intensity. * - First Order::45th Percentile: \f$ P_{45\%} \f$ The 45% percentile. 45% of all voxel do have this or a lower intensity. * - First Order::50th Percentile: \f$ P_{50\%} \f$ The 50% percentile. 50% of all voxel do have this or a lower intensity. * - First Order::55th Percentile: \f$ P_{55\%} \f$ The 55% percentile. 55% of all voxel do have this or a lower intensity. * - First Order::60th Percentile: \f$ P_{60\%} \f$ The 60% percentile. 60% of all voxel do have this or a lower intensity. * - First Order::65th Percentile: \f$ P_{65\%} \f$ The 65% percentile. 65% of all voxel do have this or a lower intensity. * - First Order::70th Percentile: \f$ P_{70\%} \f$ The 70% percentile. 70% of all voxel do have this or a lower intensity. * - First Order::75th Percentile: \f$ P_{75\%} \f$ The 75% percentile. 75% of all voxel do have this or a lower intensity. * - First Order::80th Percentile: \f$ P_{80\%} \f$ The 80% percentile. 80% of all voxel do have this or a lower intensity. * - First Order::85th Percentile: \f$ P_{85\%} \f$ The 85% percentile. 85% of all voxel do have this or a lower intensity. * - First Order::90th Percentile: \f$ P_{90\%} \f$ The 90% percentile. 90% of all voxel do have this or a lower intensity. * - First Order::95th Percentile: \f$ P_{95\%} \f$ The 95% percentile. 95% of all voxel do have this or a lower intensity. * - First Order::Maximum: The maximum is defined as the minimum of the all intensities in the ROI. * - First Order::Range: The range of intensity values is defined as the difference between the maximum * and minimum intensity in the ROI. * - First Order::Interquartile Range: The difference between the 75% and 25% quantile. * - First Order::Mean Absolute Deviation: The mean absolute deviation gives the mean distance of each * voxel intensity to the overal mean intensity and is a measure of the dispersion of the intensity form the * mean value: * \f[ \textup{Mean Absolute Deviation} = \frac{1}{N_v} \sum \left \| x_i - \mu \right \| \f] * - First Order::Robust Mean: The mean intensity within the ROI for all voxels between the 10% and 90% quantile: * \f[ \textup{Robust Mean}= \mu_R = \frac{1}{N_{vr}} \sum x_i \f] * - First Order::Robust Mean Absolute Deviation: The absolute deviation of all intensities within the ROI for * all voxels between the 10% and 90% quantilefrom the robust mean intensity: * \f[ \textup{Robust Mean Absolute Deviation}= \mu_R = \frac{1}{N_{vr}} \sum \left \| x_i - \mu_R \right \| \f] * - First Order::Median Absolute Deviation: Similar to the mean absolute deviation, but uses the median * instead of the mean to measure the center of the distribution. * - First Order::Coefficient Of Variation: Measures the dispersion of the intensity distribution: * \f[ \textup{Coefficient Of Variation} = \frac{sigma}{\mu} \f] * - First Order::Quantile Coefficient Of Dispersion: A robust alternative to teh coefficient of variance: * \f[ \textup{Quantile Coefficient Of Dispersion} = \frac{P_{75\%} - P_{25\%} }{P_{75\%} + P_{25\%}} \f] * - First Order::Energy: The intensity energy: * \f[ \textup{Energy} = \sum x_i ^2 \f] * - First Order::Root Mean Square: Root mean square is an important measure for the error. * \f[ \textup{Root Mean Square} = \sqrt{\frac{\sum x_i ^2}{N_v}} \f] * - First Order::Uniformity: * \f[ \textup{Uniformity} = \sum p_x^2 \f] * - First Order::Entropy: * \f[ \textup{Entropy} = - \sum p_x \textup{log}_2(p_x) \f] * - First Order::Entropy: * \f[ \textup{Entropy} = - \sum p_x \textup{log}_2(p_x) \f] * - First Order::Covered Image Intensity Range: Percentage of the image intensity range (maximum - minimum in whole * image) that is covered by the ROI. * - First Order::Sum: The sum of all intensities. It is correlated to the mean intensity. * \f[ \textup{Sum} = \sum x_i \f] * - First Order::Mode: The most common intensity. * - First Order::Mode Probability: The likelihood of the most common intensity. * - First Order::Number Of Voxels: \f$ N_v \f$ the number of voxels covered by the ROI. * - First Order::Image Dimension: The dimensionality of the image (e.g. 2D, 3D, etc.). * - First Order::Number Of Voxels: The product of all spacing along all dimensions. In 3D, this is equal to the * volume. * - First Order::Number Of Voxels: The volume of a single voxel. If the dimensionality is only 2D, this is the * surface of an voxel. */ mitkClassMacro(GIFFirstOrderNumericStatistics,AbstractGlobalImageFeature) itkFactorylessNewMacro(Self) itkCloneMacro(Self) GIFFirstOrderNumericStatistics(); /** * \brief Calculates the First Order Features based on a binned version of the image. */ FeatureListType CalculateFeatures(const Image::Pointer & image, const Image::Pointer &feature) override; /** * \brief Returns a list of the names of all features that are calculated from this class */ FeatureNameListType GetFeatureNames() override; virtual std::string GetCurrentFeatureEncoding() override; virtual void CalculateFeaturesUsingParameters(const Image::Pointer & feature, const Image::Pointer &mask, const Image::Pointer &maskNoNAN, FeatureListType &featureList); virtual void AddArguments(mitkCommandLineParser &parser); - - struct ParameterStruct { - double MinimumIntensity; - double MaximumIntensity; - int Bins; - std::string prefix; - }; - }; } #endif //mitkGIFFirstOrderStatistics_h diff --git a/Modules/Classification/CLUtilities/src/GlobalImageFeatures/mitkGIFFirstOrderNumericStatistics.cpp b/Modules/Classification/CLUtilities/src/GlobalImageFeatures/mitkGIFFirstOrderNumericStatistics.cpp index a173f9a588..56dd133b10 100644 --- a/Modules/Classification/CLUtilities/src/GlobalImageFeatures/mitkGIFFirstOrderNumericStatistics.cpp +++ b/Modules/Classification/CLUtilities/src/GlobalImageFeatures/mitkGIFFirstOrderNumericStatistics.cpp @@ -1,327 +1,390 @@ /*=================================================================== 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. ===================================================================*/ #include // MITK #include #include #include // ITK #include #include +#include // STL #include +struct FirstOrderNumericParameterStruct { + mitk::IntensityQuantifier::Pointer quantifier; + double MinimumIntensity; + double MaximumIntensity; + int Bins; + std::string prefix; +}; + template void -CalculateFirstOrderStatistics(itk::Image* itkImage, mitk::Image::Pointer mask, mitk::GIFFirstOrderNumericStatistics::FeatureListType & featureList, mitk::GIFFirstOrderNumericStatistics::ParameterStruct params) +CalculateFirstOrderStatistics(itk::Image* itkImage, mitk::Image::Pointer mask, mitk::GIFFirstOrderNumericStatistics::FeatureListType & featureList, FirstOrderNumericParameterStruct params) { - typedef itk::Image ImageType; - typedef itk::Image MaskType; - typedef itk::LabelStatisticsImageFilter FilterType; - typedef typename FilterType::HistogramType HistogramType; - typedef typename HistogramType::IndexType HIndexType; - typedef itk::MinimumMaximumImageCalculator MinMaxComputerType; + typedef itk::Image ImageType; + typedef itk::Image MaskType; typename MaskType::Pointer maskImage = MaskType::New(); mitk::CastToItkImage(mask, maskImage); + // + // Calculate the Volume of Voxel (Maximum to the 3th order) + // double voxelVolume = 1; for (unsigned int i = 0; i < std::min(3, VImageDimension); ++i) voxelVolume *= itkImage->GetSpacing()[i]; + + // + // Calculate the Hypervolume of Voxel + // double voxelSpace = 1; for (unsigned int i = 0; i < VImageDimension; ++i) voxelSpace *= itkImage->GetSpacing()[i]; + double minimum = std::numeric_limits::max(); + double maximum = std::numeric_limits::lowest(); + + double sum = 0; + double sumTwo= 0; + double sumThree = 0; + + unsigned int numberOfVoxels = 0; + + itk::ImageRegionIterator imageIter(itkImage, itkImage->GetLargestPossibleRegion()); + itk::ImageRegionIterator maskIter(maskImage, maskImage->GetLargetPossibleRegion()); + + while (!imageIter.IsAtEnd()) + { + if (maskIter.Get() > 0) + { + value = imageIter.Get(); + + minimum = std::min(minimum, value); + maximum = std::max(maximum, value); + + sum += value; + sumTwo += value * value; + sumThree += value * value*value; + + ++numberOfVoxels; + } + ++maskiter; + ++imageIter; + } + + double mean = sum / numberOfVoxels; + double energy = sumTwo; + double rootMeanSquare = std::sqrt(sumTwo / numberOfVoxels); + + + + featureList.push_back(std::make_pair(params.prefix + "Mean", mean)); + featureList.push_back(std::make_pair(params.prefix + "Energy", energy)); + featureList.push_back(std::make_pair(params.prefix + "Root mean square", rootMeanSquare)); + + return; + + + + + + + + + typedef itk::LabelStatisticsImageFilter FilterType; + typedef typename FilterType::HistogramType HistogramType; + typedef typename HistogramType::IndexType HIndexType; + typedef itk::MinimumMaximumImageCalculator MinMaxComputerType; + + + + typename MinMaxComputerType::Pointer minMaxComputer = MinMaxComputerType::New(); minMaxComputer->SetImage(itkImage); minMaxComputer->Compute(); double imageRange = minMaxComputer->GetMaximum() - minMaxComputer->GetMinimum(); typename FilterType::Pointer labelStatisticsImageFilter = FilterType::New(); labelStatisticsImageFilter->SetInput( itkImage ); labelStatisticsImageFilter->SetLabelInput(maskImage); labelStatisticsImageFilter->SetUseHistograms(true); double min = params.MinimumIntensity; double max = params.MaximumIntensity; labelStatisticsImageFilter->SetHistogramParameters(params.Bins, min,max); labelStatisticsImageFilter->Update(); // --------------- Range -------------------- double range = labelStatisticsImageFilter->GetMaximum(1) - labelStatisticsImageFilter->GetMinimum(1); // --------------- Uniformity, Entropy -------------------- double count = labelStatisticsImageFilter->GetCount(1); //double std_dev = labelStatisticsImageFilter->GetSigma(1); double mean = labelStatisticsImageFilter->GetMean(1); double median = labelStatisticsImageFilter->GetMedian(1); auto histogram = labelStatisticsImageFilter->GetHistogram(1); bool histogramIsCalculated = histogram; HIndexType index; index.SetSize(1); double uniformity = 0; double entropy = 0; double squared_sum = 0; double kurtosis = 0; double mean_absolut_deviation = 0; double median_absolut_deviation = 0; double skewness = 0; double sum_prob = 0; double binWidth = 0; double p05th = 0, p10th = 0, p15th = 0, p20th = 0, p25th = 0, p30th = 0, p35th = 0, p40th = 0, p45th = 0, p50th = 0; double p55th = 0, p60th = 0, p65th = 0, p70th = 0, p75th = 0, p80th = 0, p85th = 0, p90th = 0, p95th = 0; double voxelValue = 0; if (histogramIsCalculated) { binWidth = histogram->GetBinMax(0, 0) - histogram->GetBinMin(0, 0); p05th = histogram->Quantile(0, 0.05); p10th = histogram->Quantile(0, 0.10); p15th = histogram->Quantile(0, 0.15); p20th = histogram->Quantile(0, 0.20); p25th = histogram->Quantile(0, 0.25); p30th = histogram->Quantile(0, 0.30); p35th = histogram->Quantile(0, 0.35); p40th = histogram->Quantile(0, 0.40); p45th = histogram->Quantile(0, 0.45); p50th = histogram->Quantile(0, 0.50); p55th = histogram->Quantile(0, 0.55); p60th = histogram->Quantile(0, 0.60); p65th = histogram->Quantile(0, 0.65); p70th = histogram->Quantile(0, 0.70); p75th = histogram->Quantile(0, 0.75); p80th = histogram->Quantile(0, 0.80); p85th = histogram->Quantile(0, 0.85); p90th = histogram->Quantile(0, 0.90); p95th = histogram->Quantile(0, 0.95); } double Log2=log(2); double mode_bin; double mode_value = 0; double variance = 0; if (histogramIsCalculated) { for (int i = 0; i < (int)(histogram->GetSize(0)); ++i) { index[0] = i; double prob = histogram->GetFrequency(index); if (prob < 0.00000001) continue; voxelValue = histogram->GetBinMin(0, i) + binWidth * 0.5; if (prob > mode_value) { mode_value = prob; mode_bin = voxelValue; } sum_prob += prob; squared_sum += prob * voxelValue*voxelValue; prob /= count; mean_absolut_deviation += prob* std::abs(voxelValue - mean); median_absolut_deviation += prob* std::abs(voxelValue - median); variance += prob * (voxelValue - mean) * (voxelValue - mean); kurtosis += prob* (voxelValue - mean) * (voxelValue - mean) * (voxelValue - mean) * (voxelValue - mean); skewness += prob* (voxelValue - mean) * (voxelValue - mean) * (voxelValue - mean); uniformity += prob*prob; if (prob > 0) { entropy += prob * std::log(prob) / Log2; } } } entropy = -entropy; double uncorrected_std_dev = std::sqrt(variance); double rms = std::sqrt(squared_sum / count); kurtosis = kurtosis / (variance * variance); skewness = skewness / (variance * uncorrected_std_dev); double coveredGrayValueRange = range / imageRange; double coefficient_of_variation = (mean == 0) ? 0 : std::sqrt(variance) / mean; double quantile_coefficient_of_dispersion = (p75th - p25th) / (p75th + p25th); //Calculate the robust mean absolute deviation //First, set all frequencies to 0 that are <10th or >90th percentile double meanRobust = 0.0; double robustMeanAbsoluteDeviation = 0.0; if (histogramIsCalculated) { for (int i = 0; i < (int)(histogram->GetSize(0)); ++i) { index[0] = i; if (histogram->GetBinMax(0, i) < p10th) { histogram->SetFrequencyOfIndex(index, 0); } else if (histogram->GetBinMin(0, i) > p90th) { histogram->SetFrequencyOfIndex(index, 0); } } //Calculate the mean for (int i = 0; i < (int)(histogram->GetSize(0)); ++i) { index[0] = i; meanRobust += histogram->GetFrequency(index) * 0.5 * (histogram->GetBinMin(0, i) + histogram->GetBinMax(0, i)); } meanRobust = meanRobust / histogram->GetTotalFrequency(); for (int i = 0; i < (int)(histogram->GetSize(0)); ++i) { index[0] = i; robustMeanAbsoluteDeviation += std::abs(histogram->GetFrequency(index) * ((0.5 * (histogram->GetBinMin(0, i) + histogram->GetBinMax(0, i))) - meanRobust )); } robustMeanAbsoluteDeviation = robustMeanAbsoluteDeviation / histogram->GetTotalFrequency(); } featureList.push_back(std::make_pair(params.prefix + "Mean", labelStatisticsImageFilter->GetMean(1))); featureList.push_back(std::make_pair(params.prefix + "Unbiased Variance", labelStatisticsImageFilter->GetVariance(1))); //Siehe Definition von Unbiased Variance estimation. (Wird nicht durch n sondern durch n-1 normalisiert) featureList.push_back(std::make_pair(params.prefix + "Biased Variance", variance)); featureList.push_back(std::make_pair(params.prefix + "Skewness", skewness)); featureList.push_back(std::make_pair(params.prefix + "Kurtosis", kurtosis)); featureList.push_back(std::make_pair(params.prefix + "Median", labelStatisticsImageFilter->GetMedian(1))); featureList.push_back(std::make_pair(params.prefix + "Minimum", labelStatisticsImageFilter->GetMinimum(1))); featureList.push_back(std::make_pair(params.prefix + "Maximum", labelStatisticsImageFilter->GetMaximum(1))); featureList.push_back(std::make_pair(params.prefix + "Range", range)); featureList.push_back(std::make_pair(params.prefix + "Mean Absolute Deviation", mean_absolut_deviation)); featureList.push_back(std::make_pair(params.prefix + "Robust Mean Absolute Deviation", robustMeanAbsoluteDeviation)); featureList.push_back(std::make_pair(params.prefix + "Median Absolute Deviation", median_absolut_deviation)); featureList.push_back(std::make_pair(params.prefix + "Coefficient Of Variation", coefficient_of_variation)); featureList.push_back(std::make_pair(params.prefix + "Quantile Coefficient Of Dispersion", quantile_coefficient_of_dispersion)); featureList.push_back(std::make_pair(params.prefix + "Energy", squared_sum)); featureList.push_back(std::make_pair(params.prefix + "Root Mean Square", rms)); typename HistogramType::MeasurementVectorType mv(1); mv[0] = 0; typename HistogramType::IndexType resultingIndex; histogram->GetIndex(mv, resultingIndex); featureList.push_back(std::make_pair(params.prefix + "Robust Mean", meanRobust)); featureList.push_back(std::make_pair(params.prefix + "Uniformity", uniformity)); featureList.push_back(std::make_pair(params.prefix + "Entropy", entropy)); featureList.push_back(std::make_pair(params.prefix + "Excess Kurtosis", kurtosis - 3)); featureList.push_back(std::make_pair(params.prefix + "Covered Image Intensity Range", coveredGrayValueRange)); featureList.push_back(std::make_pair(params.prefix + "Sum", labelStatisticsImageFilter->GetSum(1))); featureList.push_back(std::make_pair(params.prefix + "Mode", mode_bin)); featureList.push_back(std::make_pair(params.prefix + "Mode Probability", mode_value)); featureList.push_back(std::make_pair(params.prefix + "Unbiased Standard deviation", labelStatisticsImageFilter->GetSigma(1))); featureList.push_back(std::make_pair(params.prefix + "Biased Standard deviation", sqrt(variance))); featureList.push_back(std::make_pair(params.prefix + "Number Of Voxels", labelStatisticsImageFilter->GetCount(1))); featureList.push_back(std::make_pair(params.prefix + "05th Percentile", p05th)); featureList.push_back(std::make_pair(params.prefix + "10th Percentile", p10th)); featureList.push_back(std::make_pair(params.prefix + "15th Percentile", p15th)); featureList.push_back(std::make_pair(params.prefix + "20th Percentile", p20th)); featureList.push_back(std::make_pair(params.prefix + "25th Percentile", p25th)); featureList.push_back(std::make_pair(params.prefix + "30th Percentile", p30th)); featureList.push_back(std::make_pair(params.prefix + "35th Percentile", p35th)); featureList.push_back(std::make_pair(params.prefix + "40th Percentile", p40th)); featureList.push_back(std::make_pair(params.prefix + "45th Percentile", p45th)); featureList.push_back(std::make_pair(params.prefix + "50th Percentile", p50th)); featureList.push_back(std::make_pair(params.prefix + "55th Percentile", p55th)); featureList.push_back(std::make_pair(params.prefix + "60th Percentile", p60th)); featureList.push_back(std::make_pair(params.prefix + "65th Percentile", p65th)); featureList.push_back(std::make_pair(params.prefix + "70th Percentile", p70th)); featureList.push_back(std::make_pair(params.prefix + "75th Percentile", p75th)); featureList.push_back(std::make_pair(params.prefix + "80th Percentile", p80th)); featureList.push_back(std::make_pair(params.prefix + "85th Percentile", p85th)); featureList.push_back(std::make_pair(params.prefix + "90th Percentile", p90th)); featureList.push_back(std::make_pair(params.prefix + "95th Percentile", p95th)); featureList.push_back(std::make_pair(params.prefix + "Interquartile Range", (p75th - p25th))); featureList.push_back(std::make_pair(params.prefix + "Image Dimension", VImageDimension)); featureList.push_back(std::make_pair(params.prefix + "Voxel Space", voxelSpace)); featureList.push_back(std::make_pair(params.prefix + "Voxel Volume", voxelVolume)); } mitk::GIFFirstOrderNumericStatistics::GIFFirstOrderNumericStatistics() { SetShortName("fon"); SetLongName("first-order-numeric"); SetFeatureClassName("First Order Numeric"); } mitk::GIFFirstOrderNumericStatistics::FeatureListType mitk::GIFFirstOrderNumericStatistics::CalculateFeatures(const Image::Pointer & image, const Image::Pointer &mask) { InitializeQuantifier(image, mask); FeatureListType featureList; - ParameterStruct params; + FirstOrderNumericParameterStruct params; + params.quantifier = GetQuantifier(); params.MinimumIntensity = GetQuantifier()->GetMinimum(); params.MaximumIntensity = GetQuantifier()->GetMaximum(); params.Bins = GetQuantifier()->GetBins(); params.prefix = FeatureDescriptionPrefix(); AccessByItk_3(image, CalculateFirstOrderStatistics, mask, featureList, params); return featureList; } mitk::GIFFirstOrderNumericStatistics::FeatureNameListType mitk::GIFFirstOrderNumericStatistics::GetFeatureNames() { FeatureNameListType featureList; - featureList.push_back("First Order::Minimum"); - featureList.push_back("First Order::Maximum"); - featureList.push_back("First Order::Mean"); - featureList.push_back("First Order::Variance"); - featureList.push_back("First Order::Sum"); - featureList.push_back("First Order::Median"); - featureList.push_back("First Order::Standard deviation"); - featureList.push_back("First Order::No. of Voxel"); return featureList; } void mitk::GIFFirstOrderNumericStatistics::AddArguments(mitkCommandLineParser &parser) { std::string name = GetOptionPrefix(); parser.addArgument(GetLongName(), name, mitkCommandLineParser::Bool, "Use First Order Statistic (Numeric)", "calculates First Order Statistic (Numeric)", us::Any()); AddQuantifierArguments(parser); } void mitk::GIFFirstOrderNumericStatistics::CalculateFeaturesUsingParameters(const Image::Pointer & feature, const Image::Pointer &, const Image::Pointer &maskNoNAN, FeatureListType &featureList) { auto parsedArgs = GetParameter(); if (parsedArgs.count(GetLongName())) { InitializeQuantifierFromParameters(feature, maskNoNAN); MITK_INFO << "Start calculating first order features ...."; auto localResults = this->CalculateFeatures(feature, maskNoNAN); featureList.insert(featureList.end(), localResults.begin(), localResults.end()); MITK_INFO << "Finished calculating first order features...."; } } std::string mitk::GIFFirstOrderNumericStatistics::GetCurrentFeatureEncoding() { return QuantifierParameterString(); } \ No newline at end of file diff --git a/Modules/Classification/CLUtilities/test/files.cmake b/Modules/Classification/CLUtilities/test/files.cmake index 4dc57754cc..0013ebafcc 100644 --- a/Modules/Classification/CLUtilities/test/files.cmake +++ b/Modules/Classification/CLUtilities/test/files.cmake @@ -1,16 +1,17 @@ set(MODULE_TESTS mitkGIFCooc2Test mitkGIFCurvatureStatisticTest mitkGIFFirstOrderHistogramStatisticsTest + mitkGIFFirstOrderNumericStatisticsTest mitkGIFGreyLevelDistanceZoneTest mitkGIFGreyLevelSizeZoneTest mitkGIFImageDescriptionFeaturesTest mitkGIFIntensityVolumeHistogramTest mitkGIFLocalIntensityTest mitkGIFNeighbourhoodGreyToneDifferenceFeaturesTest mitkGIFNeighbouringGreyLevelDependenceFeatureTest mitkGIFVolumetricDensityStatisticsTest mitkGIFVolumetricStatisticsTest #mitkSmoothedClassProbabilitesTest.cpp #mitkGlobalFeaturesTest.cpp ) diff --git a/Modules/Classification/CLUtilities/test/mitkGIFFirstOrderNumericStatisticsTest.cpp b/Modules/Classification/CLUtilities/test/mitkGIFFirstOrderNumericStatisticsTest.cpp new file mode 100644 index 0000000000..bc64347e97 --- /dev/null +++ b/Modules/Classification/CLUtilities/test/mitkGIFFirstOrderNumericStatisticsTest.cpp @@ -0,0 +1,82 @@ +/*=================================================================== + +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. + +===================================================================*/ + +#include +#include +#include "mitkIOUtil.h" +#include + +#include + +class mitkGIFFirstOrderNumericStatisticsTestSuite : public mitk::TestFixture +{ + CPPUNIT_TEST_SUITE(mitkGIFFirstOrderNumericStatisticsTestSuite); + + MITK_TEST(ImageDescription_PhantomTest); + + CPPUNIT_TEST_SUITE_END(); + +private: + mitk::Image::Pointer m_IBSI_Phantom_Image_Small; + mitk::Image::Pointer m_IBSI_Phantom_Image_Large; + mitk::Image::Pointer m_IBSI_Phantom_Mask_Small; + mitk::Image::Pointer m_IBSI_Phantom_Mask_Large; + +public: + + void setUp(void) override + { + m_IBSI_Phantom_Image_Small = mitk::IOUtil::Load(GetTestDataFilePath("Radiomics/IBSI_Phantom_Image_Small.nrrd")); + m_IBSI_Phantom_Image_Large = mitk::IOUtil::Load(GetTestDataFilePath("Radiomics/IBSI_Phantom_Image_Large.nrrd")); + m_IBSI_Phantom_Mask_Small = mitk::IOUtil::Load(GetTestDataFilePath("Radiomics/IBSI_Phantom_Mask_Small.nrrd")); + m_IBSI_Phantom_Mask_Large = mitk::IOUtil::Load(GetTestDataFilePath("Radiomics/IBSI_Phantom_Mask_Large.nrrd")); + } + + void ImageDescription_PhantomTest() + { + mitk::mitkGIFFirstOrderNumericStatistics::Pointer featureCalculator = mitk::mitkGIFFirstOrderNumericStatistics::New(); + + featureCalculator->SetUseBinsize(true); + featureCalculator->SetBinsize(1.0); + featureCalculator->SetUseMinimumIntensity(true); + featureCalculator->SetUseMaximumIntensity(true); + featureCalculator->SetMinimumIntensity(0.5); + featureCalculator->SetMaximumIntensity(6.5); + + auto featureList = featureCalculator->CalculateFeatures(m_IBSI_Phantom_Image_Large, m_IBSI_Phantom_Mask_Large); + + std::map results; + for (auto valuePair : featureList) + { + MITK_INFO << valuePair.first << " : " << valuePair.second; + results[valuePair.first] = valuePair.second; + } + CPPUNIT_ASSERT_EQUAL_MESSAGE("Image Diagnostics should calculate 46 features.", std::size_t(46), featureList.size()); + + // These values are obtained by a run of the filter. + // The might be wrong! + CPPUNIT_ASSERT_DOUBLES_EQUAL_MESSAGE("First Order Histogram::Mean Value should be 2.15 with Large IBSI Phantom Image", 2.15, results["First Order Histogram::Mean Value"], 0.01); + + + // These values are taken from the IBSI Initiative to ensure compatibility + // The values are given with an accuracy of 0.01 + CPPUNIT_ASSERT_DOUBLES_EQUAL_MESSAGE("First Order Histogram::Mean Index should be 2.15 with Large IBSI Phantom Image", 2.15, results["First Order Histogram::Mean Index"], 0.01); + + } + +}; + +MITK_TEST_SUITE_REGISTRATION(mitkGIFFirstOrderNumericStatistics ) \ No newline at end of file