Metric measures (medpy.metric)

This package provides a number of metric measures that e.g. can be used for testing and/or evaluation purposes on two binary masks (i.e. measuring their similarity) or distance between histograms.

Binary metrics (medpy.metric.binary)

Metrics to compare binary objects and classification results.

Compare two binary objects

dc(result, reference) Dice coefficient
jc(result, reference) Jaccard coefficient
hd(result, reference[, voxelspacing, …]) Hausdorff Distance.
asd(result, reference[, voxelspacing, …]) Average surface distance metric.
assd(result, reference[, voxelspacing, …]) Average symmetric surface distance.
precision(result, reference) Precison.
recall(result, reference) Recall.
sensitivity(result, reference) Sensitivity.
specificity(result, reference) Specificity.
true_positive_rate(result, reference) True positive rate.
true_negative_rate(result, reference) True negative rate.
positive_predictive_value(result, reference) Positive predictive value.
ravd(result, reference) Relative absolute volume difference.

Compare two sets of binary objects

obj_tpr(result, reference[, connectivity]) The true positive rate of distinct binary object detection.
obj_fpr(result, reference[, connectivity]) The false positive rate of distinct binary object detection.
obj_asd(result, reference[, voxelspacing, …]) Average surface distance between objects.
obj_assd(result, reference[, voxelspacing, …]) Average symmetric surface distance.

Compare to sequences of binary objects

volume_correlation(results, references) Volume correlation.
volume_change_correlation(results, references) Volume change correlation.

Image metrics (medpy.metric.image)

Some more image metrics (e.g. sls and ssd) can be found in medpy.filter.image.

mutual_information(i1, i2[, bins]) Computes the mutual information (MI) (a measure of entropy) between two images.

Histogram metrics (medpy.metric.histogram)

chebyshev(h1, h2) Chebyshev distance.
chebyshev_neg(h1, h2) Chebyshev negative distance.
chi_square(h1, h2) Chi-square distance.
correlate(h1, h2) Correlation between two histograms.
correlate_1(h1, h2) Correlation distance.
cosine(h1, h2) Cosine simmilarity.
cosine_1(h1, h2) Cosine simmilarity.
cosine_2(h1, h2) Cosine simmilarity.
cosine_alt(h1, h2) Alternative implementation of the cosine distance measure.
euclidean(h1, h2) Equal to Minowski distance with \(p=2\).
fidelity_based(h1, h2) Fidelity based distance.
histogram_intersection(h1, h2) Calculate the common part of two histograms.
histogram_intersection_1(h1, h2) Turns the histogram intersection similarity into a distance measure for normalized, positive histograms.
jensen_shannon(h1, h2) Jensen-Shannon divergence.
kullback_leibler(h1, h2) Kullback-Leibler divergence.
manhattan(h1, h2) Equal to Minowski distance with \(p=1\).
minowski(h1, h2[, p]) Minowski distance.
noelle_1(h1, h2) Extension of fidelity_based proposed by [R39].
noelle_2(h1, h2) Extension of fidelity_based proposed by [R40].
noelle_3(h1, h2) Extension of fidelity_based proposed by [R41].
noelle_4(h1, h2) Extension of fidelity_based proposed by [R42].
noelle_5(h1, h2) Extension of fidelity_based proposed by [R43].
quadratic_forms(h1, h2) Quadrativ forms metric.
relative_bin_deviation(h1, h2) Calculate the bin-wise deviation between two histograms.
relative_deviation(h1, h2) Calculate the deviation between two histograms.