net.imglib2.algorithm.metrics.segmentation

Class LazySEGMetrics

java.lang.Object

net.imglib2.algorithm.metrics.segmentation.LazySEGMetrics

public class LazySEGMetrics
extends Object

The LazySEGMetrics computes a running SEGMetrics over all images added by calling addTimePoint. The images are expected to be of dimension XYZ, where Z can be of depth 1.

The score is computed by accumulating the number of ground-truth labels and their corresponding IoU with respect to matching prediction labels. Each contribution is computed by adding images using addTimePoint. The final score can be queried by computeScore().

Each image's contribution is calculated independently. Therefore, the same LazYSEGMetrics object can be called from multiple threads in order to speed up computation. For instance, if the total stack does not fit in memory, lazy loading and multithreading can be used to compute the SEG score by splitting the XYZ images between threads and adding them one by one. The final score can then be calculated once all threads have finished.

Reference: Ulman, V., Maška, M., Magnusson, K. et al. An objective comparison of cell-tracking algorithms. Nat Methods 14, 1141–1152 (2017).

Author:

Joran Deschamps

See Also:

SEGMetrics, Original implementation by Martin Maška and Vladimír Ulman

Constructor Summary

Constructors

Constructor and Description
LazySEGMetrics()

Method Summary

Modifier and Type	Method and Description
<T,I extends net.imglib2.type.numeric.IntegerType<I>,U,J extends net.imglib2.type.numeric.IntegerType<J>> void	addTimePoint(net.imglib2.roi.labeling.ImgLabeling<T,I> groundTruth, net.imglib2.roi.labeling.ImgLabeling<U,J> prediction) Add a new image pair and compute its contribution to the SEG metrics.
<I extends net.imglib2.type.numeric.IntegerType<I>,J extends net.imglib2.type.numeric.IntegerType<J>> void	addTimePoint(net.imglib2.RandomAccessibleInterval<I> groundTruth, net.imglib2.RandomAccessibleInterval<J> prediction) Add a new image pair and compute its contribution to the SEG metrics.
double	computeScore() Compute the total SEG score.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

LazySEGMetrics

public LazySEGMetrics()

Method Detail

addTimePoint

public <T,I extends net.imglib2.type.numeric.IntegerType<I>,U,J extends net.imglib2.type.numeric.IntegerType<J>> void addTimePoint(net.imglib2.roi.labeling.ImgLabeling<T,I> groundTruth,
                                                                                                                                   net.imglib2.roi.labeling.ImgLabeling<U,J> prediction)

Add a new image pair and compute its contribution to the SEG metrics. The current SEG score can be computed by calling computeScore(). This method is not compatible with ImgLabeling with intersecting labels.

Type Parameters:

T - Label type associated to the ground-truth

I - Ground-truth pixel type

U - Label type associated to the prediction

J - Prediction pixel type

Parameters:

groundTruth - Ground-truth image

prediction - Predicted image

addTimePoint

public <I extends net.imglib2.type.numeric.IntegerType<I>,J extends net.imglib2.type.numeric.IntegerType<J>> void addTimePoint(net.imglib2.RandomAccessibleInterval<I> groundTruth,
                                                                                                                               net.imglib2.RandomAccessibleInterval<J> prediction)

Add a new image pair and compute its contribution to the SEG metrics. The current SEG score can be computed by calling computeScore().

Type Parameters:

I - Ground-truth pixel type

J - Prediction pixel type

Parameters:

groundTruth - Ground-truth image

prediction - Predicted image

computeScore

public double computeScore()

Compute the total SEG score. If no image was added, or all images were empty, then the SEG score is NaN.

Returns:

SEG score