Medical Imaging Interaction Toolkit  2021.10.00
Medical Imaging Interaction Toolkit


To check the correctness of the hotspot calculation, this special class has been created, which generates images with known hotspot location and statistics. A number of unit tests use this class to first generate an image of known properties and then verify that mitk::ImageStatisticsCalculator is able to reproduce the known statistics.

Every testcase has a defined hotspot, maximum and minimum including their corresponding index-values and mean value. The XML-files to each testcase is located in Modules/ImageStatistics/Testing/Data.

This test checks the hotspot-statistics of images with a spacing[x,y,z] of [3,3,3], [4,4,3] and [5,5,5], according to common PET-resolutions.

The following cases describe situations of hotspot-calculation and their supposed results.

Note: Below only the behaviour of maximum is mentioned mostly, but the other statistics (minimum and mean) behave in the same way like maximum.

Testcase 1: No values outside of hotspot are used for statistic-calculation

The purpose of this testcase is primarily to confirm the correct detection of the hotspot even if there is an global maximum which is "hotter" than the mean value itself. On the other hand the test verifies that only voxels are used for statistic-calculation which are located in the hotspot.



Assumed results:

Testcase 2: Correct detection of hotspot

In this testcase we want to make sure that when a segmentation is available the origin of the hotspot-sphere is located within it. The image is so structured that there are two hot regions: One region inside and another one, which is hotter than the other region, outside the segmentation. So we can assume that the segmentation is also considered when detecting the hotspot, even an actual hotspot outside the segmentation exists.



Assumed results:

Testcase 3: Correct calculation of statistics in hotspot, although the whole hotspot is not inside segmentation

The difficulty of calculating the hotspot statistics in testcase 3 is that the origin of the hotspot is close to the segmentation-borders. So if the whole hotspot is not inside the segmentation (or even the segmentation is smaller than the hotspot itself) this test checks that calculation of hotspot statistics is possible anyway.



Assumed results:

Testcase 4 and 5: Hotspot must (not) be completely inside image

Testcase 4 and 5 are very similar so we mention it at the same time: In testcase 4 the hotspot is not completely inside the image and just voxels are considered for calculation which are located inside the image. But in testcase 5 the hotspot must be completely inside the image even there is an possible hotspot-location at the borders of the image.



Assumed results in testcase 4:

Assumed results in testcase 5:

Testcase 6: Multi label mask

This testcase confirms that mitkImageStatisticsCalculator has the possibility to calculate hotspot statistics even if there are multiple regions of interest.



Assumed results: