Creatis- Univesity of Lyon || NTNU || Technical University of Denmark


Evaluation framework
Speckle quality

Speckle is an important feature of ultrasound images. It can be the basis for tissue classification, structure segmentation, motion estimation, etc... As a consequence, we assess in the PICMUS framework the ability of the reconstruction methods in preserving fully-developped speckle situation.

To this aim, we involve numerical and physical phantoms that should lead to images with a fully developped speckle for the background. For instance, for the numerical phantom, this property is ensure by randomly putting 20 scatterers per resolution cell. In this condition, it is well-know that the intensity of the envelope image (before log compression) should follow a Rayleigh distribution. For a set of predefined regions, the Kolmogorov-Smirnov (KS) test is thus applied. KSis a widely used statistical hypothesis test that verifies in our case whether there is enough evidence in data to deduce that the hypothesis under consideration (i.e. the data follows a Rayleigh distribution) is true. The tested regions that pass the KS test with significance level α = 0.05 are considered as region where the speckle quality is preserved.

In the figure bellow we show an example of the evaluation of the speckle quality on the numerical phantom where the KS test was succesfully passed for two pre-defined regions.

illustration of speckle quality assessment