Speckle classification for sensorless freehand 3-D ultrasound

Novel reconstruction and feature exploitation techniques for sensorless freehand 3D ultrasound

“…18,19 In freehand experiments the images are not parallel as they are in our experiments, and therefore the rotations between the images need to be found. 11,13,20 We showed before 15 that performing beam steering significantly increases the accuracy of out-of-plane motion estimation. In this work, we showed that MLE can also be used to enhance the out-of-plane motion estimation.…”

“…11 Gee et al 13 proposed a heuristic technique that is robust to the lack of FDS patches in the ultrasound image. This method allows the calculation of the elevational distance for all patches of the image, regardless of their level of coherency, by measuring the axial and lateral correlation of each patch.…”

Learning to estimate out-of-plane motion in ultrasound imagery of real tissue

“…Methods for detecting fully developed speckle have mostly relied on the well-understood first order statistical properties of US signals under Rayleigh scattering conditions, namely the Gaussianity of the radio-frequency (RF) signal (Georgiou and Cohen, 1998), the exponential statistics of the echo intensity signal (Tuthill et al, 1998) and the fixed normalised first order moments of fractional powers of the echo envelope Rivaz et al, 2006). Alternatively, Hassenpflug et al (2005) proposed to detect fully developed speckle using the transducer-specificity of second order image statistics, based on the degree of agreement between the axial decorrelation curve of the data and that obtained for speckle phantom imagery.…”

“…Another fundamental problem with the speckle decorrelation approach is that, even when accounting for all of the aforementioned sources of error, the relationship between image correlation and displacement is not one-to-one except under the narrowly defined Rayleigh scattering conditions. This seriously limits the accuracy of the approach when imaging biological tissue because these conditions are rarely met closely enough for the nominal model to apply (Hassenpflug et al, 2005;Gee et al, 2006). The micro-structure of real tissue may not be as densely packed as that of a speckle phantom and may exhibit non-random scatterer configurations in the form of clustered or periodic patterns.…”

“…To address this issue, most prior work relies on filters for finding regions of fully developed speckle where the nominal model applies, which leads to poor or overly sparse motion estimates in imagery of arbitrary tissues (Hassenpflug et al, 2005). A more challenging, but more promising approach is to locally adapt the decorrelation model to the characteristics of the medium .…”