This configuration dialog controls all of the parameters which affect how strain images are calculated from frames of ultrasound RF data. See this page for a high level overview of Stradwin's 2D and 3D strain imaging facilities. The strain settings are grouped into four boxes, roughly in order of processing, from left to right:
The 'elevational' and '3D' checkboxes are only relevant when using 3D data from a mechanically swept probe. For 2D strain images, the 'elevational' setting is ignored, and checking the '3D' box has the same effect as the '2D' box.
The first list box determines which algorithm is used to calculate the strain image: EPZS (Efficient Phase Zero Search), WPS (Weighted Phase Separation) or EWPS (Efficient Weighted Phase Separation). AMC (Amplitude Modulation Correction), which can be added to either algorithm, is used to reposition the displacement estimates to their ideal locations. PV (Phase Variance) uses the variance of the residual phase after window matching as the quality indication, rather than using signal correlation.
All of the above algorithms have various weighting options which can be used to modify their behaviour. These are controlled by the 'weightings' section. EPZS can be run with log compression of the amplitude of the RF data. With WPS, the phase and amplitude weightings can be controlled separately. There is also an option to use weightings more appropriate for very low strain data sets.
The next slider controls the lateral range of data examined in performing dropout correction in any of the above algorithms. Setting this slider to '0' will disable dropout correction. Clicking the 'top' checkbox causes tracking to start at the top, and progress downwards, rather than starting at the bottom of the image. The 'zero' box uses an initial guess of zero displacement at the start: if left blank, the displacement is initialised with a set of guesses which vary laterally across the image.
The next line controls whether to track the displacements laterally (across the image) and elevationally (across frames in a 3D data set) as well as axially (down the image). The 'smooth' box enables lateral and elevational tracking with a different algorithm which generates smoother estimates and is slightly faster.
In order to speed up processing, the RF signal can be sub-sampled after conversion to a baseband analytic signal. The sub-sampling factor is set by the first slider in this section. The maximum possible sub-sampling before performance is degraded is determined by the initial sampling rate (set in the image configuration dialog) and the bandwidth of the ultrasound pulse.
The next two sliders control the window length (in cycles of the ultrasound centre frequency) and spacing (in display pixels) for the matching windows used to locate the same data in different frames, and subsequently to derive a relative displacement of that data. Longer window lengths and smaller spacings will result in strain images with lower noise, but also lower resolution. Short windows and small spacings increase the speed of the algorithm but at the cost of increased signal noise. Checking the 'lat' box will cause Stradwin to use a 2D window for matching, whose width (lateral dimension) is the same as the length (axial dimension). The '2D' box will allow Stradwin to skip RF vectors as well as vertical pixels when calculating strain. The 'el' and '3D' checkboxes have similar effects for 3D data. A line of text below these sliders indicates the resulting window size and overlap using these two parameters.
The strain image is related to the gradient of the calculated displacements: the next slider controls how many display pixels are used in estimating each gradient. Setting this to '2' will use a simple centred-difference for gradient estimation. Larger values perform gradient estimation by fitting a line, in the least-squares sense, to the displacement estimates. Different gradient estimates are calculated for every image pixel. The '2D' check box causes Stradwin to fit a plane, rather than a line, and then take the axial gradient of this plane. The '3D' check box has a similar effect for 3D data.
Note that setting the gradient slider to '0' will turn off gradient calculation: in this case the raw displacement values are displayed, rather than the strain.
The normalisation controls determine the way in which Stradwin estimates the stress in the acquired RF data, in order to display a pseudo-strain image, which is partially normalised for stress. The first list box controls the function which is fitted to (and then removed from) the strain data, and the sliders determine the range of normalised strain values which are displayed and the slope in degrees of the strain-to-pseudo-strain curve at 0. Setting 'range' to '2.0' and 'angle' to '45' results in the linear display of pseudo-strain values from 0 to twice the normalised average strain.
If you do not wish to display pseudo-strain, but instead wish to see the actual strain, set the normalisation function to 'Disabled'. The 'range' slider now sets the range (in % strain) of values displayed on a linear colourmap scale.
Hovering the mouse above Stradwin's top-left strain image causes the local strain (or pseudo-strain) at the pixel below the pointer to be displayed in Stradwin's status bar.
The next two controls allow a filter to be applied to the strain data. This can be either a mean, median or Gaussian filter (or none at all), over a number of displacement windows (in all dimensions) according to the last slider. Larger filters will obviously smooth the data more, but also reduce the resolution and take more time to perform. There is also a 'Regularise' option which uses nonparametric regression to smooth the strain data such that the resulting data has uniform precision. In this case, the slider controls the smoothing strength: the effective size of the filter will vary according to the quality of the image data.
Finally, the last slider controls how many frames can be used to generate a weighted strain image in a similar way to the persistence controls on a conventional ultrasound machine. A larger value will give a higher quality strain image so long as the probe remains substantially over the same anatomy during the acquisition of these frames. Smaller values will improve the response time of the strain image.
The first display option controls what data to display. For a normal strain image, this is 'axial strain', but other options are 'axial shear strain', 'lateral strain', 'elevational strain' or 'quality value'. The lateral and elevational strains can only be visualised if tracking was enabled in the corresponding direction. The 'quality value' display allows direct display of the strain quality value (mapped to a colour using the 'acceptable quality' slider) used in all the processing steps.
The strain can be displayed using several different colour schemes, which can be selected with the next list box. Unlike most other controls, changing the display colour does not cause the strain image to be recalculated. It is also possible to invert the mapping between strain and colour in each scheme.
The 'acceptable quality level' slider controls whether strain data of a given quality will be displayed at full resolution, or as black/red, depending on the colour map. Lastly, the 'image opacity' slider controls whether the strain data is overlayed on a B-scan image. A value less than '100' will allow the B-scan image to be seen. This slider only affects the overall opacity, locally the image can be more translucent if the quality value is low.