Sweep Compounding

Reconstructs a 3D volume from one or multiple ultrasound sweeps.

Input

The algorithm requires one or several ultrasound sweeps, each with a non-zero selection of frames.

Output

A reconstructed 3D volume containing the ultrasound sweeps is created. The input data is not modified.

Description

Using the geometry of the input sweep(s), a predominant axis, bounding box and spacing is automatically determined for the resulting 3D volume. The following parameters may be used to modify the default behavior of the algorithm:

  • Spacing allows to specify a custom spacing, rather than automatic selection.

  • Method selects which device and which gathering method shall be used. The options include:

    • GPU: GPU-based direct compounding scheme with linear interpolation.
    • GPU Nearest: GPU-based direct compounding scheme with nearest neighbor interpolation.
    • GPU Backward: GPU-based high-quality backward compounding scheme with linear interpolation
    • CPU Nearest: CPU-based scheme with nearest-neighbor interpolation
    • CPU Gaussian: CPU-based scheme with Gaussian interpolation
    • CPU Median: CPU-based scheme with median interpolation
    • CPU Max: CPU-based scheme with maximum interpolation
    • CPU Inverse Distance: CPU-based scheme with inverse-distance based interpolation
    • CPU Trajectory: CPU-based scheme with linear interpolation along trajectory

    Note that all CPU-based methods are slow and deprecated backward compounding variants. It is advised to use a GPU method instead.

  • Background intensity defines the image intensity outside of US frames.

  • Frame Thickness describes the maximum distance of a US frame to voxel of the output volume to contribute to its intensity.

  • Range sets the backward compounding range and if the full neighborhood shall be used.

  • Neighborhood defines whether the full neighborhood is to be considered.

  • Bounding Box Mode defines how the resulting volume should be oriented in space. Supported modes are:

    1. Global Coordinates: Volume will be axis-aligned.
    2. Frame Normal: Uses the mean frame normal direction for the z-direction.
    3. Heuristic Alignment: Uses a PCA-based heuristic to find an well containing bounding box for most use cases.
    4. Fit: Runs an iterative optimizer to find the minimal bounding box.

  • Reconstruct the sweep into individual volumes let’s you choose whether all input sweeps are compounded together into a single volume or whether each sweep is compounded into an separate volume (only available if multiple input sweeps selected).