Skip to content

Limited Acoustic FOV

Limited acoustic FOV

Running the full k-Wave acoustic simulation on a whole-head grid can be slow and memory-intensive when only the focal zone and near-field matter. The acoustic FOV feature crops the k-Wave domain to a beam-axis-aligned box before simulation and back-projects the result onto the full grid afterwards. The thermal simulation always runs on the full head grid, regardless of any acoustic FOV setting.

Quick start

Add the following keys to your study YAML (or set them on the parameters struct before calling prestus_pipeline_start):

grid:
  acoustic_fov_diameter_mm: 60   # lateral width [mm], centred on the focus
  acoustic_fov_length_mm:  120   # axial depth [mm], starting at the transducer

Choose values that comfortably enclose the focal zone and any near-field region of interest. The pipeline prints the resulting sub-grid dimensions and offset at runtime so you can verify the crop before committing to a full run.

How the FOV box is defined

The FOV is a rectangular box in the beam-axis-aligned acoustic grid:

Dimension Extent Anchor
Axial (beam axis, z in transducer_axis mode) acoustic_fov_length_mm starts at the transducer centre
Lateral (x, y) acoustic_fov_diameter_mm (full width) centred on the focal point

The axial start is placed at the transducer centre rather than at the transducer face or the focus. This ensures the entire near-field between the transducer and the focus is captured, which is important for skull heating metrics. The axial end should extend at least a few wavelengths beyond the focus to avoid boundary artefacts from the PML.

The box is clamped to the full acoustic grid. If acoustic_fov_length_mm is too short for the focus to fall inside, a warning is raised and acoustic_fov_length_mm should be increased.

Pipeline stages affected

Stage Effect
Medium setup (Stage 4) Full-grid state is saved before the crop.
FOV crop acoustic_grid_fov crops kwave_medium, medium_masks, segmentation, trans_pos, focus_pos, and parameters.grid.dims to the sub-grid.
Source/sensor setup (Stage 5) kgrid and the source matrix are built on the sub-grid.
Acoustic simulation (Stage 6) k-Wave runs on the sub-grid; sensor_data.p_max_all is sub-grid sized.
Acoustic analysis & NIfTI export (Stage 7) Analysis runs on the sub-grid; outputs are back-projected to the full grid before NIfTI writing so spatial location is correct.
Thermal setup (Stage 8) Full-grid state is fully restored. sensor_data.p_max_all is back-projected to the full grid via fov_backproject_volume. kgrid is rebuilt from the full-grid parameters via thermal_grid_setup.
Thermal simulation & analysis (Stages 8–9) Runs on the full grid. Results are physically correct.

The acoustic cache (.mat file) stores the sub-grid pressure field together with the crop metadata (acoustic_provenance.fov_offset_ac and acoustic_provenance.full_ac_dims). This keeps the cache small and lets assemble_limited_fov_fields back-project and combine multiple limited-FOV results from independent transducer placements.

Why thermal is always full-grid

Thermal conduction has a much longer spatial scale than the acoustic focal zone. Heat deposited at the focus diffuses into surrounding tissue over the sonication duration; cropping the thermal domain would:

  • suppress heat loss to surrounding tissue, overestimating peak temperature rise;
  • miss heating of non-target structures just outside the FOV (e.g. skull surface);
  • produce physically incorrect boundary conditions at the crop edges.

For these reasons the acoustic FOV crop is never propagated to the thermal stage. If faster thermal simulations are needed, use grid.thermal_resolution_mm (coarser grid) or grid.thermal_fov_mm (explicit thermal crop, for prototyping only — see the caveat in doc_parameters.md).

Multi-transducer setups

When multiple transducers each target a different focus with independent limited-FOV acoustic runs, use assemble_limited_fov_fields to combine them before thermal:

assemble_limited_fov_fields( ...
    {'sub-001_layered_results_t1.mat', 'sub-001_layered_results_t2.mat'}, ...
    'sub-001_layered_results_combined.mat', ...
    [target_t1_wcm2, target_t2_wcm2]);

The assembler back-projects each sub-grid onto the shared full acoustic grid, applies incoherent intensity summation, and writes a combined cache that the thermal pipeline reads as a normal full-grid result. The full-head domain is therefore always used for thermal regardless of how many limited-FOV acoustic runs contributed.

See doc_multitransducer.md for the full multi-transducer workflow.

Choosing FOV dimensions

Parameter Too small Too large
acoustic_fov_length_mm Focus outside FOV → warning, clamped focal position, incorrect simulation No harm; wastes grid voxels
acoustic_fov_diameter_mm Lateral beam clipped, boundary reflections near focus No harm; wastes grid voxels

Recommended starting values for a single-element transducer at a ~70 mm depth:

acoustic_fov_diameter_mm: 60    # 3–4× the −6 dB beam diameter at focus
acoustic_fov_length_mm:  120    # transducer–focus distance + ~30 mm margin

At 2 mm resolution the above box is roughly 30 × 30 × 60 voxels (54 k), compared to a typical whole-head grid of ~90 × 90 × 200 voxels (1.6 M): a ~97 % reduction in voxel count.

Print the actual reduction at runtime:

fprintf('FOV: [%s]  full: [%s]  reduction: %.1f %%\n', ...
    num2str(parameters.grid.dims), num2str(full_ac_dims), ...
    100*(1 - prod(parameters.grid.dims)/prod(full_ac_dims)));

Relevant functions

Function Role
acoustic_grid_fov Crops medium, masks, and positions to the FOV sub-grid; returns crop offset.
fov_backproject_volume Zero-pads a FOV-sized array back to a given full-grid size.
assemble_limited_fov_fields Back-projects and incoherently combines N limited-FOV acoustic caches for multi-transducer thermal simulation.
thermal_grid_setup Rebuilds kgrid and resamples medium/masks for the thermal stage; always operates on the full acoustic grid when has_acoustic_fov is true.