Repository

CLI Overview

The maintained entry point is:

julia --project=. scripts/run_pam.jl --option=value

The focusing scripts are also available:

julia --project=. scripts/run_focus.jl --option=value
julia --project=. scripts/compare_focus_estimators.jl --option=value

All scripts use --name=value arguments. Positional arguments and space-separated forms such as --name value are not supported.

Output Directories

Runs write to an automatically named directory under outputs/. The name includes the date, main run type, medium, source model, grid size, and other identifying parameters. Use --out-dir=/path/to/output when a fixed location is needed.

Typical outputs include:

  • summary.json: run settings, source metadata, metrics, and output paths.
  • result.jld2: numerical arrays and structured data for later analysis.
  • overview.png: PAM source/reconstruction overview where applicable.
  • activity_boundaries.png: detection-threshold visualization for activity sources.
  • pressure.png or comparison.png: focusing figures.

GPU Flags

PAM separates the forward simulator and reconstruction GPU switches:

  • --kwave-use-gpu=true controls k-Wave forward simulation.
  • --recon-use-gpu=true controls CUDA.jl ASA/HASA reconstruction.

Older notes may mention --use-gpu; the current PAM runner uses the two explicit flags above.

Coordinate Conventions

PAM coordinates are specified in millimeters relative to the receiver/transducer plane:

  • 2D point or anchor coordinates use depth:lateral.
  • 3D point or anchor coordinates use depth:y:z.
  • Comma-separated values specify multiple coordinates.

Examples:

--sources-mm=30:0
--sources-mm=25:-6,32:0,40:8
--sources-mm=30:2:-1
--anchors-mm=45:0:0

Depth is positive into the medium. Lateral coordinates are centered around the transducer/receiver axis.