Batch Runs

Use mrax.core.batch.run_batch_studies() when you want to process multiple studies back-to-back without any interdependencies between them. The helper:

• Discovers subfolders under a studies root (or a user-specified subset).

• Loads each folder into a mrax.data_struct.study.Study via a study_loader.

• Runs fitting/analysis (planner + executor by default, override with analysis_fn).

• Persists each run in the standard internal layout under a shared output root so the manifests can be loaded later for further analysis.

Quickstart example

Create two synthetic study folders, then batch fit and export them into a common batch_output directory:

from pathlib import Path
import json

from mrax.core.batch import run_batch_studies
from mrax.core.synthetic import make_example_suite
from mrax.data_struct import Quantity, Study, load_study_results

studies_root = Path("batch_inputs")
for seed, name in ((0, "subj_a"), (1, "subj_b")):
    study_dir = studies_root / name
    study_dir.mkdir(parents=True, exist_ok=True)
    (study_dir / "config.json").write_text(json.dumps({"seed": seed}))

def loader(study_dir: Path):
    cfg = json.loads((study_dir / "config.json").read_text())
    examples = make_example_suite(noise_level=0.02, seed=cfg["seed"])[:1]
    study_globals = {"field_strength_mhz": Quantity(400.0, "MHz")}
    return Study.from_examples(examples, study_globals=study_globals), {"config": cfg}

runs = run_batch_studies(
    studies_root=studies_root,
    study_loader=loader,
    output_root="batch_output",
    analysis_kwargs={"num_steps": 10, "step_size": 0.02},
)

for run in runs:
    data = load_study_results(run["manifest_path"])
    print(run["study_name"], "->", data["manifest"]["targets"])

Selective execution

• Pass selected=["subject_001", "subject_005"] to limit processing to specific study folders.

• When using target_names in analysis_kwargs, those targets are forwarded to mrax.data_struct.storage.save_study_results() so only the requested experiments are exported.

Custom analysis hooks

Provide analysis_fn to run a custom pipeline per study. The function receives the Study plus study_dir and metadata (returned by the loader):

def compute_metrics(study, study_dir=None, metadata=None, **kwargs):
    from mrax.core.study_runner import execute_study_plan, plan_study_fits

    plan = plan_study_fits(study, target_names=kwargs.pop("target_names", None))
    results, _ = execute_study_plan(plan, run_fit_kwargs=kwargs)
    return results  # structure consumed by save_study_results

run_batch_studies(
    studies_root=studies_root,
    study_loader=loader,
    output_root="batch_output",
    analysis_fn=compute_metrics,
    analysis_kwargs={"num_steps": 5, "step_size": 0.02},
    save_kwargs={"file_extension": ".mrd"},
)

Outputs land in batch_output/<study_name>/ with study_manifest.json (including structured summary) plus per-experiment ISMRMRD data, mirroring the single-study export.