ONLINE RESOURCE 1 INDEX

Manuscript: Controlled Observation Fibres and Robust Causal Prefix Stopping
            in Sequential MRI
Journal:    Journal of Mathematical Imaging and Vision
Author:     Takashi Matsuhisa
Affiliation: MRI BSBH Foundation for Scientific Research, Mito, Japan;
             IAMR, KarRC RAS, Petrozavodsk, Russia
E-mail:     takashimatsuhisa.mri.bsbh@gmail.com;
            takashi.matsuhisa@krc.karelia.ru
Archive:    ESM_1.zip (Online Resource 1)
Metadata as supplied; author confirmation is required before submission.

Online Resource 1 contains no patient or scanner data.

1. code/run_prefix_phantom.py
   Reproduces the finite multicoil DCE-MRI phantom, first-singleton prefix
   stopping, nonlinear SPGR stress, exact-collision control, conservative
   observable-signature audit, bounded MRI contamination control, robust scalar
   control, paired summaries, figures, and numerical manifest.

2. code/run_retrospective_prefix.py
   Accepts authorised NPZ or DICOM image series for chronological prefix
   replay. Its JSON output includes a subject-level gate, p-value, prediction-
   set, stop, decision, and failure/status audit trail. A stopped test subject
   is removed from all later feature computations. DICOM input requires a
   predeclared temporal start and phase count. Real-data identifiers must be
   handled as sensitive data. The distributed result is a synthetic self-test
   only; no real DICOM fixture has been validated in this package.

3. code/audit_package.py
   In full-package mode, checks the outer checksum, class/style hashes, source
   structure, citations, labels, numerical manifest, and byte closure of both
   nested archives. With --supplement-only it checks the exact Online Resource
   file set, ESM_SHA256SUMS.txt closure, the manifest, and optionally both
   numerical self-tests.

4. code/requirements.txt, code/requirements-numerical-lock.txt
   Supported Python dependency ranges and the exact NumPy/Matplotlib versions
   used for the distributed synthetic outputs. pydicom is needed only for
   authorised conventional single-frame DICOM input.

5. data/results.npz, data/runs.csv, data/summary.csv,
   data/hypothesis_summary.csv, data/paired_differences.csv
   Synthetic phantom arrays, balanced run-level results, hypothesis summaries,
   and paired policy contrasts with uncertainty intervals.

6. data/results_macros.tex
   Generated numerical macros used by the manuscript.

7. data/test_report.json, data/robust_control.json,
   data/bounded_mri_control.csv, data/bounded_mri_control_summary.csv,
   data/outer_fibre_audit.json, data/statistical_methods.json,
   data/retrospective_prefix_selftest.json
   Numerical unit checks, scalar and MRI robustness controls, the exact-rational
   outer-signature audit, declared interval methods, and the stored-image prefix
   software self-test. The latter contains synthetic subject-level audit records
   and post-stop future-access guards.

8. data/manifest.json
   Configuration, runtime environment, scope statements, and hashes for
   generated outputs.

9. DATA_DICTIONARY.md
   Defines every distributed CSV field family, JSON object, NPZ array, missing-
   value convention, and the distinct switching quantities used for policy
   scoring and descriptive summaries.

10. ESM_SHA256SUMS.txt
   SHA-256 records for every regular Online Resource file except this checksum
   file itself.

11. figures/fig1_pipeline.pdf through figures/fig5_failures.pdf
   Vector figures generated by run_prefix_phantom.py.

Correct reruns reproduce the numerical arrays, tables, JSON results, and
figures. The runtime macro in results_macros.tex and the creation time,
runtime, and platform provenance in manifest.json are run dependent.

The program output is synthetic and does not estimate clinical accuracy,
counterfactual acquisition effects, or state-of-the-art raw-k-space
performance.
