Experimental research software · Synthetic data · Not for clinical use · Not scientifically validated

Source: docs/applied_la_mitral_lv.md @ e3d560e · Source SHA-256 6ddc08648bfe163b69f86a2d7a2c453278b67a95be05f7165950ca71042cd915 · Markdown source

This document is reproduced verbatim. Only sensitive information (server host names, OS user names, file paths, account and billing details, AI usage, internal request IDs, and private repository names) was redacted in 1 places. No other content was changed.

LA–mitral–LV descriptive sidecar v1

This is a thin metadata sidecar for four_chamber_m1@1.0.0, restricted to LA, mitral transport and LV. It reads the closed M1 manifest and leaves equations, parameters, units, outputs and the legacy registry untouched. It does not create a new physiological model or an anatomical ontology equivalence. All source roles are generic knowledge templates. No patient, CT, mask, overlay, measurement or observation material is received, including public or deidentified material.

The baseline snapshot 456abc8c77134cd8909d603a1d96ed1384bad41f is UNREVIEWED. Inherited M1 batch-TDD history remains unresolved. This sidecar does not approve it. There is no clinical, anatomical or physiological validation, body-wide application, right-heart change or M2/M3 change.

Pinned evidence and internal-use notice

Source: [private repository] revision 70784bcdb9c14fba5271ee73f3c62b5f94ccbcfb. The retained knowledge-only snapshot was read locally; no source scripts, instructions, inference environment, network or full TriG were used. Source material and prior model results remain untrusted reference evidence. applied_la_mitral_lv_source_v1.json retains original JSON records in record, with original file, JSON pointer and full-file SHA256 in provenance. Only selected literature-source metadata is retained.

The Python module independently pins the complete canonical JSON document, section digests and original full-file digests. Rewriting both the receipt and its records cannot change these code pins. Canonical JSON is UTF-8, sorted keys, ASCII escaping, compact separators and no nonfinite numbers. Hashes describe integrity and provenance, not scientific validity or numerical permission. The complete-document digest includes the subset-digest receipt but never itself. The loader requires the pinned revision, unique template/profile keys and claim IDs, closed context references, and knowledge-only flags. The two null contexts are explicitly heart/organ and heart/mitral_orifice. No missing context is filled by inference. All four chamber/wall/myocardium claims questioned in the plan exist in this slice; absent_claims is empty.

Source THIRD_PARTY_NOTICES.md:3–14 reports no separate own-code open-source license. This copy is owner-authorized internal research use only; no source relicensing or redistribution right is asserted. Uberon is CC BY 3.0, FMA CC BY 4.0 and RO CC0 as recorded by the source notice. Other referenced works retain separate access/license conditions in sources. Source scripts are not copied. Term release versions are not invented; the source revision is the version pin.

Exact source locations and roles

Paths below are relative to experiments/medmlx_anatomy_mapping/ in that revision. Catalog pointers are array positions in results/contents_catalog.json.

Every catalog record retains KNOWLEDGE_TEMPLATE_REVIEW and observed_in_CT=false. The identical blood concept ID does not merge the LA/LV contextual templates.

Directed claims are verbatim in results/relation_literature_review.json at #/claims/20,22,23,26,27,32,33,34,39,41,42,45 (the machine receipt is authoritative for each pointer). Each retains its claim ID, exact IRI, direction, evidence locator, reviewer, scope, original relation status, formalization decision and expert-review status. No absent field is invented: the derived status view marks an absent anatomical-evidence status NOT_RECORDED while retaining the verbatim record. All selected claims are NOT_INDEPENDENTLY_EXPERT_REVIEWED.

Special reviews preserve mitral_orifice_boundary_equivalence=REVIEW and mitral_annulus_complete_fibrous_ring_interpretation=REVISE. Boundary definitions are ontology/boundary_definitions.json#/proposals/5 (attachment) and #/proposals/6 (transfer interface), version 0.1.0, DRAFT_MODEL_CONVENTION, NOT_MEASURED. The symbolic oriented spanning interface is separate from the anatomical valve, orifice, annulus, open gap and effective orifice area. EOA is NOT_REPRESENTED.

Relations and independent gates

Source blood→cavity RO_0001025 (located_in), source cavity→chamber RO_0002572 (IRI preserved without a new interpretation), target C5 part_of (docs/decisions.md:27, runtime adaptation deferred), and M1 blood_contained_in (four_chamber_m1_manifest.py:54–61) remain separate. The C5 entry is explicitly unapplied. No sameAs, equivalentClass, verified bearer or value equivalence is produced. V2 anatomy pointers are recorded beside source IDs only; matching spelling does not establish equivalence. V2 LA's Smith role and separate M1 LA state remain an unresolved interpretation difference.

Three independent axes are retained: source decisions/original relation status/ expert reviews; target crosswalk PROPOSED; external numeric permission NONE. ACCEPT only supports descriptive context. REVIEW, REVISE and DRAFT remain visible. They never imply owner review or numerical authorization.

• G1: pinned representation and exact model contract checks.
• G2: synthetic execution allowed; public gate_report reports executed=false until the demo actually executes and verifies its two runs.
• G3: external numeric binding CLOSED. request_numeric_binding always raises PermissionError, including requests carrying ACCEPT or a purported approval.

The validator is deliberately a closed model-specific contract. It rejects extra or changed fields, model versions, contexts, units, references, temporal meaning, direction, identity/bearer substitutions, aorta→systemic-bed mappings, boundary/EOA equivalences, status promotion and changed receipts. No actual binder or connect_value use is added; no resistance, area or elastance is inferred.

Symbols, parameters and law references

variables and parameters are exact copies of their selected M1 contracts; sidecar role annotations live separately and never extend the canonical manifest.

Pressure reference is exactly: "model gauge pressure relative to one common fixed external pressure of 0 mmHg; no intrathoracic/pericardial pressure". Signed negative gauge pressure is permitted. Nonpressure references remain not applicable.

Required parameter references: elastance_la, emin_lv, emax_lv (mmHg/mL), resistance_mitral (mmHg*s/mL), heart_rate (1/min), systolic_fraction (dimensionless), unstressed_volumes and initial_volumes (mL). The volume arrays retain the complete M1 order in their exact parameter contracts; slice references select LA index 7 and LV index 0. Initial volumes are initial-state inputs; V0 references select unstressed-volume parameters. No values are obtained from anatomy. SV_lv/CO_lv belong to the aortic edge and are intentionally rejected here.

E_lv(t) is a separate constitutive-law node, never a manifest variable. Its activation inputs are heart_rate and systolic_fraction; the implementation refers to physiosim.models.cardiovascular.elastance. Verbatim target equation anchors and declared line/section references remain traceable to docs/physiology_four_chamber_m1.md:62–70: activation, E_v(t), P_v, P_a, ideal-diode q_e and cumulative integral. These are references to existing model equations, not adopted anatomical axioms. Constitutive, definition and signed conservation-contribution dependencies are separate from value equivalence; the slice does not replace the other M1 edges.

M1's ideal diode is a restricted no-reverse-flow approximation relative to the source interface allowing signed/regurgitant transport. It does not model regurgitation, effective area, patient geometry or valve mechanics.

Public API and private demo

From physiosim.ontology.applied_la_mitral_lv: load_source_slice(path=None), validate_source_slice(document), build_applied_sidecar(), validate_applied_sidecar(sidecar), gate_report(sidecar), request_numeric_binding(*args, **kwargs). Validation returns None on success and raises ValueError for contract defects. Missing local resources fail locally; no network fallback exists. Expected source hashes are code pins. The M1-specific law/version contract checks the model ID, version, common pressure reference, LA/LV state order, mitral LA → LV diode and required parameter fields. This declaration does not prove arbitrary numerical implementation integrity. Runtime file hashes are separate observed provenance; checkout tests check the law anchors against documentation. The existing M1 manifest remains closed.

From physiosim.experiments.applied_la_mitral_lv_demo: result_fingerprint(result), run_applied_demo(output), main(argv=None). The demo runs the existing synthetic_fixture() with cycles=2, samples_per_cycle=50, and SolverSettings(max_step=period/200) twice, before and after building/validating the sidecar. Full time/volume/pressure/flow/cumulative arrays, dtype/shape and configuration hashes must match exactly. The manifest must also remain identical. These short runs are numerical invariance checks, not steady-state or physiological validation.

PYTHONDONTWRITEBYTECODE=1 PYTHONPATH="$PWD/src" \
.venv/bin/python \
scripts/run_applied_la_mitral_lv.py --output results/applied-demo

--output is required. Existing nonempty directories and final-component symlinks are rejected; symlinked ancestors (such as macOS /tmp) are accepted. umask 077 is restored afterward; the output directory is mode 0700 and all six files are mode 0600 with exclusive, no-follow writes. Files: applied_sidecar.json, gate_report.json, source_receipt.json, trace_la_mitral_lv.csv, index.html, run_receipt.json. CSV values come from actual computed M1 arrays, with units in column names. The local HTML uses inline SVG, synthetic labels and source statuses; no external assets or scripts. Output checksums explicitly exclude run_receipt.json itself. No output paths or timestamps enter deterministic file contents.

Execution evidence and review boundary

For the PR #5 portability repair, chronological focused commands, actual exit codes and RED/GREEN logs are retained in the private task artifacts under pr5-wheel-portability-v1/tdd/index.jsonl. The initial wheel evidence reproduced two failures; it is not evidence for all 30 reported CI failures. The router runs both complete source and freshly installed wheel suites separately. These software checks do not approve mappings or physiological validity. Inherited M1 batch-TDD history remains UNREVIEWED.

Run focused source tests from this checkout with:

PYTHONDONTWRITEBYTECODE=1 \
PYTHONPATH="$PWD/src" \
.venv/bin/python -m pytest \
tests/test_applied_la_mitral_lv.py -q -p no:cacheprovider -o addopts= \
-o pythonpath="$PWD/src" --basetemp=.pytest_tmp/focused

test_physiosim_package_matches_checkout_source compares complete imported package file hashes against this checkout, accepting an identical installed wheel and rejecting changed submodules, missing files and extra files. Wheel tests use python -I -m pytest without source-path redirection. A byte-exact check compares the bundled runtime anatomy JSON with docs/evidence_review/anatomy_nodes_v2.json in both modes, so a later documentation-source change requires an explicit resource update. This preserves source provenance without requiring repository docs at runtime.

Law and port metadata explicitly retain unit, temporal meaning, sign, reference and aggregation. time_contract fixes seconds elapsed from run start and the periodic activation phase. parameter_roles distinguishes caller parameters from state; target_anatomy_source preserves the existing v2 version/provenance contract.