Home

MAR ILS Dataset Generator — Project Home

Welcome to the official wiki for the Metal Artifact Reduction (MAR) Interlaboratory Study (ILS) reference implementation, compliant with ASTM WKXXXXX Revision 05 (soon ASTM FXXXX), IEC 60601-2-44 Ed. 4 §203.6.7.101, and DICOM 2026b CP-2575.

This project provides the reference implementation for generating and analyzing standardized synthetic CT datasets used to quantify the diagnostic-task impact of MAR algorithms via a Channelized Hotelling Observer (CHO).

---

Current Normative Reference

Attribute	Value
Standard of Truth	v7.0.0-dicom2026b (ballot candidate v7.1.0-dicom2026b)
Geometry	Fan-beam, SID = 570 mm, SDD = 1040 mm
Projections	720 angles × 512 equi-angular detectors, full 360°
Realizations	40 LP + 40 LA (screening: 20 + 20)
Lesion	Circular disc, 5-voxel radius, slice 128 only
CHO	2D single-slice, 121×121 ROI — 3D integration shall not be performed (§A1.5.3)
Lesion contrast	12.0 HU (sinogram-domain only, no post-FBP hard-set)
AUC tolerance	±0.005

---

Locked Metrology Baseline (2026-04-07)

Parameter	Value
BASELINE_AUC_noMAR	0.8294, CI [0.7612, 0.9025]
Internal noise σ (normative default)	15
Calibrated I₀	310,853
Background HU	40.0
Metal HU	3000.0
Noise σ (soft tissue)	30 HU target

All 79 metrology guards in tests/ pass green against these constants. The single source of truth lives in mar_ils_core/constants.py.

---

Technical Documentation

• Technical Foundation — SKE/BKE detection task, 2D CHO implementation, Laguerre-Gauss channelization.
• Phantom Design — Fan-beam geometry, elliptical body, centered iron rod, single-slice lesion disc.
• Standard Operating Procedure — Checksum verification, submission protocol.
• Compliance Checklist — Pre-submission verification.
• MAR ILS Results Submission Template — Reporting form for labs.

---

Regulatory Framework (Layered Approach)

Layer	Document	Role
1	IEC 60601-2-44 Ed. 4 §203.6.7.101	MAR method available (.1), user info (.2), DICOM recording (.3)
2	ASTM FXXXX (formerly WKXXXXX Rev 05) — bound to §203.6.7.101.1 via post-publication Corrigendum/Amendment, submitted after ASTM FXXXX publishes	TYPE TEST — measures ΔAUC
3	FDA guidance	Acceptance thresholds (non-degradation, superiority)

See docs_and_references/IEC_203_6_7_101_compliance_statement_proposal.md in the main repo for the PRVC comment text.

---

Strategic Rationale

1. Accessibility — fully digital workflow; no physical phantom shipping.
2. Determinism — 64-bit floating-point CHO replaces subjective artifact-width measurements.
3. Regulatory readiness — fan-beam geometry matches real clinical scanners; DICOM 2026b CP-2575 metadata exposed for automated compliance tooling.

---

For Task Group participation, open a GitHub Issue or contact the Technical Contact specified in the Work Item.