Reliability Report  •  Sample
Reliability Report
Two independent instruments. One acquisition event.
NOD-001  ·  RUL
Solid pulmonary nodule  —  7.8 mm
2026-06-18  ·  GE Revolution CT  ·  CT Chest LDCT
Image Reliability
Acquisition-induced measurement variability
Low
±1.2mm
Expected SD (k=2)  ·  range 6.6–9.0 mm
Contributing factors
!
Soft reconstruction kernel
!
Near resolution limit
~
Complex boundary geometry
AI Robustness
Protocol estimate  ·  Reference model
Low
This model showed large output changes under clinically realistic protocol variations. This acquisition profile falls into a protocol region where the model demonstrated lower robustness during validation.
Reference robustness profile  ·  Tested across reconstruction kernel, slice thickness, dose, and vendor variation.
Clinical impact
⚠
Lung-RADS threshold risk Diameter range (6.6–9.0 mm) spans the Lung-RADS 2→3 boundary. Category assignment unreliable without protocol standardization.
Recommendation
Confirm diameter on equivalent reconstruction before assessing interval growth or upgrading Lung-RADS category.
Technical details
ARS score 15.4%  (95% CI 11.2–19.6%) Sphericity 0.412 S/V ratio 0.81 mm⁻¹ Boundary ΔHU 184.3 HU  (mean |Δ|, 3-voxel shell) Kernel / Slice / IRB60f  ·  5.0 mm  ·  ASIR-V 50% CTDIvol 2.1 mGy Series UID 1.2.840.114350…4609 Processed 2026-06-18T14:32:07Z  ·  Engine v2.3.1  ·  RPT-20260618-00274
NOD-002  ·  LLL
Solid pulmonary nodule  —  18.3 mm
2026-06-18  ·  Siemens SOMATOM Force  ·  CT Chest LDCT
Image Reliability
Acquisition-induced measurement variability
High
±0.4mm
Expected SD (k=2)  ·  range 17.9–18.7 mm
Contributing factors
✓
Standard reconstruction kernel
✓
Well above resolution limit
✓
Compact, well-defined boundary
AI Robustness
Protocol estimate  ·  Reference model
High
This model showed stable output across clinically realistic protocol variations. This acquisition profile falls into a protocol region where the model demonstrated higher robustness during validation.
Reference robustness profile  ·  Tested across reconstruction kernel, slice thickness, dose, and vendor variation.
Clinical impact
✓
Low threshold risk Measurement variability unlikely to affect Lung-RADS category or AI output on repeat or cross-site acquisition.
Recommendation
No protocol adjustment required. Standard interval comparison applies.
Technical details
ARS score 1.8%  (95% CI 0.4–3.2%) Sphericity 0.738 S/V ratio 0.22 mm⁻¹ Boundary ΔHU 41.2 HU  (mean |Δ|, 3-voxel shell) Kernel / Slice / IRB30f  ·  3.0 mm  ·  ADMIRE 3 CTDIvol 3.4 mGy Series UID 1.2.276.0.7230010…5162 Processed 2026-06-18T14:32:09Z  ·  Engine v2.3.1  ·  RPT-20260618-00275