{
  "@context": "../../../../../spec/context/v0.5.jsonld",
  "id": "https://uofa.net/morrison/cou2",
  "type": "UnitOfAssurance",
  "conformsToProfile": "https://uofa.net/vocab#ProfileComplete",
  "name": "Morrison COU2 — VAD hemolysis credibility assessment",
  "description": "V&V 40 credibility assessment for CFD hemolysis prediction of the FDA generic centrifugal blood pump under ventricular assist device (VAD) conditions. COU2 evaluates whether the computational model can predict absolute hemolysis levels for a Class III VAD device at Model Risk Level 5. Higher model risk drives stricter credibility requirements; Monte Carlo uncertainty quantification is performed but all credibility factor gaps remain. Full 13-factor assessment: 7 factors assessed (all with achieved levels below required levels), 6 factors not assessed (flagged as epistemic gaps at MRL 5). Decision: Not accepted.",
  "bindsRequirement": "https://uofa.net/morrison/req/hemolysis-safety",
  "bindsClaim": {
    "id": "https://uofa.net/morrison/claim/cou2-hemolysis-adequacy",
    "type": "AssuranceClaim",
    "name": "COU2 hemolysis prediction adequacy claim",
    "description": "The CFD model can predict absolute hemolysis index values for a Class III VAD device across operating conditions with quantified uncertainty, sufficient to support device safety evaluation.",
    "wasDerivedFrom": [
      "https://uofa.net/morrison/validation/hemolysis-comparison-cou2",
      "https://uofa.net/morrison/validation/piv-velocity-comparison",
      "https://uofa.net/morrison/validation/mesh-convergence",
      "https://uofa.net/morrison/uq/monte-carlo-hemolysis-cou2"
    ],
    "acceptanceCriteria": {
      "id": "https://uofa.net/morrison/criteria/cou2-acceptance",
      "type": "AcceptanceCriteria",
      "name": "COU2 VAD acceptance criteria",
      "description": "All credibility factors must meet or exceed required levels for Model Risk Level 5. Hemolysis predictions must include Monte Carlo uncertainty quantification with 95% confidence intervals. Predicted hemolysis index must fall below absolute clinical threshold (MIH < 0.1 g/100L) across all five flow-speed operating conditions."
    }
  },
  "bindsModel": "https://uofa.net/morrison/model/ansys-cfx-15-sst-kw",
  "bindsDataset": [
    "https://uofa.net/morrison/data/piv-velocity",
    "https://uofa.net/morrison/data/hemolysis-invitro"
  ],
  "hasContextOfUse": {
    "id": "https://uofa.net/morrison/cou/cou2-vad",
    "type": "ContextOfUse",
    "name": "COU2: Ventricular assist device use (Class III)",
    "description": "A CFD model is used to predict absolute hemolysis levels across operating conditions for a centrifugal blood pump used as a ventricular assist device (VAD). Model influence is HIGH (model is primary basis for hemolysis acceptance); decision consequence is HIGH (incorrect prediction could result in chronic patient hemolysis requiring clinical intervention). Model Risk Level 5 — the highest in the Morrison assessment.",
    "intendedUse": "Use CFD to predict absolute hemolysis index values for a Class III VAD device across 5 flow-speed combinations. Results must demonstrate hemolysis below absolute clinical threshold (not relative to predicate). Requires Monte Carlo uncertainty quantification on hemolysis predictions.",
    "deviceClass": "Class III",
    "modelInfluence": "High",
    "decisionConsequence": "High"
  },
  "hasValidationResult": [
    {
      "id": "https://uofa.net/morrison/validation/mesh-convergence",
      "type": "ValidationResult",
      "name": "Mesh convergence study",
      "description": "Grid convergence index (GCI) computed for three mesh densities (coarse, medium, fine) at 3500 RPM / 6 LPM. Fine mesh selected for all subsequent simulations. GCI < 2% for velocity field; GCI < 5% for hemolysis index.",
      "outcome": "pass",
      "wasDerivedFrom": "https://uofa.net/morrison/comparator/mesh-self-convergence",
      "wasGeneratedBy": {
        "id": "https://uofa.net/morrison/activity/mesh-convergence-study",
        "type": "VerificationActivity",
        "name": "Mesh convergence verification",
        "description": "Systematic grid refinement study using three mesh densities with GCI evaluation per Roache (1998)."
      },
      "comparedAgainst": {
        "id": "https://uofa.net/morrison/comparator/mesh-self-convergence",
        "type": "Comparator",
        "name": "Self-convergence (grid refinement)",
        "description": "Richardson extrapolation applied to three mesh levels to estimate discretization error."
      },
      "hasUncertaintyQuantification": {
        "id": "https://uofa.net/morrison/uq/gci-mesh",
        "type": "UncertaintyQuantification",
        "name": "Grid convergence index uncertainty",
        "description": "Discretization uncertainty estimated via grid convergence index. GCI < 2% velocity, GCI < 5% hemolysis."
      }
    },
    {
      "id": "https://uofa.net/morrison/validation/piv-velocity-comparison",
      "type": "ValidationResult",
      "name": "PIV velocity field comparison",
      "description": "CFD-predicted velocity fields compared against particle image velocimetry (PIV) measurements at 5 flow-speed operating conditions. Qualitative agreement in flow structures; quantitative comparison of velocity magnitude profiles at selected planes.",
      "outcome": "pass",
      "wasDerivedFrom": "https://uofa.net/morrison/comparator/piv-experimental",
      "wasGeneratedBy": {
        "id": "https://uofa.net/morrison/activity/piv-comparison",
        "type": "VerificationActivity",
        "name": "PIV velocity validation",
        "description": "Comparison of CFD velocity predictions against experimental PIV measurements at matched operating conditions."
      },
      "comparedAgainst": {
        "id": "https://uofa.net/morrison/comparator/piv-experimental",
        "type": "Comparator",
        "name": "PIV experimental velocity data",
        "description": "Particle image velocimetry measurements of velocity field in the FDA generic blood pump at 5 operating conditions. Data collected by FDA OSEL."
      },
      "hasUncertaintyQuantification": {
        "id": "https://uofa.net/morrison/uq/piv-velocity-uncertainty",
        "type": "UncertaintyQuantification",
        "name": "PIV comparison uncertainty",
        "description": "PIV measurement uncertainty propagated through comparison metric. CFD prediction uncertainty from Monte Carlo input perturbation included in comparison bounds."
      }
    },
    {
      "id": "https://uofa.net/morrison/validation/hemolysis-comparison-cou2",
      "type": "ValidationResult",
      "name": "Hemolysis prediction comparison (COU2 — absolute threshold)",
      "description": "CFD-predicted modified index of hemolysis (MIH) compared against in vitro hemolysis measurements at 5 flow-speed operating conditions. COU2 requires absolute threshold comparison (MIH < 0.1 g/100L) with Monte Carlo uncertainty quantification, unlike COU1 which uses relative predicate comparison.",
      "outcome": "pass",
      "wasDerivedFrom": "https://uofa.net/morrison/comparator/hemolysis-invitro",
      "wasGeneratedBy": {
        "id": "https://uofa.net/morrison/activity/hemolysis-comparison-cou2",
        "type": "VerificationActivity",
        "name": "COU2 hemolysis validation with UQ",
        "description": "Monte Carlo uncertainty quantification on hemolysis predictions. 1000 samples with perturbed inputs (blood properties, boundary conditions). 95% confidence intervals computed for MIH at each operating condition."
      },
      "comparedAgainst": {
        "id": "https://uofa.net/morrison/comparator/hemolysis-invitro",
        "type": "Comparator",
        "name": "In vitro hemolysis measurements",
        "description": "Modified index of hemolysis (MIH) measured from bovine blood experiments in the FDA generic blood pump at 5 operating conditions. Data collected by FDA OSEL."
      },
      "hasUncertaintyQuantification": {
        "id": "https://uofa.net/morrison/uq/monte-carlo-hemolysis-cou2",
        "type": "UncertaintyQuantification",
        "name": "Monte Carlo hemolysis UQ",
        "description": "1000-sample Monte Carlo propagation of input uncertainties (blood viscosity, inlet boundary conditions, hemolysis model constants) through CFD pipeline. 95% confidence intervals on predicted MIH at each of 5 operating conditions. This is the key structural difference from COU1, which has no UQ on hemolysis predictions."
      }
    }
  ],
  "wasDerivedFrom": "https://doi.org/10.1115/1.4043024",
  "wasAttributedTo": "https://uofa.net/org/FDA-CDRH",
  "generatedAtTime": "2019-02-01T00:00:00Z",
  "hash": "sha256:245b9cb4adec4df51f7a7387779d28c3cfbbe4874e40d28b2a76a81aae946881",
  "signature": "ed25519:c0c38e0a660d917cd1b68577c46e7f7392776457e51a2171bd56cb93f0a83901affadbacac05655bb02368d3e1696ad8b87277b4776a657ab9194c60cb4cca04",
  "signatureAlg": "ed25519",
  "canonicalizationAlg": "json-sortkeys/v1",
  "hasCredibilityFactor": [
    {
      "type": "CredibilityFactor",
      "factorType": "Software quality assurance",
      "requiredLevel": 3,
      "achievedLevel": 1,
      "rationale": "ANSYS CFX is a commercial solver with internal SQA processes, but documented evidence of SQA practices was not available to the assessment team at the level required for MRL 5. Achieved Level 1 (commercial solver assumed) vs Required Level 3 (independent SQA audit or third-party certification evidence).",
      "acceptanceCriteria": "Software must meet minimum SQA requirements per V&V 40 Table 3 for MRL 5: documented version control, regression testing, user manual, and independent SQA audit trail.",
      "factorStatus": "assessed",
      "factorStandard": "ASME-VV40-2018"
    },
    {
      "type": "CredibilityFactor",
      "factorType": "Numerical code verification",
      "requiredLevel": 4,
      "achievedLevel": 3,
      "rationale": "Mesh convergence study performed with GCI. No method of manufactured solutions (MMS) or independent benchmark verification against analytical solutions. Achieved Level 3 (convergence study) vs Required Level 4 (formal code verification with MMS).",
      "acceptanceCriteria": "MRL 5 requires formal numerical code verification including MMS benchmarks, systematic convergence studies, and documented numerical error bounds for all primary quantities of interest.",
      "factorStatus": "assessed",
      "factorStandard": "ASME-VV40-2018"
    },
    {
      "type": "CredibilityFactor",
      "factorType": "Discretization error",
      "requiredLevel": 4,
      "achievedLevel": 3,
      "rationale": "GCI computed for velocity and hemolysis on three mesh levels. Asymptotic convergence range not formally demonstrated for hemolysis metric. Achieved Level 3 (GCI computed) vs Required Level 4 (formal convergence demonstration).",
      "acceptanceCriteria": "MRL 5 requires quantified discretization error bounds with demonstrated asymptotic convergence and formal GCI for all quantities of interest including hemolysis.",
      "factorStatus": "assessed",
      "factorStandard": "ASME-VV40-2018"
    },
    {
      "type": "CredibilityFactor",
      "factorType": "Test conditions",
      "requiredLevel": 4,
      "achievedLevel": 2,
      "rationale": "Five flow-speed operating conditions tested, but operating envelope boundaries not systematically explored. Boundary conditions documented but experimental uncertainty on inlet/outlet conditions not fully characterized. Achieved Level 2 (representative conditions) vs Required Level 4 (systematic envelope coverage with quantified boundary condition uncertainties).",
      "acceptanceCriteria": "MRL 5 requires test conditions that span the full intended operating envelope with documented uncertainty on all controlled parameters and systematic boundary exploration.",
      "factorStatus": "assessed",
      "factorStandard": "ASME-VV40-2018"
    },
    {
      "type": "CredibilityFactor",
      "factorType": "Output comparison",
      "requiredLevel": 4,
      "achievedLevel": 2,
      "rationale": "Hemolysis predictions compared quantitatively against in vitro data at matched operating conditions. However, formal validation metrics with prediction intervals and systematic comparison methodology required at Level 4 were not applied. Achieved Level 2 (quantitative comparison performed) vs Required Level 4 (formal validation metrics with prediction intervals).",
      "acceptanceCriteria": "MRL 5 requires quantitative output comparison with uncertainty bounds on both computational and experimental results, using formal validation metrics and prediction interval methodology.",
      "factorStatus": "assessed",
      "factorStandard": "ASME-VV40-2018"
    },
    {
      "type": "CredibilityFactor",
      "factorType": "Relevance of the quantities of interest",
      "requiredLevel": 3,
      "achievedLevel": 2,
      "rationale": "Modified index of hemolysis (MIH) is a recognized surrogate for clinical hemolysis but the mapping from MIH to clinical hemolysis outcome is not fully validated. Achieved Level 2 (recognized surrogate) vs Required Level 3 (validated clinical relevance).",
      "acceptanceCriteria": "MRL 5 requires that the quantities of interest be directly linked to clinical endpoints with documented relevance justification.",
      "factorStatus": "assessed",
      "factorStandard": "ASME-VV40-2018"
    },
    {
      "type": "CredibilityFactor",
      "factorType": "Relevance of the validation activities to the COU",
      "requiredLevel": 3,
      "achievedLevel": 2,
      "rationale": "Validation performed on the same FDA generic pump geometry under operating conditions that match the VAD COU. However, bovine blood analog introduces species-specific hemolysis pathway uncertainty that reduces relevance for human clinical translation. Achieved Level 2 (same geometry, accepted analog) vs Required Level 3 (validated clinical relevance with human-equivalent hemolysis data).",
      "acceptanceCriteria": "MRL 5 requires that validation activities use the same pump geometry, operating conditions, and blood analog as the intended COU application, with demonstrated relevance to human hemolysis outcomes.",
      "factorStatus": "assessed",
      "factorStandard": "ASME-VV40-2018"
    },
    {
      "id": "https://uofa.net/morrison/cou2/factor/numerical-solver-error",
      "type": "CredibilityFactor",
      "factorType": "Numerical solver error",
      "factorStatus": "not-assessed",
      "factorStandard": "ASME-VV40-2018"
    },
    {
      "id": "https://uofa.net/morrison/cou2/factor/use-error",
      "type": "CredibilityFactor",
      "factorType": "Use error",
      "factorStatus": "not-assessed",
      "factorStandard": "ASME-VV40-2018"
    },
    {
      "id": "https://uofa.net/morrison/cou2/factor/model-form",
      "type": "CredibilityFactor",
      "factorType": "Model form",
      "factorStatus": "not-assessed",
      "factorStandard": "ASME-VV40-2018"
    },
    {
      "id": "https://uofa.net/morrison/cou2/factor/model-inputs",
      "type": "CredibilityFactor",
      "factorType": "Model inputs",
      "factorStatus": "not-assessed",
      "factorStandard": "ASME-VV40-2018"
    },
    {
      "id": "https://uofa.net/morrison/cou2/factor/test-samples",
      "type": "CredibilityFactor",
      "factorType": "Test samples",
      "factorStatus": "not-assessed",
      "factorStandard": "ASME-VV40-2018"
    },
    {
      "id": "https://uofa.net/morrison/cou2/factor/equivalency-of-input-parameters",
      "type": "CredibilityFactor",
      "factorType": "Equivalency of input parameters",
      "factorStatus": "not-assessed",
      "factorStandard": "ASME-VV40-2018"
    }
  ],
  "hasWeakener": [],
  "hasDecisionRecord": {
    "id": "https://uofa.net/morrison/decision/cou2-overall",
    "type": "DecisionRecord",
    "actor": "https://uofa.net/org/FDA-CDRH",
    "role": "Credibility assessment team",
    "outcome": "Not accepted",
    "rationale": "Model not deemed sufficiently credible for COU2. All seven assessed credibility factors have achieved levels below the required levels for Model Risk Level 5. Key gaps: software quality assurance (achieved 1 vs required 3), numerical code verification (achieved 3 vs required 4), discretization error (achieved 3 vs required 4), test conditions (achieved 2 vs required 4), output comparison (achieved 2 vs required 4), relevance of quantities of interest (achieved 2 vs required 3), and relevance of validation activities to the COU (achieved 2 vs required 3). Additionally, 6 of 13 credibility factors were not assessed, representing epistemic gaps at MRL 5. The evidence shows the model does not meet the credibility bar for a Class III VAD application. Additional verification and validation activities required before the model can support Class III VAD hemolysis claims.",
    "decidedAt": "2019-02-01T00:00:00Z"
  },
  "assuranceLevel": "Low",
  "criteriaSet": "https://uofa.net/criteria/ASME-VV40-2018",
  "credibilityIndex": {
    "@value": "0.60",
    "@type": "xsd:decimal"
  },
  "traceCompleteness": {
    "@value": "1.00",
    "@type": "xsd:decimal"
  },
  "verificationCoverage": {
    "@value": "0.60",
    "@type": "xsd:decimal"
  },
  "validationCoverage": {
    "@value": "0.75",
    "@type": "xsd:decimal"
  },
  "uncertaintyCIWidth": {
    "@value": "0.15",
    "@type": "xsd:decimal"
  },
  "modelRiskLevel": 5,
  "deviceClass": "Class III",
  "couName": "VAD use",
  "decision": "Not accepted",
  "hasUncertaintyQuantification": true
}