The reviewed record of science sign in
Pith

arxiv: 2607.06210 · v1 · pith:C7PXVCPW · submitted 2026-07-07 · physics.med-ph · math.DS· q-bio.OT

Validation of a Computational Respiratory System Model for Mechanical Ventilation

Carlotta Hennigs , Charlott Danielson , Franziska Bilda , Dimitrios Karachalios , Niklas Hackelberg , Helene Selpien , Georg M\"annel , Dirk Sch\"adler
show 2 more authors
Folker Spitzenberger Philipp Rostalski
This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-07-08 13:11 UTCglm-5.2pith:C7PXVCPWrecord.jsonopen to challenge →

classification physics.med-ph math.DSq-bio.OT PACS 87.85.-d87.19.Hh
keywords mechanical ventilationin silico clinical trialscredibility assessmentASME V&V 40FDA guidancepatient-device modelautomated weaningrespiratory mechanics
0
0 comments X

The pith

Ventilation simulation model validated for preclinical weaning trials

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper argues that a coupled patient-device computational model integrating respiratory mechanics, gas exchange, respiratory control, and ventilator dynamics is credible enough to serve as a stand-in for patients in preclinical in-silico trials of automated weaning protocols. The authors apply a nine-step risk-informed credibility assessment workflow derived from ASME V&V 40 and FDA guidance, tailoring it to mechanical ventilation, a domain where no broadly accepted regulator-aligned reference simulator currently exists. They define a specific context of use (preclinical evaluation of an automated weaning protocol based on SmartCare/PS), classify the model risk as medium-low, and then run nine validation tests spanning gas exchange, respiratory control, lung mechanics in COPD, patient-ventilator asynchrony, calibration against clinical data from three ventilated patients, model plausibility review, and emergent behavior reproduction. The model achieves an applicability score of 2.5 out of 3 (rated good), with strong calibration (R-squared approximately 95% for lung volume, MAPE approximately 15% for capnography) and plausibility (3.0 each), and solid but more limited population-based validation (2.0) and emergent behavior coverage (2.0). The authors conclude the model is fit for purpose for medium-low risk preclinical in-silico clinical trials of automated weaning strategies, while transparently documenting gaps including limited patient cohorts, incomplete code verification, and absence of external validation.

Core claim

The central object is the patient-device model (PDM), a deterministic, mechanistic ODE-based representation coupling lung mechanics, four-compartment gas exchange, a neural-oscillator-driven respiratory center, and an idealized ventilator model. The paper's core claim is that this PDM, when assessed through a standards-aligned credibility framework adapted to ventilation, meets acceptance criteria for preclinical use in automated weaning evaluation. The evidence rests on four pillars: calibration against clinical data from three ventilated patients showing strong waveform fit, population-based comparison against literature envelopes for gas exchange and respiratory control responses, expert-

What carries the argument

reviewed plausibility of governing equations, and reproduction of emergent phenomena (auto-PEEP, patient-ventilator asynchrony, biphasic hypoxic response, CO2 apnea threshold) that arise from submodel interactions rather than explicit programming. The applicability score Sapp=2.5 is computed as an unweighted mean of subscores across four FDA evidence categories.

If this is right

  • If the model's credibility holds after verification closure and expanded validation, it could replace or reduce animal testing in the preclinical development of automated ventilation controllers, lowering cost and accelerating translation of closed-loop ventilator algorithms.
  • The nine-step credibility workflow adapted here for ventilation could become a template for validating computational physiological models in other under-served medical device domains lacking a regulator-accepted reference simulator.
  • The documented gaps (limited cohorts, no external validation, incomplete verification) define a concrete research agenda: any group seeking to elevate this or a similar model to higher-risk contexts must close these gaps before regulatory submission.
  • The emergent behavior demonstrations (auto-PEEP, asynchrony emergence, CO2 apnea threshold) suggest the model captures enough coupled physiology to stress-test controller logic in regimes that single-subsystem models cannot reach.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The medium-low risk classification is self-assigned by the authors; a regulator might classify the same evidence bundle differently, particularly given the absence of systematic code verification, which ASME V&V 40 treats as a co-equal prerequisite to validation rather than an optional add-on.
  • The calibration on only three patients, with only two measurement signals (airway flow and capnogram), means the model's parameter identifiability is likely limited, and the strong fit may not generalize to broader populations or different ventilator modes.
  • The absence of cardiovascular coupling (constant heart rate and cardiac output assumed) creates a blind spot for weaning scenarios where hemodynamic instability interacts with ventilatory control, potentially limiting the model's utility for sicker patient populations.

Load-bearing premise

The validation results are conditional on the correctness of the software implementation, yet systematic code verification including structured code review, formal unit tests for all submodels, and convergence studies of the numerical solver has not been completed. Without verification, observed agreements between model and data could reflect implementation artifacts rather than genuine model adequacy.

What would settle it

If the completed verification campaign reveals numerical errors or implementation bugs in the ODE solver or submodel equations, the calibration fits and validation comparisons reported here could shift substantially, potentially dropping the applicability score below the threshold for even medium-low risk use.

Figures

Figures reproduced from arXiv: 2607.06210 by Carlotta Hennigs, Charlott Danielson, Dimitrios Karachalios, Dirk Sch\"adler, Folker Spitzenberger, Franziska Bilda, Georg M\"annel, Helene Selpien, Niklas Hackelberg, Philipp Rostalski.

Figure 1
Figure 1. Figure 1: High-level functional representation of the automated weaning protocol and detailed [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Credibility evidence categories focus￾ing in validation applicable for the defined use case and the derived validation plan are pre￾sented, with the numbering of categories based on [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Results of the uncertainty quantification and sensitivity analysis. [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Population based evidence: Simulation results of validation step 1 (a) and 2 (b) for the [PITH_FULL_IMAGE:figures/full_fig_p014_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Test V–7: Patient 3 simulation results (blue line) in comparison with clinical data [PITH_FULL_IMAGE:figures/full_fig_p015_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Test V–7: Patient 1 simulation results (blue line) in comparison with clinical data (dotted, [PITH_FULL_IMAGE:figures/full_fig_p031_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Test V–7: Patient 2 simulation results (blue line) in comparison with clinical data (dotted, [PITH_FULL_IMAGE:figures/full_fig_p031_7.png] view at source ↗
read the original abstract

Computational modeling and simulation are powerful tools for the assessment of medical device performance and safety, particularly for in silico clinical trials for automated medical systems. In ventilation, where managing gas exchange, respiratory mechanics, and patient-ventilator interaction is required under evolving pathophysiology, the clinical translation of automated control strategies remains slow and resource-intensive. This paper applies a standards-aligned framework for the credibility assessment of a computational respiratory model, demonstrated using an automated weaning case study. The framework operationalizes ASME V&V 40 and FDA principles within a structured, guidance-based validation workflow. The computational physiological model integrates respiratory mechanics, gas exchange, respiratory control, and a ventilator representation, validated under a clearly defined context of use and explicit questions of interest. Model credibility is assessed through calibration, physiological plausibility, population-based evaluation, and reproduction of emergent behavior. All model requirements derived from the intended context of use are addressed within the proposed credibility assessment plan, and documented gaps are transparently reported. The resulting credibility argument supports the applicability of the model for its context of use. Strengths are demonstrated in population-based comparison and mechanistic plausibility, while residual limitations relate to the extent of in vivo evidence, population representativeness, and external validation. Overall, the model is considered fit for purpose for medium-low risk preclinical in silico clinical trials of automated weaning strategies. Furthermore, the validation procedure outlined in this article provides a blueprint for the validation of this and similar models in other mechanical ventilation algorithms and related use cases.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 12 minor

Summary. This manuscript presents a credibility assessment of a computational patient-device model (PDM) for mechanical ventilation, framed around an automated weaning protocol (AWP) case study. The authors operationalize ASME V&V 40 and FDA guidance through a nine-step workflow, defining a context of use (COU), model risk (medium-low), and quantities of interest (VT, RR, etCO2, asynchrony, Pplat). Nine validation tests are proposed covering gas exchange, respiratory control, lung mechanics, calibration, plausibility, and emergent behavior. The authors conclude the model is fit for purpose for preclinical in silico clinical trials under the stated COU, with a composite applicability score Sapp = 2.5 (good). The framework is well-structured and the COU/risk framing is a genuine contribution to a domain where such operationalization remains limited.

Significance. The paper addresses a real gap: unlike the diabetes simulator domain (UVA/Padova), mechanical ventilation lacks a broadly accepted, regulator-aligned reference simulator. The explicit mapping from clinical questions of interest to credibility evidence categories, and thence to graded acceptance criteria, is a useful blueprint. The transparency about limitations (incomplete verification, small calibration cohort, partial observability) is commendable and aligns with the spirit of ASME V&V 40. The UQ/SA results (Figure 3) and the governing-equation grading (Table 11) are concrete artifacts that strengthen the plausibility argument. The calibration results (Table 3, R2 ~0.94-0.97) on clinical capnogram data are a positive signal, though limited in scope.

major comments (3)
  1. §3.2.6, Tests V-3 through V-6: Four of nine validation tests defer all quantitative results to self-cited companion papers [15, 16] without presenting any inline figures, tables, or summary statistics. V-4 states only 'For results see reference [15]' with no quantitative claim whatsoever. V-3, V-5, and V-6 state acceptance criteria (MAPE <= 10%, differ < 5%) but provide no data to substantiate them. These tests feed directly into the population-based validation subscore (2.0) and the emergent behavior subscore (2.0) in Table 6, which in turn support the composite Sapp = 2.5 claim. Without at least summary statistics or a figure per test, the reader cannot independently assess whether the stated criteria are met. This is load-bearing for the central fitness-for-purpose claim and should be addressed by including inline evidence (at minimum a table of MAPE/bias values and one representative
  2. §3.2.6, Test V-7 and §4.5.1: The calibration uses three patients with only two measurement signals (airway flow and capnogram). The paper itself acknowledges (§4.5.1) that 'the available measurement set (flow and end-tidal CO2) does not fully satisfy the observability conditions required for complete state reconstruction.' Despite this, the calibration subscore is rated 3.0 (good) in Table 6, contributing to the Sapp = 2.5. The gap between the acknowledged observability limitation and the 'good' rating should be reconciled. At minimum, the authors should clarify which specific credibility factors in Table 12 drive the 3.0 subscore and why the observability limitation does not reduce the rating below 'good.'
  3. §3.2.6, Verification: The paper explicitly states that 'a systematic and complete verification of the model implementation... has not yet been fully executed and documented.' The authors are transparent about this, and the Discussion (§4.1) acknowledges that 'verification activities are currently ongoing.' However, ASME V&V 40 treats verification as a prerequisite for credible validation. The paper's own framework (§3.1, step 7) lists verification as a required step. The current text states validation results are 'conditional on the correctness of the current implementation' but does not factor this conditionality into the Sapp score or the adequacy assessment. The authors should either (a) explicitly state how the incomplete verification status affects the credibility score's interpretation, or (b) add a conditional qualifier to the 'fit for purpose' conclusion in the abstract and §3.2.
minor comments (12)
  1. §2.2.3: 'dynamically account with a device-specific ventilator control model' — grammatical issue, likely 'dynamically coupled with.'
  2. §2.2.3: 'account ordinary differential equations (ODEs)' — likely 'a set of ordinary differential equations.'
  3. §3.2.6, Test V-5: 'For results see For results see reference [16]' — duplicated text.
  4. Table 3: Patient 2 MAPE for FCO2 is listed as '-' with no explanation. Please clarify whether MAPE could not be computed or was not reported.
  5. Table 3 caption: 'R2 FCO2 R2 V' column headers are ambiguous; consider 'R² (FCO2)' and 'R² (V)' for clarity.
  6. §3.2.7: The text states 'The total score is 2.13' (from Table 5) but then Table 6 reports Sapp = 2.5. The relationship between these two numbers should be clarified — are they different metrics? The current presentation risks confusion.
  7. Table 5: The scoring formula uses d = 4 ('very good'), but the gradation scale in §3.2.4 defines only a, b, c. The 'd' level is introduced without prior definition in the main text.
  8. Figure 5 caption: 'R2 = 0.982, MAPE = 7.84%' — these values match Patient 3 in Table 3, but the figure label says 'Patient 3' while the caption text does not specify. Minor inconsistency.
  9. §4.5.1, Results paragraph: 'Patient 2 represents normal lung mechanics (E=6.94)' and 'Patient 2 exhibits obstructive pathology' — both sentences refer to Patient 2 but describe different profiles. One likely should read 'Patient 1.'
  10. Reference [8] is cited as 'In preperaation' (2026). If still unpublished at submission, this should be noted or replaced with a citable preprint.
  11. Table 11: The equation for GE1 appears to have a formatting issue in the last term — the nesting of diffusion terms is difficult to parse. Consider typesetting more clearly.
  12. §3.2.4: 'All categories except Bench test validation and in vivo validation were deemed relevant' — the rationale for excluding bench validation (lung simulators being model-based) is reasonable but could cite a specific example to strengthen the argument.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for a thorough and constructive review. The referee correctly identifies that the paper's central fitness-for-purpose claim is currently undermined by (1) missing inline evidence for four validation tests, (2) an insufficiently reconciled gap between the acknowledged observability limitation and the calibration subscore, and (3) the absence of any explicit accounting for incomplete verification in the credibility score and fitness-for-purpose conclusion. We agree with all three major comments and will revise the manuscript accordingly. The revisions are substantive but do not change the paper's fundamental contribution, which is the operationalization of ASME V&V 40 and FDA guidance for mechanical ventilation models.

read point-by-point responses

  1. Referee: §3.2.6, Tests V-3 through V-6: Four of nine validation tests defer all quantitative results to self-cited companion papers [15, 16] without presenting any inline figures, tables, or summary statistics. V-4 states only 'For results see reference [15]' with no quantitative claim whatsoever. V-3, V-5, and V-6 state acceptance criteria (MAPE <= 10%, differ < 5%) but provide no data to substantiate them. These tests feed directly into the population-based validation subscore (2.0) and the emergent behavior subscore (2.0) in Table 6, which in turn support the composite Sapp = 2.5 claim. Without at least summary statistics or a figure per test, the reader cannot independently assess whether the stated criteria are met. This is load-bearing for the central fitness-for-purpose claim and should be addressed by including inline evidence (at minimum a table of MAPE/bias values and one representative

    Authors: The referee is correct. Tests V-3 through V-6 currently defer all quantitative evidence to companion papers without presenting inline summary statistics, figures, or even a concise statement of whether acceptance criteria are met. This is a genuine gap: the reader cannot independently verify the claims that MAPE <= 10% (V-3, V-5) or < 5% difference (V-6) are satisfied, and V-4 provides no quantitative claim at all. Since these tests directly support the population-based validation subscore (2.0) and, through V-6, the emergent behavior subscore (2.0), the absence of inline evidence is load-bearing for the Sapp = 2.5 conclusion. We will revise the manuscript to include, for each of Tests V-3 through V-6, at minimum: (a) a summary table of the key quantitative metrics (MAPE, bias, or percentage difference as applicable) against the stated acceptance criteria, (b) one representative figure per test showing model output versus comparator data, and (c) an explicit pass/fail statement relative to the acceptance criteria. Where the companion papers contain additional detail, we will summarize the essential results inline and retain the reference for full methodological detail. We will also verify that the subscores in Table 6 and Table 12 are consistent with the inline evidence presented. revision: yes

  2. Referee: §3.2.6, Test V-7 and §4.5.1: The calibration uses three patients with only two measurement signals (airway flow and capnogram). The paper itself acknowledges (§4.5.1) that 'the available measurement set (flow and end-tidal CO2) does not fully satisfy the observability conditions required for complete state reconstruction.' Despite this, the calibration subscore is rated 3.0 (good) in Table 6, contributing to the Sapp = 2.5. The gap between the acknowledged observability limitation and the 'good' rating should be reconciled. At minimum, the authors should clarify which specific credibility factors in Table 12 drive the 3.0 subscore and why the observability limitation does not reduce the rating below 'good.'

    Authors: The referee identifies a legitimate inconsistency. The manuscript acknowledges in §4.5.1 that the measurement set (flow and capnogram) does not fully satisfy nonlinear observability conditions for complete state reconstruction, yet the calibration subscore in Table 6 is rated 3.0 (good) without explicitly addressing why this limitation does not reduce the rating. We will revise the manuscript to reconcile this gap. Specifically, we will: (a) clarify which credibility factors in Table 12 drive the calibration subscore — the applicability-relevant factors for calibration are 'Relevance of QoI' and 'Relevance to COU,' both rated (c), but we will make explicit how the full set of calibration factors (quality of data, quantity of data, inputs vs. params, goodness of fit) contributes; (b) add an explicit discussion of how the observability limitation affects the calibration rating, noting that it constrains the identifiability of certain internal states but does not prevent accurate reproduction of the measured outputs (flow and capnogram) that are the QoIs for the COU; and (c) consider whether the subscore should be adjusted downward or whether the limitation is better captured as a documented residual gap with a corresponding mitigation action (expanded measurement set in future work). We lean toward the latter interpretation — the observability limitation is a scope constraint on what can be validated, not evidence that the validated outputs are inaccurate — but we will make the reasoning transparent and let the reader assess it. revision: partial

  3. Referee: §3.2.6, Verification: The paper explicitly states that 'a systematic and complete verification of the model implementation... has not yet been fully executed and documented.' The authors are transparent about this, and the Discussion (§4.1) acknowledges that 'verification activities are currently ongoing.' However, ASME V&V 40 treats verification as a prerequisite for credible validation. The paper's own framework (§3.1, step 7) lists verification as a required step. The current text states validation results are 'conditional on the correctness of the current implementation' but does not factor this conditionality into the Sapp score or the adequacy assessment. The authors should either (a) explicitly state how the incomplete verification status affects the credibility score's interpretation, or (b) add a conditional qualifier to the 'fit for purpose' conclusion in the abstract and §3.2.

    Authors: The referee is correct that ASME V&V 40 treats verification as a prerequisite for credible validation, and that the current manuscript does not adequately factor the incomplete verification status into either the Sapp score or the fitness-for-purpose conclusion. The statement that validation results are 'conditional on the correctness of the current implementation' is present but does not propagate to the quantitative score or the abstract. We will address this in two ways. First, we will add an explicit conditional qualifier to the fitness-for-purpose conclusion in both the abstract and §3.2.7/§3.2.8, stating that the conclusion is provisional pending completion of systematic verification. Second, we will add a discussion paragraph in §3.2.7 explaining how the incomplete verification status affects the interpretation of Sapp = 2.5: the score reflects the strength of validation evidence collected to date, but the credibility argument is not fully closed until verification is complete. We considered whether to introduce a separate verification subscore or to adjust Sapp downward, but we believe the more honest approach is to keep Sapp as a characterization of the validation evidence while making the conditional status of the overall credibility argument explicit and prominent. This is consistent with the spirit of ASME V&V 40, which requires verification and validation to be co-equal prerequisites — we should not present the validation score in isolation without flagging that the verification pillar is incomplete. revision: yes

Circularity Check

3 steps flagged

Tests V-3 through V-6 defer quantitative results to self-cited companion papers [15, 16]; Test V-7 reports calibration fit as validation evidence.

specific steps
  1. self citation load bearing [Section 3.2.6, Tests V-3, V-5, V-6]
    "Test V–3: ... For results see reference [16], , simulation results are within 90% confidence interval and MAPE of simulation results is less than 10%. ... Test V–5: ... For results see For results see reference [16], MAPE of simulation results is less than 10%. ... Test V–6: ... For results see reference [16], simulation results differ less than 5% compared to literature data."

    Tests V-3, V-5, and V-6 state quantitative acceptance criteria (MAPE < 10%, differ < 5%) but defer all supporting results to reference [16], which is a companion paper by the same author group (Hennigs et al.). The paper claims these tests pass its acceptance criteria without presenting any inline data, figures, or analysis. The credibility score Sapp=2.5 partly rests on these tests contributing to the population-based validation subscore of 2.0. Since the evidence is entirely outsourced to a same-author companion paper with no inline substantiation, the validation claim is not independently supported within this paper.

  2. self citation load bearing [Section 3.2.6, Test V-4]
    "Test V–4: Lung Mechanics, COPD The model's representation of lung mechanics in patients with COPD is evaluated. Clinical data from ventilated COPD patients [11] provide the comparison. For results see reference [15]."

    Test V-4 is the only clinical-data comparison for COPD lung mechanics, directly relevant to the stated context of use. It provides no quantitative claim, no figure, and no acceptance criterion inline — it simply says 'For results see reference [15],' which is another companion paper by the same author group (Hennigs et al.). This test contributes to the model calibration subscore of 3.0, which is one of the two pillars lifting Sapp above the 'good' threshold of 2.3. The load-bearing evidence is entirely external to this paper and resides in a same-author publication.

  3. fitted input called prediction [Section 3.2.6, Test V-7; Appendix 4.5.1]
    "The model calibration which employed gradient-based optimization (Gauss-Newton with Levenberg-Marquardt regularization) to fit the ODE-based patient model to clinical data. ... Calibration results demonstrate strong agreement: estimated lung volume achieves R2 ≈95 %after leakage correction. Capnography signals show R2 ≈94 %... These results, obtained on clinical datasets of ventilated patients, confirm that the model accurately reproduces respiratory mechanics and gas exchange dynamics under realistic conditions."

    Test V-7 fits model parameters to clinical data from 3 patients using only 2 measurement signals (airway flow and capnogram), then reports R² values as evidence that 'the model accurately reproduces respiratory mechanics and gas exchange dynamics.' The paper itself acknowledges the measurement set 'does not fully satisfy the observability conditions required for complete state reconstruction.' Reporting goodness-of-fit on the same data used for parameter identification is a calibration result, not independent validation. Yet Test V-7 contributes to the calibration subscore of 3.0, which lifts Sapp to 2.5 ('good'). The R² values are statistically forced by the fitting procedure.

full rationale

The paper's central claim of Sapp=2.5 ('good') rests on four subscores: population-based validation (2.0), model calibration (3.0), model plausibility (3.0), and emergent behavior (2.0). Two of these pillars — calibration (3.0) and population-based validation (2.0) — are substantially weakened by circularity. Tests V-3, V-4, V-5, and V-6 defer all quantitative results to companion papers [15, 16] by the same author group, presenting no inline evidence to support stated acceptance criteria. Test V-7 reports calibration fit statistics (R²≈95%) as validation evidence when parameters were fitted to the same data. The model plausibility (3.0) and emergent behavior (2.0) subscores are more self-contained, relying on expert opinion and qualitative demonstrations. The paper is transparent about limitations (small sample, incomplete verification, partial observability), which is commendable, but the Sapp=2.5 score is not fully supported by independently verifiable evidence within the paper itself.

Axiom & Free-Parameter Ledger

11 free parameters · 7 axioms · 2 invented entities

The model contains approximately 11+ fitted parameter groups from Test V-7 calibration on three patients, plus literature-derived parameters with uncertainty. The axioms are mostly domain assumptions justified by the narrow context of use (preclinical, no cardiovascular coupling needed). The ad hoc acceptance criteria and scoring formula are the most fragile invented elements. No code or data is shipped, making independent verification of fitted values impossible.

free parameters (11)
  • E_L (lung elastance) = 13 mmHg/L (baseline); 6.94-23.42 mmHg/L (fitted per patient)
    Patient-specific parameter fitted via Gauss-Newton optimization in Test V-7 calibration
  • R (airway resistance) = 2 mmHg/L/s (baseline); 5.04-7.00 mmHg/L/s (fitted)
    Patient-specific parameter fitted in Test V-7 calibration
  • EELV (end-expiratory lung volume) = 3.50-4.00 L
    Fitted per patient in Test V-7
  • VD (anatomical dead space) = 0.172-0.545 L
    Fitted per patient in Test V-7
  • DL,O2 / DL,CO2 (diffusion capacities) = 9.05e-5 to 1.03e-4 / 1.81e-3 to 2.06e-3
    Fitted per patient in Test V-7
  • Vc (capillary volume) = 0.0634-0.0778 L
    Fitted per patient in Test V-7
  • GC (central chemoreflex gain) = 27.72 exc/ΔpH (baseline); 35 (UQ mean)
    Literature-derived but uncertain; included in sensitivity analysis
  • RRbase, RRamp, kRR,sens, RRoff, RRsc = 10, 2.1, 1.8, 0.1212, 6.545
    Respiratory rate parameters fitted to literature data [36,35]
  • rho_sed, rho_resp (sedation/failure scaling) = 0-1 range
    Adjustable scaling factors for clinical scenarios
  • eta (capnogram flow delay) = 0.115-0.232 s
    Fitted per patient in Test V-7
  • tau1, tau2 (capnogram time constants) = 2.89-28.72 s
    Fitted per patient in Test V-7
axioms (7)
  • domain assumption Constant heart rate and cardiac output assumed (no cardiovascular model)
    Section 2.2.2: justified because AWP does not adjust PEEP based on cardiovascular parameters
  • domain assumption Intubated patient with no airway leakage
    Section 2.2.3: compartmental lung structure with homogeneous pressure/volume/gas within each compartment
  • domain assumption Neurological control is normal, excluding disorders altering neural drive
    Section 2.2.3: limits applicability to patients with normal respiratory drive
  • domain assumption Model outputs are noise-free except for stochastic breath generation
    Section 2.2.3: deterministic model with stochastic component for physiological variability
  • domain assumption Lung simulators are model-based and cannot serve as independent bench comparators
    Section 1/3.2.4: justifies exclusion of bench test validation category from credibility assessment
  • standard math ASME V&V 40 and FDA guidance frameworks are applicable to ventilation PDMs
    Section 3.1: foundational assumption that general credibility frameworks transfer to this domain
  • ad hoc to paper MAPE <=10% or 90% CI containment constitutes acceptable agreement
    Section 3.2.4: acceptance criteria chosen by authors; some tests do not meet these and use 'well-reasoned justification' instead
invented entities (2)
  • Patient-device model (PDM) integration independent evidence
    purpose: Coupled respiratory mechanics, gas exchange, respiratory control, and ventilator model
    The integrated PDM is validated against clinical data and literature; emergent behaviors (auto-PEEP, asynchrony) arise from model interactions rather than explicit coding
  • Credibility scoring formula (Total Grade = LQ + DQ + V - 2) no independent evidence
    purpose: Composite quality grading for governing equations
    Equation (1) is an ad hoc scoring scheme; weights are equal and unvalidated; no external benchmark for the scoring rubric

pith-pipeline@v1.1.0-glm · 33224 in / 3507 out tokens · 580518 ms · 2026-07-08T13:11:35.658335+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

55 extracted references · 55 canonical work pages

  1. [1]

    T Bates.Lung Mechanics: an Inverse Modeling Approach

    Jason H. T Bates.Lung Mechanics: an Inverse Modeling Approach. English. Leiden: Cambridge University Press, 2009.isbn: 978-0-511-59547-9. (Visited on 07/12/2021)

    work page 2009

  2. [2]

    Simplified models for gas exchange in the human lungs

    Alona Ben-Tal. “Simplified models for gas exchange in the human lungs”. en. In: Journal of Theoretical Biology238.2 (Jan. 2006),pp.474–495.doi:10.1016/j.jtbi. 2005.06.005

    work page doi:10.1016/j.jtbi 2006

  3. [3]

    Unified mathematical model structure for in silico simulations

    F. Bilda et al. “Unified mathematical model structure for in silico simulations”. en. In:IFAC-PapersOnLine58.24 (2024), pp. 49–54.doi:10.1016/j.ifacol.2024. 11.010

    work page doi:10.1016/j.ifacol.2024 2024

  4. [4]

    Neu- ral network implementation of a three- phase model of respiratory rhythm gener- ation

    S. M. Botros and E. N. Brace. “Neu- ral network implementation of a three- phase model of respiratory rhythm gener- ation”. en. In:Biological Cybernetics63.2 (June 1990), pp. 143–153.doi:10.1007/ BF00203037

    work page 1990

  5. [5]

    A comprehensive simula- tor of the human respiratory system: Val- idation with experimental and simulated data

    Lorenzo Chiari, Guido Avanzolini, and Mauro Ursino. “A comprehensive simula- tor of the human respiratory system: Val- idation with experimental and simulated data”. en. In:Annals of Biomedical Engi- neering25.6(Nov.1997),pp.985–999.doi: 10.1007/BF02684134

    work page doi:10.1007/bf02684134 1997

  6. [6]

    De- veloping the UVA/Padova Type 1 Dia- betes Simulator: Modeling, Validation, Re- finements, and Utility

    Claudio Cobelli and Boris Kovatchev. “De- veloping the UVA/Padova Type 1 Dia- betes Simulator: Modeling, Validation, Re- finements, and Utility”. In:Journal of Di- abetes Science and Technology17 (Nov. 2023), pp. 1493–1505.doi:10 . 1177 / 19322968231195081

    work page 2023

  7. [7]

    Meal Simulation Model of the Glucose-Insulin System

    Chiara Dalla Man, Robert A. Rizza, and Claudio Cobelli. “Meal Simulation Model of the Glucose-Insulin System”. In:IEEE Transactions on Biomedical Engineering 54.10 (Oct. 2007), pp. 1740–1749.doi:10. 1109/TBME.2007.893506

    work page arXiv 2007

  8. [8]

    Considerations for the Credibility Assessment of Compu- tational Physiological Models,In prepera- tion

    Charlott Danielson et al. “Considerations for the Credibility Assessment of Compu- tational Physiological Models,In prepera- tion”. In: (2026)

    work page 2026

  9. [9]

    Measuring the ventilatory response to hypoxia

    James Duffin. “Measuring the ventilatory response to hypoxia”. en. In:The Journal of Physiology584.1 (Oct. 2007), pp. 285– 293.doi:10 . 1113 / jphysiol . 2007 . 138883

    work page 2007

  10. [10]

    Breathing: Motor Control of Diaphragm Muscle

    Matthew J. Fogarty, Carlos B. Mantilla, and Gary C. Sieck. “Breathing: Motor Control of Diaphragm Muscle”. en. In: Physiology33.2 (Mar. 2018), pp. 113–126. doi:10.1152/physiol.00002.2018

    work page doi:10.1152/physiol.00002.2018 2018

  11. [11]

    Surface EMG-based quantification of inspiratory effort: a quan- titative comparison with Pes

    Jan Graßhoff et al. “Surface EMG-based quantification of inspiratory effort: a quan- titative comparison with Pes”. In:Critical Care25.1 (2021).doi:10.1186/s13054- 021-03833-w

    work page doi:10.1186/s13054- 2021

  12. [12]

    Latin hyper- cube sampling and the propagation of un- certainty in analyses of complex systems

    J.C. Helton and F.J. Davis. “Latin hyper- cube sampling and the propagation of un- certainty in analyses of complex systems”. en. In:Reliability Engineering & System 22 Safety81.1 (July 2003), pp. 23–69.doi: 10.1016/S0951-8320(03)00058-9

    work page doi:10.1016/s0951-8320(03)00058-9 2003

  13. [13]

    Mathematical lung modelforlocalgasexchangebasedonEIT- measurements

    Carlotta Hennigs, Tobias Becher, and Philipp Rostalski. “Mathematical lung modelforlocalgasexchangebasedonEIT- measurements”. In:Current Directions in Biomedical Engineering8, no. 2 (2022)

    work page 2022

  14. [14]

    A switching lung mechanics model for detection of expira- tory flow limitation

    Carlotta Hennigs et al. “A switching lung mechanics model for detection of expira- tory flow limitation”. In:at - Automa- tisierungstechnik72.5 (2024), pp. 417–428. doi:doi:10.1515/auto-2023-0206.url: https://doi.org/10.1515/auto-2023- 0206

    work page doi:10.1515/auto-2023-0206.url: 2024

  15. [15]

    A switching lung mechanics model for detection of expira- tory flow limitation

    Carlotta Hennigs et al. “A switching lung mechanics model for detection of expira- tory flow limitation”. In:at - Automa- tisierungstechnik72.5 (2024), pp. 417–428. doi:doi:10.1515/auto-2023-0206

    work page doi:10.1515/auto-2023-0206 2024

  16. [16]

    Patient-Ventilator Interaction - Development of a Mathe- matical Model of the Respiratory Center

    Carlotta Hennigs et al. “Patient-Ventilator Interaction - Development of a Mathe- matical Model of the Respiratory Center”. In:Computer Methods and Programs in Biomedicine(2026), p. 109329.doi:10 . 1016/j.cmpb.2026.109329

    work page arXiv 2026

  17. [17]

    pH and the Henderson- Hasselbalch equation

    A.Gorman Hills. “pH and the Henderson- Hasselbalch equation”. en. In:The Ameri- can Journal of Medicine55.2 (Aug. 1973), pp. 131–133.doi:10 . 1016 / 0002 - 9343(73)90160-5

    work page 1973

  18. [18]

    Virtual pa- tient model for evaluating automated in- spired oxygen control

    Jacob Herrmann et al. “Virtual pa- tient model for evaluating automated in- spired oxygen control”. In:Comput. Biol. Medicine(2025).doi:10 . 1016 / j . compbiomed.2025.110505

    work page arXiv 2025

  19. [19]

    Sources of breathing pattern variability in the respiratory feedback control loop

    Jacek Jaworski and Jason H.T. Bates. “Sources of breathing pattern variability in the respiratory feedback control loop”. In: Journal of Theoretical Biology469 (2019), pp. 148–162.doi:10.1016/j.jtbi.2019. 03.001

    work page doi:10.1016/j.jtbi.2019 2019

  20. [20]

    Advancing in Silico Clinical Trials for Regulatory Adop- tion and Innovation

    Georgia Karanasiou et al. “Advancing in Silico Clinical Trials for Regulatory Adop- tion and Innovation”. In:IEEE Journal of Biomedical and Health Informatics29.4 (Apr. 2025), pp. 2654–2668.doi:10.1109/ JBHI.2024.3486538

    work page arXiv 2025

  21. [21]

    Experimental de- sign principles to choose the number of Monte Carlo replicates for stochastic eco- logical models

    Maureen C. Kennedy. “Experimental de- sign principles to choose the number of Monte Carlo replicates for stochastic eco- logical models”. en. In:Ecological Mod- elling394 (Feb. 2019), pp. 11–17.doi:10. 1016/j.ecolmodel.2018.12.022

    work page 2019

  22. [22]

    Field Quantization,

    Steffen Leonhardt and Marian Walter. Medizintechnische Systeme. Heidelberg: Springer Nature Link, 2016.isbn: 978-3- 642-41238-7.doi:10.1007/978- 3- 642- 41239-4

    work page doi:10.1007/978- 2016

  23. [23]

    Relationship between whole blood base excess and CO2 con- tent in vivo

    J.A. Loeppky et al. “Relationship between whole blood base excess and CO2 con- tent in vivo”. en. In:Respiration Physi- ology94.1 (Oct. 1993), pp. 109–120.doi: 10.1016/0034-5687(93)90060-N

    work page doi:10.1016/0034-5687(93)90060-n 1993

  24. [24]

    On Latin hypercube sam- pling

    Wei-Liem Loh. “On Latin hypercube sam- pling”. In:The Annals of Statistics24.5 (Oct. 1996).doi:10 . 1214 / aos / 1069362310

    work page 1996

  25. [25]

    Lumb.Nunn’s applied respira- tory physiology

    Andrew B. Lumb.Nunn’s applied respira- tory physiology. Eighth. Edinburgh; New York: Elsevier, 2017.isbn: 978-0-7020- 6294-0

    work page 2017

  26. [26]

    A lung model con- sidering age, sex and height: an in-silico study-gas exchange

    Sophia Merrath et al. “A lung model con- sidering age, sex and height: an in-silico study-gas exchange”. In:Proceedings on Automation in Medical Engineering Proc AUTOMED2, No 1 (2023).doi:10 . 18416/AUTOMED.2023

    work page 2023

  27. [27]

    A lung model con- sidering age, sex and height: an in-silico study-lung mechanics

    Sophia Merrath et al. “A lung model con- sidering age, sex and height: an in-silico study-lung mechanics”. In:Proceedings on Automation in Medical Engineering Proc AUTOMED2, No 1 (2023).doi:10 . 18416/AUTOMED.2023. 23

    work page 2023

  28. [28]

    A Mathemati- cal Model of the Human Respiratory Con- trol System

    Howard T. Milhorn et al. “A Mathemati- cal Model of the Human Respiratory Con- trol System”. In:Biophysical Journal5.1 (1965), pp. 27–46.doi:10.1016/S0006- 3495(65)86701-7

    work page doi:10.1016/s0006- 1965

  29. [29]

    Automatic de- tection of ineffective triggering and dou- ble triggering during mechanical venti- lation

    Qestra Mulqueeny et al. “Automatic de- tection of ineffective triggering and dou- ble triggering during mechanical venti- lation”. en. In:Intensive Care Medicine 33.11 (Nov. 2007), pp. 2014–2018.doi:10. 1007/s00134-007-0767-z

    work page 2007

  30. [30]

    SmartCare/PS - Das au- tomatisierte Weaning Protokoll

    Andreas Neumann and Hartmunt Schmidt. “SmartCare/PS - Das au- tomatisierte Weaning Protokoll”. In: (2015)

    work page 2015

  31. [31]

    Physiological and pathophysiological implications of upper airway reflexes in humans

    Takashi Nishino. “Physiological and pathophysiological implications of upper airway reflexes in humans”. In:Japanese Journal of Physiology50.1 (Feb. 2000), pp. 3–14.doi:10.2170/jjphysiol.50.3

    work page doi:10.2170/jjphysiol.50.3 2000

  32. [32]

    Oberkampf and Christopher J

    William L. Oberkampf and Christopher J. Roy.Verification, Validation, and Un- certainty Quantification in Scientific Com- puting. 2nd ed. Cambridge University Press, Apr. 2025.isbn: 978-1-00-903100- 4 978-1-316-51613-3.doi:10 . 1017 / 9781009031004

    work page 2025

  33. [33]

    In silico clin- ical trials: concepts and early adoptions

    Francesco Pappalardo et al. “In silico clin- ical trials: concepts and early adoptions”. In:Briefings in Bioinformatics20.5 (Sept. 2019), pp. 1699–1708.doi:10.1093/bib/ bby043

    work page doi:10.1093/bib/ 2019

  34. [34]

    Credibility as- sessment of in silico clinical trials for med- ical devices

    Pras Pathmanathan et al. “Credibility as- sessment of in silico clinical trials for med- ical devices”. In:PLOS Computational Bi- ology20.8 (Aug. 2024). Ed. by Mukesh Dhamala.doi:10.1371/journal.pcbi. 1012289. (Visited on 08/07/2025)

    work page doi:10.1371/journal.pcbi 2024

  35. [35]

    Tran- sient ventilatory response to hypoxia with andwithoutcontrolledalveolarPCO2

    W. J. Reynolds and H. T. Milhorn. “Tran- sient ventilatory response to hypoxia with andwithoutcontrolledalveolarPCO2”.In: Journal of applied physiology35.2 (1973), pp. 187–196.doi:10.1152/jappl.1973. 35.2.187

    work page doi:10.1152/jappl.1973 1973

  36. [36]

    Transient ventilatory response to graded hypercapnia in man

    W. J. Reynolds, H.T. Milhorn, and G. H. Holloman. “Transient ventilatory response to graded hypercapnia in man”. In:Jour- nal of Applied Physiology33.1 (July 1972), pp. 47–54.doi:10.1152/jappl.1972.33. 1.47

    work page doi:10.1152/jappl.1972.33 1972

  37. [37]

    Ac- curate determination of O2 dissociation curve of human blood above 98.7 percent saturation with data on O2 solubility in unmodified human blood from 0 degrees to 37 degrees C

    F J Roughton and J W Severinghaus. “Ac- curate determination of O2 dissociation curve of human blood above 98.7 percent saturation with data on O2 solubility in unmodified human blood from 0 degrees to 37 degrees C”. en. In:Journal of Applied Physiology35.6 (Dec. 1973), pp. 861–869. doi:10.1152/jappl.1973.35.6.861

    work page doi:10.1152/jappl.1973.35.6.861 1973

  38. [38]

    Automated characteri- zation of patient-ventilator interaction us- ing surface electromyography

    Julia Sauer et al. “Automated characteri- zation of patient-ventilator interaction us- ing surface electromyography”. en. In:An- nals of Intensive Care14.1 (Feb. 2024), p. 32.doi:10.1186/s13613-024-01259- 5

    work page doi:10.1186/s13613-024-01259- 2024

  39. [39]

    Dirk Schädler.Non-invasive estimation of the work rate of breathing during automated weaning with SmartCare/PS: a prospective observational trial, Clinical data. 2010

    work page 2010

  40. [40]

    Simple, accurate equations for human blood O2 dissocia- tion computations

    J. W. Severinghaus. “Simple, accurate equations for human blood O2 dissocia- tion computations”. In:Journal of Applied Physiology46.3 (Mar. 1979), pp. 599–602. doi:10.1152/jappl.1979.46.3.599

    work page doi:10.1152/jappl.1979.46.3.599 1979

  41. [41]

    Béla Suki and Jason H. T. Bates.Mathe- matical Modeling of the Healthy and Dis- eased Lung: Linking Structure, Biome- chanics, and Mechanobiology. en. Springer Nature Switzerland, 2025.isbn: 978-3-031- 95321-7 978-3-031-95322-4.doi:10.1007/ 978-3-031-95322-4

    work page 2025

  42. [42]

    The American Society of Mechanical En- gineers.ASME V&V 40-2018 - Assess- ing Credibility of Computational Modeling through Verification and Validation: Appli- cation to Medical Devices. 2018. 24

    work page 2018

  43. [43]

    Patient-ventilator asynchrony during assisted mechanical ventilation

    Arnaud W. Thille et al. “Patient-ventilator asynchrony during assisted mechanical ventilation”. In:Intensive Care Medicine 32.10 (Oct. 2006), pp. 1515–1522.doi:10. 1007/s00134-006-0301-8

    work page 2006

  44. [44]

    U.S. Department of Health and Human Services, Food and Drug Administration, and Center for Devices and Radiologi- cal Health.Assessing the Credibility of Computational Modeling and Simulation in Medical Device Submissions - Guidance for Industry and FDA Staff. 2023

    work page 2023

  45. [45]

    Rate Constants for O2 and CO Binding to the a and b Sub- units within the R and T States of Human Hemoglobin

    Satoru Unzai et al. “Rate Constants for O2 and CO Binding to the a and b Sub- units within the R and T States of Human Hemoglobin”. en. In:Journal of Biological Chemistry273.36 (Sept. 1998), pp. 23150– 23159.doi:10.1074/jbc.273.36.23150

    work page doi:10.1074/jbc.273.36.23150 1998

  46. [46]

    Assessment of Physiologic Variables and Subjective Com- fort Under Different Levels of Pressure Support Ventilation

    Michele Vitacca et al. “Assessment of Physiologic Variables and Subjective Com- fort Under Different Levels of Pressure Support Ventilation”. In:CHEST126.3 (Sept. 2004). Publisher: Elsevier, pp. 851– 859.doi:10.1378/chest.126.3.851

    work page doi:10.1378/chest.126.3.851 2004

  47. [47]

    Computational physiological models for individualised me- chanical ventilation: a systematic litera- ture review focussing on quality, availabil- ity, and clinical readiness

    R. S. P. Warnaar et al. “Computational physiological models for individualised me- chanical ventilation: a systematic litera- ture review focussing on quality, availabil- ity, and clinical readiness”. en. In:Critical Care27.1 (July 2023), p. 268.doi:10 . 1186/s13054-023-04549-9

    work page 2023

  48. [48]

    West and Andrew M

    John B. West and Andrew M. Luks.West’s pulmonary pathophysiology: the essentials. Ninth.Philadelphia:WoltersKluwer,2017. isbn: 978-1-4963-3944-7

    work page 2017

  49. [49]

    Spearman Rank Cor- relation

    Jerrold H. Zar. “Spearman Rank Cor- relation”. In:Encyclopedia of Biostatis- tics(Feb. 2005). Ed. by Peter Armitage and Theodore Colton.doi:10 . 1002 / 0470011815.b2a15150

    work page 2005

  50. [50]

    Lactate after exercise in man: II. Mathematical model

    P. Zouloumian and H. Freund. “Lactate after exercise in man: II. Mathematical model”. en. In:European Journal of Ap- plied Physiology and Occupational Physi- ology46.2 (May 1981), pp. 135–147.doi: 10.1007/BF00428866. 25 Appendix 4.3 Patient Model States and Parameters Table 7: Baseline patient model input parameters. Param.Description Value Unit Literatur...

    work page doi:10.1007/bf00428866 1981

  51. [51]

    5 GE2 Gas exchange O 2 in lungs dfO2(t) dt = 1 V(t) ( DL,O2 ( PcO2(t)−PAO2 ) + ( finsp,O2−fO2(t) )˙QAi− fO2(t) ( DL,CO2 ( PcCO2(t)−PACO2 ) +D L,O2 ( PcO2(t)−PAO2 )))

    work page

  52. [52]

    5 GE3 Gas exchange CO 2 in capillaries dPcCO2(t) dt = DL,CO2 σCO2 ·Vc (PACO2(t)−PcCO2(t)) + d·l2 σCO2 h·[HCO− 3 ](t)−δ·r2·PcCO2(t)[2] 5 GE4 Gas exchange z in capillaries d[HCO− 3 ](t) dt =δ·r2·σCO2·PcCO2(t)−δ·l2·h·[HCO− 3 ](t)[2] 5 GE5 Gas exchange O 2 in capillaries dPcO2(t) dt = DL,O2 σO2 ·Vc [ 1 + 4·Th σO2 d ˜f(PcO2(t)) dPcO2 ]−1 ( PAO2(t)−PcO2(t) ) .[...

    work page

  53. [53]

    36 Table 11 – Model governing equations

    7 ... 36 Table 11 – Model governing equations. ID Description Equation Literature Grade BL3 Conversion of partial pressure to concentration (CO2) cCO2 =( 1− 0.02924 L/g·Th 2.244−0.422·SO2 ,vir·(8.74−pH) ) ·0.0301 L/mmHg·pCO2· ( 1 + 10pH−6.1 ) ·2.226

    work page

  54. [54]

    6 BL4 Conversion of partial pressure to concentration (O2) pO2,vir =pO 2 (40 mmHg pCO2 )0.3

    work page

  55. [55]

    37 Table 11 – Model governing equations

    7 cO2 =k 1·SO2,vir +k 2·pO2 [5] 7 RC1 Central chemical drive of pH level in medullar dDC(t) dt = 1 τC ( GC(θpH,m−pHm)−DC(t) ) modified from [19, 4] 5 Peripheral chemical drive ofP aO2 dDO2(t) dt = 1 τP ( GO2(PaO2,th−PaO2)−DO2(t) ) modified from [19, 4] 5 Peripheral chemical drive ofP aCO2 and pH dDCO2(t) dt = 1 τP ( GCO2PaCO2 +G CO2,pH(θpH,a−pHa)−DCO2(t) ...

    work page