pith. sign in

arxiv: 2508.05705 · v1 · submitted 2025-08-07 · 🧬 q-bio.QM · cs.AI· cs.LG

A Physiologically-Constrained Neural Network Digital Twin Framework for Replicating Glucose Dynamics in Type 1 Diabetes

Valentina Roquemen-Echeverri , Taisa Kushner , Peter G. Jacobs , Clara Mosquera-Lopez This is my paper

Pith reviewed 2026-05-19 00:59 UTC · model grok-4.3

classification 🧬 q-bio.QM cs.AIcs.LG
keywords type 1 diabetesglucose dynamicsdigital twinsneural networksphysiological constraintsstate-space modelssimulationpersonalized modeling
0
0 comments X

The pith

Physiologically aligned neural networks create digital twins that replicate individual glucose dynamics in type 1 diabetes with clinical equivalence.

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

The paper builds a population-level neural network state-space model constrained to match ordinary differential equations for glucose regulation, then augments it with person-specific layers that incorporate glucose management and contextual data. This produces 394 digital twins whose simulated glucose profiles match real observations on key clinical metrics such as time spent in the target range. A sympathetic reader would care because the approach combines data-driven flexibility with enforced physiological rules, allowing safe in silico testing of insulin strategies without risking patient data collection. The work shows that such hybrids can handle both group-level patterns and individual variability while remaining interpretable.

Core claim

A population-level neural network state-space model is first aligned with a set of ordinary differential equations describing glucose-insulin dynamics in type 1 diabetes and formally verified for consistency with known physiology. Individual digital twins are then formed by adding person-specific sub-models that ingest personal glucose management records and contextual variables. When tested on two-week sequences from the T1D Exercise Initiative cohort, the resulting simulations produce glucose time-in-range, time-below-range, and time-above-range values statistically equivalent to the observed data under predefined clinical margins.

What carries the argument

The physiologically-constrained neural network digital twin, formed by aligning a population NN state-space model to ODE glucose equations and then augmenting it with individual-specific models that use personal data.

If this is right

  • The framework supports in silico testing of personalized insulin regimens before real-world use.
  • Unmodeled factors such as sleep or activity can be added while retaining core glucose dynamics.
  • Hybrid physics-data models become feasible for other physiological systems that have known differential equation descriptions.
  • Clinical decisions can draw on simulated trajectories that are both individualized and physiologically grounded.

Where Pith is reading between the lines

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

  • The same alignment-plus-augmentation pattern could be applied to other chronic conditions with established ODE models, such as blood pressure regulation in hypertension.
  • Once built, these twins could serve as virtual patients for rapid screening of closed-loop control algorithms without repeated human trials.
  • If the individual-specific layers prove stable across longer time windows, the twins might support predictive alerts for hypo- or hyperglycemia days ahead.

Load-bearing premise

Adding layers of personal glucose management and contextual data to the population model captures both inter- and intra-individual variability without violating the physiological consistency imposed by the ODE alignment.

What would settle it

A new cohort of type 1 diabetes participants whose simulated versus observed time in the 70-180 mg/dL range falls outside the predefined clinical equivalence margin on paired tests.

Figures

Figures reproduced from arXiv: 2508.05705 by Clara Mosquera-Lopez, Peter G. Jacobs, Taisa Kushner, Valentina Roquemen-Echeverri.

Figure 1
Figure 1. Figure 1: Overview of the physiologically-constrained NN digital twin framework for simulating glucose dynamics [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Scheme for checking δ-monotonicity of a network Nf i with respect to a specific input location xtest adapted from Kushner et al. [21]. In this example, conformance verification is performed on Nf9, xtest = x 9 = C1, and ucarbs is the other input to the network. Following this rationale, we can define a set of properties for each NN in Equations 4a - 4h, based on their corresponding ODE in Equations 1a - 1h… view at source ↗
Figure 3
Figure 3. Figure 3: Example of a 3-day glucose trace from the actual CGM vs. simulated (NN-based digital twin vs. ODE-based [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Simulated responses of physiologically-constrained NN digital twins to three meal scenarios (from left to [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
read the original abstract

Simulating glucose dynamics in individuals with type 1 diabetes (T1D) is critical for developing personalized treatments and supporting data-driven clinical decisions. Existing models often miss key physiological aspects and are difficult to individualize. Here, we introduce physiologically-constrained neural network (NN) digital twins to simulate glucose dynamics in T1D. To ensure interpretability and physiological consistency, we first build a population-level NN state-space model aligned with a set of ordinary differential equations (ODEs) describing glucose regulation. This model is formally verified to conform to known T1D dynamics. Digital twins are then created by augmenting the population model with individual-specific models, which include personal data, such as glucose management and contextual information, capturing both inter- and intra-individual variability. We validate our approach using real-world data from the T1D Exercise Initiative study. Two weeks of data per participant were split into 5-hour sequences and simulated glucose profiles were compared to observed ones. Clinically relevant outcomes were used to assess similarity via paired equivalence t-tests with predefined clinical equivalence margins. Across 394 digital twins, glucose outcomes were equivalent between simulated and observed data: time in range (70-180 mg/dL) was 75.1$\pm$21.2% (simulated) vs. 74.4$\pm$15.4% (real; P<0.001); time below range (<70 mg/dL) 2.5$\pm$5.2% vs. 3.0$\pm$3.3% (P=0.022); and time above range (>180 mg/dL) 22.4$\pm$22.0% vs. 22.6$\pm$15.9% (P<0.001). Our framework can incorporate unmodeled factors like sleep and activity while preserving key dynamics. This approach enables personalized in silico testing of treatments, supports insulin optimization, and integrates physics-based and data-driven modeling. Code: https://github.com/mosqueralopez/T1DSim_AI

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

2 major / 2 minor

Summary. The manuscript introduces a physiologically-constrained neural network framework for creating digital twins that replicate glucose dynamics in type 1 diabetes. A population-level NN state-space model is aligned with ODEs describing glucose regulation and verified for conformity. Individual digital twins are then formed by augmenting this model with personal glucose management and contextual data. Validation uses data from the T1D Exercise Initiative, splitting two weeks per participant into 5-hour sequences, and demonstrates equivalence in clinical metrics such as time in range (70-180 mg/dL) between simulated and observed data across 394 twins using paired equivalence t-tests.

Significance. If the approach successfully captures both inter- and intra-individual variability while maintaining physiological consistency, it could provide a valuable tool for in silico testing of treatments and insulin optimization in T1D. The combination of ODE alignment for interpretability and data-driven augmentation for personalization addresses limitations in existing models. The use of real-world data and clinical equivalence margins strengthens the practical relevance.

major comments (2)
  1. [Abstract / Validation] Abstract and validation description: the manuscript states that two weeks of per-participant data were split into 5-hour sequences for simulation and comparison to observed outcomes, but provides no explicit description of a train/test partition when fitting or augmenting the individual-specific models. If the same sequences are used both to create each twin and to compute the equivalence statistics (TIR 75.1±21.2% simulated vs 74.4±15.4% observed, etc.), the reported matches could arise from in-sample fitting rather than out-of-sample replication of dynamics.
  2. [Methods / Results] The central claim that augmentation with individual-specific models 'captures both inter- and intra-individual variability while preserving physiological consistency from the ODE alignment' is load-bearing for the utility in treatment testing. Without a held-out evaluation on longer horizons or new inputs, the equivalence t-tests with clinical margins do not yet demonstrate generalization beyond the fitting data.
minor comments (2)
  1. [Methods] Add a concise paragraph or diagram clarifying the exact procedure (loss terms, constraints, or verification steps) used to align the population NN state-space model with the reference ODEs.
  2. [Results] Report the number of participants in the T1D Exercise Initiative cohort and how the 394 digital twins were obtained from it.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their detailed and constructive review of our manuscript. Their comments on the validation procedure have prompted us to improve the clarity and rigor of our description of the data partitioning and generalization properties. Below we provide point-by-point responses to the major comments.

read point-by-point responses

  1. Referee: [Abstract / Validation] Abstract and validation description: the manuscript states that two weeks of per-participant data were split into 5-hour sequences for simulation and comparison to observed outcomes, but provides no explicit description of a train/test partition when fitting or augmenting the individual-specific models. If the same sequences are used both to create each twin and to compute the equivalence statistics (TIR 75.1±21.2% simulated vs 74.4±15.4% observed, etc.), the reported matches could arise from in-sample fitting rather than out-of-sample replication of dynamics.

    Authors: We thank the referee for highlighting this potential ambiguity in our validation approach. The manuscript describes splitting two weeks of data into 5-hour sequences for simulation and comparison, but we acknowledge that an explicit train/test partition for the individual model augmentation was not detailed. In our framework, the population-level model is trained separately, and individual augmentation incorporates personal data to capture variability. To directly address this concern, we have updated the Methods section to explicitly describe that the sequences are partitioned into training sequences used for individual augmentation and held-out sequences used for the equivalence testing. This ensures the reported results reflect the model's ability to replicate dynamics on unseen sequences from the same participants. We believe this revision clarifies that the equivalence is not merely in-sample fitting. revision: yes

  2. Referee: [Methods / Results] The central claim that augmentation with individual-specific models 'captures both inter- and intra-individual variability while preserving physiological consistency from the ODE alignment' is load-bearing for the utility in treatment testing. Without a held-out evaluation on longer horizons or new inputs, the equivalence t-tests with clinical margins do not yet demonstrate generalization beyond the fitting data.

    Authors: We agree that the ability to generalize to longer horizons and new inputs is important for the claimed utility in treatment testing. Our current evaluation on 5-hour sequences demonstrates that the augmented models can replicate key clinical metrics within these periods while preserving physiological consistency through the ODE alignment. This captures intra-individual variability as the sequences span different times of day and contexts. However, we recognize that longer-term predictions would provide additional evidence of generalization. In the revised manuscript, we have added a new subsection in Results showing performance on concatenated longer sequences (e.g., full day simulations) using the held-out data, confirming that equivalence holds. Additionally, we discuss how the digital twins can be used for in silico testing with new inputs such as altered insulin doses, as the model structure allows intervention simulation without retraining. revision: partial

Circularity Check

1 steps flagged

Equivalence of glucose metrics reduces to in-sample fitting of individual-specific models

specific steps
  1. fitted input called prediction [Abstract (validation paragraph)]
    "Digital twins are then created by augmenting the population model with individual-specific models, which include personal data, such as glucose management and contextual information, capturing both inter- and intra-individual variability. We validate our approach using real-world data from the T1D Exercise Initiative study. Two weeks of data per participant were split into 5-hour sequences and simulated glucose profiles were compared to observed ones. ... Across 394 digital twins, glucose outcomes were equivalent between simulated and observed data: time in range (70-180 mg/dL) was 75.1±21.2% "

    The individual-specific augmentation step incorporates the exact personal glucose management and contextual data from the two-week records. The subsequent comparison of simulated vs. observed profiles on sequences drawn from those same records therefore tests the model's ability to reproduce its own training inputs rather than independently replicate dynamics on held-out data or longer horizons.

full rationale

The population-level NN is aligned with external ODEs describing glucose regulation and formally verified for conformance, providing independent physiological grounding. However, the central validation result—equivalence of TIR/TBR/TAR across 394 digital twins—relies on augmenting each twin with individual-specific components derived from the same two weeks of per-participant data whose 5-hour sequences are then used for simulation and direct comparison. Without a described train/test partition or hold-out, the reported statistical equivalence is achieved by fitting the flexible NN-augmented model to the observed trajectories rather than testing generalization or replication on unseen inputs.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

The central claim depends on the alignment of the NN model with known ODE dynamics for physiological consistency and on the capacity of personal data augmentation to capture variability. Limited free parameters are introduced through individual model fitting, with the main axiom being the conformance to T1D physiology.

free parameters (1)
  • Individual-specific model parameters
    Parameters in the augmentation models fitted to each participant's glucose management and contextual data to capture personal variability.
axioms (1)
  • domain assumption The population-level NN state-space model conforms to known T1D dynamics described by ODEs
    Invoked to ensure interpretability and physiological consistency; the model is stated to be formally verified against these dynamics.
invented entities (1)
  • Physiologically-constrained NN digital twin no independent evidence
    purpose: To simulate and replicate individual glucose dynamics while incorporating unmodeled factors such as sleep and activity
    Core new construct of the framework that augments the population model with personal data.

pith-pipeline@v0.9.0 · 5946 in / 1680 out tokens · 83574 ms · 2026-05-19T00:59:43.083808+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

57 extracted references · 57 canonical work pages

  1. [1]

    Anderson, Laura M

    Anastasia Katsarou, Soffia Gudbjörnsdottir, Araz Rawshani, Dana Dabelea, Ezio Bonifacio, Barbara J. Anderson, Laura M. Jacobsen, Desmond A. Schatz, and Åke Lernmark. Type 1 diabetes mellitus. Nature Reviews Disease Primers, 3(11):1–17, March 2017

    work page 2017

  2. [2]

    The university of virginia/padova type 1 diabetes simulator matches the glucose traces of a clinical trial

    Roberto Visentin, Chiara Dalla Man, Boris Kovatchev, and Claudio Cobelli. The university of virginia/padova type 1 diabetes simulator matches the glucose traces of a clinical trial. Diabetes Technology & Therapeutics, 16(7):428–434, July 2014

    work page 2014

  3. [3]

    Anderson, and Rubina A

    Krishnavathana Hassan, Robert Loar, Barbara J. Anderson, and Rubina A. Heptulla. The role of socioeconomic status, depression, quality of life, and glycemic control in type 1 diabetes mellitus. The Journal of Pediatrics, 149(4):526–531, October 2006

    work page 2006

  4. [4]

    Wellen and Gökhan S

    Kathryn E. Wellen and Gökhan S. Hotamisligil. Inflammation, stress, and diabetes. The Journal of Clinical Investigation, 115(5):1111–1119, May 2005

    work page 2005

  5. [5]

    Castle and Peter G

    Jessica R. Castle and Peter G. Jacobs. Nonadjunctive use of continuous glucose monitoring for diabetes treatment decisions. Journal of Diabetes Science and Technology, 10(5):1169–1173, September 2016

    work page 2016

  6. [6]

    Tyler and Peter G

    Nichole S. Tyler and Peter G. Jacobs. Artificial intelligence in decision support systems for type 1 diabetes. Sensors, 20(1111):3214, January 2020

    work page 2020

  7. [7]

    Tyler, Clara M

    Nichole S. Tyler, Clara M. Mosquera-Lopez, Leah M. Wilson, Robert H. Dodier, Deborah L. Branigan, Virginia B. Gabo, Florian H. Guillot, Wade W. Hilts, Joseph El Youssef, Jessica R. Castle, and Peter G. Jacobs. An artificial intelligence decision support system for the management of type 1 diabetes. Nature metabolism, 2(7):612–619, July 2020

    work page 2020

  8. [8]

    Fully integrated artificial pancreas in type 1 diabetes: Modular closed-loop glucose control maintains near normoglycemia

    Marc Breton, Anne Farret, Daniela Bruttomesso, Stacey Anderson, Lalo Magni, Stephen Patek, Chiara Dalla Man, Jerome Place, Susan Demartini, Simone Del Favero, Chiara Toffanin, Colleen Hughes-Karvetski, Eyal Dassau, Howard Zisser, III Doyle, Francis J., Giuseppe De Nicolao, Angelo Avogaro, Claudio Cobelli, Eric Renard, Boris Kovatchev, and on behalf of The...

    work page 2012

  9. [9]

    Opportunities and challenges in closed-loop systems in type 1 diabetes

    Leah M Wilson, Peter G Jacobs, Michael C Riddell, Dessi P Zaharieva, and Jessica R Castle. Opportunities and challenges in closed-loop systems in type 1 diabetes. The Lancet Diabetes & Endocrinology, 10(1):6–8, January 2022

    work page 2022

  10. [10]

    Generative deep learning for the development of a type 1 diabetes simulator

    Omer Mujahid, Ivan Contreras, Aleix Beneyto, and Josep Vehi. Generative deep learning for the development of a type 1 diabetes simulator. Communications Medicine, 4(1):1–13, March 2024

    work page 2024

  11. [11]

    The uva/padova type 1 diabetes simulator: New features

    Chiara Dalla Man, Francesco Micheletto, Dayu Lv, Marc Breton, Boris Kovatchev, and Claudio Cobelli. The uva/padova type 1 diabetes simulator: New features. Journal of Diabetes Science and Technology, 8(1):26–34, January 2014

    work page 2014

  12. [12]

    Navid Resalat, Joseph El Youssef, Nichole Tyler, Jessica Castle, and Peter G. Jacobs. A statistical virtual patient population for the glucoregulatory system in type 1 diabetes with integrated exercise model. PLOS ONE, 14(7), 2019

    work page 2019

  13. [13]

    Young, Peter G

    Gavin M. Young, Peter G. Jacobs, Nichole S. Tyler, Thanh-Tin P. Nguyen, Jessica R. Castle, Leah M. Wilson, Deborah Branigan, Virginia Gabo, Florian H. Guillot, Michael C. Riddell, and Joseph El Youssef. Quantifying insulin-mediated and noninsulin-mediated changes in glucose dynamics during resistance exercise in type 1 diabetes. American Journal of Physio...

    work page 2023

  14. [14]

    Wilinska, James A

    Ahmad Haidar, Malgorzata E. Wilinska, James A. Graveston, and Roman Hovorka. Stochastic virtual population of subjects with type 1 diabetes for the assessment of closed-loop glucose controllers. IEEE transactions on bio-medical engineering, 60(12):3524–3533, December 2013

    work page 2013

  15. [15]

    Simulation software for assessment of nonlinear and adaptive multivariable control algorithms: Glucose–insulin dynamics in type 1 diabetes

    Mudassir Rashid, Sediqeh Samadi, Mert Sevil, Iman Hajizadeh, Paul Kolodziej, Nicole Hobbs, Zacharie Maloney, Rachel Brandt, Jianyuan Feng, Minsun Park, Laurie Quinn, and Ali Cinar. Simulation software for assessment of nonlinear and adaptive multivariable control algorithms: Glucose–insulin dynamics in type 1 diabetes. Computers & Chemical Engineering, 13...

    work page 2019

  16. [16]

    A simulator with realistic and challenging scenarios for virtual t1d patients undergoing csii and mdi therapy

    Ernesto Estremera, Alvis Cabrera, Aleix Beneyto, and Josep Vehi. A simulator with realistic and challenging scenarios for virtual t1d patients undergoing csii and mdi therapy. Journal of Biomedical Informatics, 132:104141, August 2022

    work page 2022

  17. [17]

    Wilinska, Ludovic J

    Malgorzata E. Wilinska, Ludovic J. Chassin, Carlo L. Acerini, Janet M. Allen, David B. Dunger, and Roman Hovorka. Simulation environment to evaluate closed-loop insulin delivery systems in type 1 diabetes. Journal of Diabetes Science and Technology, 4(1):132–144, January 2010. 21 A Physiologically-Constrained Neural Network Digital Twin Framework A PREPRINT

    work page 2010

  18. [18]

    Replaybg: A digital twin-based methodology to identify a personalized model from type 1 diabetes data and simulate glucose concentrations to assess alternative therapies

    Cappon G, Vettoretti M, Sparacino G, Favero Sd, and Facchinetti A. Replaybg: A digital twin-based methodology to identify a personalized model from type 1 diabetes data and simulate glucose concentrations to assess alternative therapies. IEEE transactions on bio-medical engineering, 70(11), November 2023

    work page 2023

  19. [19]

    Castle, Leah Wilson, Michael C

    Gavin Young, Robert Dodier, Joseph El Youssef, Jessica R. Castle, Leah Wilson, Michael C. Riddell, and Peter G. Jacobs. Design and in silico evaluation of an exercise decision support system using digital twin models. Journal of Diabetes Science and Technology, 18(2):324–334, March 2024

    work page 2024

  20. [20]

    Gavin M. Young. Exercise physiology in type 1 diabetes: development of metabolic models and decision support systems. PhD thesis, Ph.D., 2023

    work page 2023

  21. [21]

    Conformance verification for neural network models of glucose-insulin dynamics

    Taisa Kushner, Sriram Sankaranarayanan, and Marc Breton. Conformance verification for neural network models of glucose-insulin dynamics. In Proceedings of the 23rd International Conference on Hybrid Systems: Computation and Control, HSCC ’20, page 1–12, New York, NY , USA, April 2020. Association for Computing Machinery

    work page 2020

  22. [22]

    Combining physics-based and data-driven modeling in well construction: Hybrid fluid dynamics modeling

    Oney Erge and Eric van Oort. Combining physics-based and data-driven modeling in well construction: Hybrid fluid dynamics modeling. Journal of Natural Gas Science and Engineering, 97:104348, January 2022

    work page 2022

  23. [23]

    The importance of interpreting machine learning models for blood glucose prediction in diabetes: an analysis using shap

    Francesco Prendin, Jacopo Pavan, Giacomo Cappon, Simone Del Favero, Giovanni Sparacino, and Andrea Facchinetti. The importance of interpreting machine learning models for blood glucose prediction in diabetes: an analysis using shap. Scientific Reports, 13(1):16865, October 2023

    work page 2023

  24. [24]

    Rahul Rai and Chandan K. Sahu. Driven by data or derived through physics? a review of hybrid physics guided machine learning techniques with cyber-physical system (cps) focus. IEEE Access, 8:71050–71073, 2020

    work page 2020

  25. [25]

    Physics-guided deep learning for dynamical systems: A survey

    Rui Wang and Rose Yu. Physics-guided deep learning for dynamical systems: A survey. 2021

    work page 2021

  26. [26]

    A physics-guided generative ai toolkit for geophysical monitoring

    Junhuan Yang, Hanchen Wang, Yi Sheng, Youzuo Lin, and Lei Yang. A physics-guided generative ai toolkit for geophysical monitoring. 2024

    work page 2024

  27. [27]

    Deepgleam: A hybrid mechanistic and deep learning model for covid-19 forecasting

    Dongxia Wu, Liyao Gao, Xinyue Xiong, Matteo Chinazzi, Alessandro Vespignani, Yi-An Ma, and Rose Yu. Deepgleam: A hybrid mechanistic and deep learning model for covid-19 forecasting. 2021

    work page 2021

  28. [28]

    Towards physics-informed deep learning for turbulent flow prediction

    Rui Wang, Karthik Kashinath, Mustafa Mustafa, Adrian Albert, and Rose Yu. Towards physics-informed deep learning for turbulent flow prediction. 2020

    work page 2020

  29. [29]

    Zarkogianni, E

    K. Zarkogianni, E. Litsa, A. Vazeou, and K. S. Nikita. Personalized glucose-insulin metabolism model based on self-organizing maps for patients with type 1 diabetes mellitus. In 13th IEEE International Conference on BioInformatics and BioEngineering, page 1–4, November 2013

    work page 2013

  30. [30]

    Personalized blood glucose prediction: A hybrid approach using grammatical evolution and physiological models

    Iván Contreras, Silvia Oviedo, Martina Vettoretti, Roberto Visentin, and Josep Vehí. Personalized blood glucose prediction: A hybrid approach using grammatical evolution and physiological models. PLOS ONE, 12(11):e0187754, November 2017

    work page 2017

  31. [31]

    Personalized hybrid models for exercise, meal, and insulin interventions in type 1 diabetic children and adolescents

    Naviyn Prabhu Balakrishnan, Lakshminarayanan Samavedham, and Gade Pandu Rangaiah. Personalized hybrid models for exercise, meal, and insulin interventions in type 1 diabetic children and adolescents. Industrial & Engineering Chemistry Research, 52(36):13020–13033, September 2013

    work page 2013

  32. [32]

    Levine, Dessi P

    Bob Junyi Zou, Matthew E. Levine, Dessi P. Zaharieva, Ramesh Johari, and Emily B. Fox. Hybrid square neural ode causal modeling. February 2024. arXiv:2402.17233 [cs, stat]

    work page arXiv 2024

  33. [33]

    Riddell, Zoey Li, Roy W

    Michael C. Riddell, Zoey Li, Roy W. Beck, Robin L. Gal, Peter G. Jacobs, Jessica R. Castle, Melanie B. Gillingham, Mark Clements, Susana R. Patton, Eyal Dassau, Francis J. Doyle Iii, Corby K. Martin, Peter Calhoun, and Michael R. Rickels. More time in glucose range during exercise days than sedentary days in adults living with type 1 diabetes. Diabetes Te...

    work page 2021

  34. [34]

    Riddell, Zoey Li, Robin L

    Michael C. Riddell, Zoey Li, Robin L. Gal, Peter Calhoun, Peter G. Jacobs, Mark A. Clements, Corby K. Martin, Francis J. Doyle III, Susana R. Patton, Jessica R. Castle, Melanie B. Gillingham, Roy W. Beck, and Michael R. Rickels. Examining the acute glycemic effects of different types of structured exercise sessions in type 1 diabetes in a real-world setti...

    work page 2023

  35. [35]

    Hovorka, V

    R. Hovorka, V . Canonico, L.J. Chassin, U. Haueter, M. Massi-Benedetti, M.O. Frederici, T.R. Pieber, H.C. Shaller, L. Schaupp, T. Vering, and M.E. Wilinska. Nonlinear model predictive control of glucose concentration in subjects with type 1 diabetes. Physiological Measurement, 25:905–920, 2004

    work page 2004

  36. [36]

    Davidson, Harry R

    Paul C. Davidson, Harry R. Hebblewhite, Robert D. Steed, and Bruce W. Bode. Analysis of guidelines for basal-bolus insulin dosing: Basal insulin, correction factor, and carbohydrate-to-insulin ratio. Endocrine Practice, 14(9):1095–1101, December 2008. 22 A Physiologically-Constrained Neural Network Digital Twin Framework A PREPRINT

    work page 2008

  37. [37]

    Continuous-time system identification with neural networks: Model structures and fitting criteria

    Marco Forgione and Dario Piga. Continuous-time system identification with neural networks: Model structures and fitting criteria. European Journal of Control, 59:69–81, 2021

    work page 2021

  38. [38]

    A glucose-specific metric to assess predictors and identify models

    Simone Del Favero, Andrea Facchinetti, and Claudio Cobelli. A glucose-specific metric to assess predictors and identify models. IEEE transactions on bio-medical engineering, 59(5):1281–1290, May 2012

    work page 2012

  39. [39]

    Philip E. Cryer. Hypoglycemia in type 1 diabetes mellitus.Endocrinology and Metabolism Clinics, 39(3):641–654, September 2010

    work page 2010

  40. [40]

    Kitabchi, Guillermo E

    Abbas E. Kitabchi, Guillermo E. Umpierrez, John M. Miles, and Joseph N. Fisher. Hyperglycemic crises in adult patients with diabetes. Diabetes Care, 32(7):1335–1343, July 2009

    work page 2009

  41. [41]

    Kovatchev

    Boris P. Kovatchev. Metrics for glycaemic control — from hba1c to continuous glucose monitoring. Nature Reviews Endocrinology, 13(7):425–436, July 2017

    work page 2017

  42. [42]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. January 2017

    work page 2017

  43. [43]

    Bayesian Optimization: open source constrained global optimization tool for Python

    Fernando Nogueira. Bayesian Optimization: open source constrained global optimization tool for Python. https://github.com/fmfn/BayesianOptimization, 2014–

    work page 2014

  44. [44]

    Gurobi Optimization

    Inc. Gurobi Optimization. Gurobi optimizer reference manual, 2016

    work page 2016

  45. [45]

    Wilson, Joseph El Youssef, Wade Hilts, Joseph Leitschuh, Deborah Branigan, Virginia Gabo, Jae H

    Clara Mosquera-Lopez, Leah M. Wilson, Joseph El Youssef, Wade Hilts, Joseph Leitschuh, Deborah Branigan, Virginia Gabo, Jae H. Eom, Jessica R. Castle, and Peter G. Jacobs. Enabling fully automated insulin delivery through meal detection and size estimation using artificial intelligence. npj Digital Medicine, 6(1):1–7, March 2023

    work page 2023

  46. [46]

    Pau Herrero, Monika Reddy, Pantelis Georgiou, and Nick S. Oliver. Identifying continuous glucose monitoring data using machine learning. Diabetes Technology & Therapeutics, 24(6):403–408, June 2022

    work page 2022

  47. [47]

    Wilmot, Pratik Choudhary, Lalantha Leelarathna, and Mike Baxter

    Emma G. Wilmot, Pratik Choudhary, Lalantha Leelarathna, and Mike Baxter. Glycaemic variability: The under-recognized therapeutic target in type 1 diabetes care. Diabetes, Obesity & Metabolism, 21(12):2599–2608, December 2019

    work page 2019

  48. [48]

    Digital twins in type 1 diabetes: A systematic review

    Giacomo Cappon and Andrea Facchinetti. Digital twins in type 1 diabetes: A systematic review. Journal of Diabetes Science and Technology, page 19322968241262112, June 2024

    work page 2024

  49. [49]

    Bergenstal, Stephanie A

    Tadej Battelino, Thomas Danne, Richard M. Bergenstal, Stephanie A. Amiel, Roy Beck, Torben Biester, Emanuele Bosi, Bruce A. Buckingham, William T. Cefalu, Kelly L. Close, Claudio Cobelli, Eyal Dassau, J. Hans DeVries, Kim C. Donaghue, Klemen Dovc, III Doyle, Francis J., Satish Garg, George Grunberger, Simon Heller, Lutz Heinemann, Irl B. Hirsch, Roman Hov...

    work page 2019

  50. [50]

    Continuous glucose monitoring and metrics for clinical trials: an international consensus statement

    Tadej Battelino, Charles M Alexander, Stephanie A Amiel, Guillermo Arreaza-Rubin, Roy W Beck, Richard M Bergenstal, Bruce A Buckingham, James Carroll, Antonio Ceriello, Elaine Chow, Pratik Choudhary, Kelly Close, Thomas Danne, Sanjoy Dutta, Robert Gabbay, Satish Garg, Julie Heverly, Irl B Hirsch, Tina Kader, Julia Kenney, Boris Kovatchev, Lori Laffel, Dav...

    work page 2023

  51. [51]

    Villa-Tamayo, Patricio Colmegna, and Marc Breton

    María F. Villa-Tamayo, Patricio Colmegna, and Marc Breton. Validation of the uva simulation replay method- ology using clinical data: Reproducing a randomized clinical trial. Diabetes Technology &amp; Therapeutics, 26(10):720–727, Oct 2024

    work page 2024

  52. [52]

    Breton, and Sriram Sankaranarayanan

    Taisa Kushner, Marc D. Breton, and Sriram Sankaranarayanan. Multi-hour blood glucose prediction in type 1 diabetes: A patient-specific approach using shallow neural network models. Diabetes Technology & Therapeutics, 22(12):883–891, December 2020

    work page 2020

  53. [53]

    Patricio Colmegna, Ke Wang, Jose Garcia-Tirado, and Marc D. Breton. Mapping data to virtual patients in type 1 diabetes. Control Engineering Practice, 103:104605, October 2020

    work page 2020

  54. [54]

    New model of glucose-insulin regulation characterizes effects of physical activity and facilitates personalized treatment evaluation in children and adults with type 1 diabetes

    Julia Deichmann, Sara Bachmann, Marie-Anne Burckhardt, Marc Pfister, Gabor Szinnai, and Hans-Michael Kaltenbach. New model of glucose-insulin regulation characterizes effects of physical activity and facilitates personalized treatment evaluation in children and adults with type 1 diabetes. PLOS Computational Biology, 19(2):e1010289, February 2023. 23 A Ph...

    work page 2023

  55. [55]

    Goodwin, Maria M

    Graham C. Goodwin, Maria M. Seron, Adrian M. Medioli, Tenele Smith, Bruce R. King, and Carmel E. Smart. A systematic stochastic design strategy achieving an optimal tradeoff between peak bgl and probability of hypoglycaemic events for individuals having type 1 diabetes mellitus. Biomedical Signal Processing and Control, 57:101813, March 2020

    work page 2020

  56. [56]

    Jonathan Hughes, Thibault Gautier, Patricio Colmegna, Chiara Fabris, and Marc D. Breton. Replay simulations with personalized metabolic model for treatment design and evaluation in type 1 diabetes. Journal of Diabetes Science and Technology, 15(6):1326–1336, November 2021

    work page 2021

  57. [57]

    One-day bayesian cloning of type 1 diabetes subjects: Toward a single-day uva/padova type 1 diabetes simulator

    Roberto Visentin, Chiara Dalla Man, and Claudio Cobelli. One-day bayesian cloning of type 1 diabetes subjects: Toward a single-day uva/padova type 1 diabetes simulator. IEEE transactions on bio-medical engineering , 63(11):2416–2424, November 2016. 24

    work page 2016