pith. sign in

arxiv: 2511.02849 · v2 · submitted 2025-10-26 · 📡 eess.SP · cs.CV· eess.IV

Benchmarking ResNet for Short-Term Hypoglycemia Classification with DiaData

Beyza Cinar , Maria Maleshkova This is my paper

Pith reviewed 2026-05-18 04:23 UTC · model grok-4.3

classification 📡 eess.SP cs.CVeess.IV
keywords hypoglycemia classificationResNetDiaDataType 1 diabetesdata imputationcontinuous glucose monitoringoutlier detectionheart rate correlation
0
0 comments X

The pith

Refining the DiaData collection of glucose and heart rate signals raises ResNet accuracy for hypoglycemia classification by 2-3 percent, with larger training sets adding another 7 percent.

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

The paper establishes that careful cleaning of a large integrated dataset from 2510 people with Type 1 diabetes lets a ResNet model detect hypoglycemia up to two hours ahead more reliably. Outliers are replaced with missing values using the interquartile range method, small gaps are filled with linear interpolation, and larger gaps receive Stineman interpolation to keep the time series realistic. The authors report that quality refinement alone improves results by 2-3 percent while adding more data contributes an extra 7 percent gain. Readers would care because earlier and more accurate warnings could reduce dangerous low-blood-sugar events in everyday diabetes care.

Core claim

The central claim is that preprocessing the DiaData collection by replacing IQR-identified outliers with missing values, imputing gaps of 25 minutes or less via linear interpolation and gaps between 30 and 120 minutes via Stineman interpolation, then training a ResNet classifier on the resulting Maindatabase and Subdatabase II produces better performance in short-term hypoglycemia classification than training on raw data, with the larger dataset contributing a 7 percent improvement and the cleaning step adding 2-3 percent.

What carries the argument

ResNet model trained on cleaned glucose and heart-rate time series to classify hypoglycemia onset up to two hours ahead, after IQR outlier removal and mixed linear-Stineman gap imputation.

If this is right

  • The cleaned DiaData supports more reliable forecasts of hypoglycemia from continuous sensor streams.
  • A moderate correlation appears between glucose and heart rate in the 15-to-60-minute window before events.
  • Stineman interpolation produces more plausible glucose traces than linear interpolation for gaps longer than 25 minutes.
  • Benchmarks on the full Maindatabase and the Subdatabase II set reference points for future classification models.

Where Pith is reading between the lines

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

  • The same outlier-and-interpolation pipeline could be tested on other wearable sensor streams for related time-series tasks.
  • The observed heart-rate correlation suggests it could be added as an auxiliary input to improve wearable alert systems.
  • Evaluating the refined model on completely new patient cohorts would show whether the gains transfer beyond the original collection.

Load-bearing premise

Replacing outliers with missing values and filling gaps with linear or Stineman interpolation leaves the true temporal patterns in glucose and heart rate signals intact enough that the ResNet learns real predictors rather than artifacts.

What would settle it

Retrain the same ResNet on the raw DiaData versus the IQR-cleaned and Stineman-imputed version; if accuracy does not rise or if switching to a different gap-filling method erases the reported 2-3 percent gain, the benefit of the described cleaning would be called into question.

Figures

Figures reproduced from arXiv: 2511.02849 by Beyza Cinar, Maria Maleshkova.

Figure 1
Figure 1. Figure 1: Missingness in Glucose Levels Before Data Imputation [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Missingness in Glucose Levels After Data Imputation [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Raw vs. Imputed Data for Subject 190.0 RT-CGM. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Confusion Matrices importance of data quality and quantity is highlighted, providing better generalization and increased performance. The correlation analysis suggests that additional context and sensor values can enhance classification performance, particularly in distinguishing between increased PHs [4]. This is also supported by the observation that even the smaller dataset exhibited similar patterns an… view at source ↗
read the original abstract

Individualized therapy is driven forward by medical data analysis, which provides insight into the patient's context. In particular, for Type 1 Diabetes (T1D), which is an autoimmune disease, relationships between demographics, sensor data, and context can be analyzed. However, outliers, noisy data, and small data volumes cannot provide a reliable analysis. Hence, the research domain requires large volumes of high-quality data. Moreover, missing values can lead to information loss. To address this limitation, this study improves the data quality of DiaData, an integration of 15 separate datasets containing glucose values from 2510 subjects with T1D. Notably, we make the following contributions: 1) Outliers are identified with the interquartile range (IQR) approach and treated by replacing them with missing values. 2) Small gaps ($\le$ 25 min) are imputed with linear interpolation and larger gaps ($\ge$ 30 and $<$ 120 min) with Stineman interpolation. Based on a visual comparison, Stineman interpolation provides more realistic glucose estimates than linear interpolation for larger gaps. 3) After data cleaning, the correlation between glucose and heart rate is analyzed, yielding a moderate relation between 15 and 60 minutes before hypoglycemia ($\le$ 70 mg/dL). 4) Finally, a benchmark for hypoglycemia classification is provided with a state-of-the-art ResNet model. The model is trained with the Maindatabase and Subdatabase II of DiaData to classify hypoglycemia onset up to 2 hours in advance. Training with more data improves performance by 7% while using quality-refined data yields a 2-3% gain compared to raw data.

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

1 major / 1 minor

Summary. The manuscript describes a data preprocessing pipeline for the integrated DiaData dataset (2510 T1D subjects) that identifies outliers via IQR and replaces them with missing values, imputes gaps ≤25 min by linear interpolation and 30-120 min gaps by Stineman interpolation (chosen after visual inspection), analyzes glucose-heart rate correlations before hypoglycemia events (≤70 mg/dL), and benchmarks a ResNet classifier for hypoglycemia onset prediction up to 2 hours ahead. The central empirical claims are a 7% performance improvement when training on the larger Maindatabase versus Subdatabase II and a 2-3% gain when using the quality-refined data versus raw data.

Significance. If the reported gains prove robust to imputation artifacts and are supported by exact metrics, statistical tests, and ablation studies, the work would provide a useful public benchmark and data-quality protocol for ML-based hypoglycemia prediction in T1D. The integration of multiple datasets and the explicit comparison of raw versus refined data are constructive steps toward reproducible research in this domain.

major comments (1)
  1. [Abstract] Abstract and § on data refinement: the headline claim of a 2-3% gain from quality-refined versus raw data is load-bearing for the paper’s contribution on preprocessing. The pipeline replaces IQR outliers with missing values and imputes via linear (≤25 min) or Stineman (30-120 min) interpolation, yet the manuscript supplies only a visual comparison for Stineman realism and no quantitative checks (e.g., pre/post autocorrelation, variance, or cross-signal correlation statistics) that the imputed series preserve the original temporal dynamics of glucose and heart-rate signals. Because ResNet is a high-capacity 1-D model, even modest imputation-induced smoothness could produce the observed small gain without reflecting genuine improvement in 2-hour-ahead hypoglycemia classification.
minor comments (1)
  1. [Abstract] Abstract and results section: exact performance numbers (accuracy, F1, AUC, etc.), cross-validation scheme, statistical significance tests, and ablation tables comparing raw vs. refined data are not reported, making the stated 7% and 2-3% deltas difficult to interpret or reproduce.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive and detailed comments on our manuscript. The concern regarding quantitative validation of the imputation procedure is well-taken and directly addresses the robustness of our central claim about the 2–3% performance gain from data refinement. We address this point below and will incorporate the suggested analyses in the revised manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract and § on data refinement: the headline claim of a 2-3% gain from quality-refined versus raw data is load-bearing for the paper’s contribution on preprocessing. The pipeline replaces IQR outliers with missing values and imputes via linear (≤25 min) or Stineman (30-120 min) interpolation, yet the manuscript supplies only a visual comparison for Stineman realism and no quantitative checks (e.g., pre/post autocorrelation, variance, or cross-signal correlation statistics) that the imputed series preserve the original temporal dynamics of glucose and heart-rate signals. Because ResNet is a high-capacity 1-D model, even modest imputation-induced smoothness could produce the observed small gain without reflecting genuine improvement in 2-hour-ahead hypoglycemia classification.

    Authors: We agree that the absence of quantitative checks on the imputed signals is a limitation of the current manuscript and that visual inspection alone is insufficient to fully rule out imputation artifacts. In the revision we will add explicit pre- and post-imputation comparisons, including autocorrelation functions at multiple lags, preservation of signal variance, and cross-correlation statistics between glucose and heart-rate time series. These metrics will be reported both for the linear and Stineman segments to demonstrate that the chosen interpolation does not introduce excessive smoothness beyond what is physiologically plausible. We selected Stineman interpolation after observing that linear interpolation over 30–120 min gaps produces unrealistically flat segments that violate the known non-linear dynamics of glucose excursions; however, we acknowledge that this rationale must be supported by the quantitative evidence the referee requests. Regarding the possibility that the modest 2–3% gain arises from smoothness rather than genuine signal improvement, we note that the gain is observed consistently across multiple prediction horizons (30 min to 2 h) and evaluation metrics. Nevertheless, the additional statistics and any necessary ablation experiments will be included to address this concern directly. We will also clarify in the text that the 7% gain from the larger Maindatabase versus Subdatabase II is obtained under identical preprocessing, so it is not confounded by the imputation step. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical ML benchmarking with direct data comparisons

full rationale

The paper presents an empirical workflow: IQR-based outlier replacement, linear/Stineman interpolation for gaps, correlation analysis, and ResNet training for hypoglycemia classification up to 2 hours ahead. Reported gains (+7% from larger database, +2-3% from refined vs raw data) are obtained by training and evaluating the same model architecture on the same underlying DiaData under different preprocessing conditions. No equations, fitted parameters, or derivations appear that reduce these performance deltas to quantities defined by the preprocessing steps themselves. The central results rest on standard train/test splits and metric comparisons rather than any self-referential construction or self-citation chain, rendering the derivation chain self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central performance claims rest on the validity of IQR outlier treatment and the choice of Stineman interpolation for larger gaps, both treated as domain-appropriate without further justification beyond visual inspection.

axioms (2)
  • domain assumption Interquartile range reliably identifies outliers in continuous glucose monitoring time series
    Used to replace outliers with missing values before imputation.
  • domain assumption Stineman interpolation yields more realistic glucose trajectories than linear interpolation for gaps of 30-120 minutes
    Chosen after visual comparison only.

pith-pipeline@v0.9.0 · 5843 in / 1217 out tokens · 46805 ms · 2026-05-18T04:23:40.350787+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.

  • IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    Outliers are identified with the interquartile range (IQR) approach and treated by replacing them with missing values. Small gaps (≤ 25 min) are imputed with linear interpolation and larger gaps (≥ 30 and < 120 min) with Stineman interpolation.

  • IndisputableMonolith/Foundation/ArithmeticFromLogic.lean LogicNat recovery unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    a benchmark for hypoglycemia classification is provided with a state-of-the-art ResNet model... Training with more data improves performance by 7% while using quality-refined data yields a 2-3% gain

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

38 extracted references · 38 canonical work pages

  1. [1]

    Brussels, Belgium: International Diabetes Federation, 11 ed., 2025

    International Diabetes Federation,IDF Diabetes Atlas. Brussels, Belgium: International Diabetes Federation, 11 ed., 2025

    work page 2025

  2. [2]

    14. Children and Adolescents: Standards of Care in Diabetes—2024,

    N. A. ElSayed, G. Aleppo, R. R. Bannuru, D. Bruemmer, B. S. Collins, L. Ekhlaspour, M. E. Hilliard, E. L. Johnson, K. Khunti, I. Lingvay, G. Matfin, R. G. McCoy, M. L. Perry, S. J. Pilla, S. Polsky, P. Prahalad, R. E. Pratley, A. R. Segal, J. J. Seley, R. C. Stanton, and R. A. Gabbay, “14. Children and Adolescents: Standards of Care in Diabetes—2024,”Diab...

    work page 2024

  3. [3]

    Type 1 Diabetes,

    R. S. W. Jessica Lucier, “Type 1 Diabetes,” Jan. 2023

    work page 2023

  4. [4]

    Data-based algorithms and models using diabetics real data for blood glucose and hypoglycaemia prediction – a systematic literature review,

    V . Felizardo, N. M. Garcia, N. Pombo, and I. Megdiche, “Data-based algorithms and models using diabetics real data for blood glucose and hypoglycaemia prediction – a systematic literature review,”Artificial Intelligence in Medicine, vol. 118, p. 102120, Aug. 2021

    work page 2021

  5. [5]

    Fear of hypoglycemia and diabetes distress: Expected reduction by glucose prediction,

    D. Ehrmann, L. Laviola, L.-S. Priesterroth, N. Hermanns, N. Babion, and T. Glatzer, “Fear of hypoglycemia and diabetes distress: Expected reduction by glucose prediction,”Journal of Diabetes Science and Technology, vol. 18, p. 1027–1034, Aug. 2024

    work page 2024

  6. [6]

    GlucoBench: Curated List of Continuous Glucose Monitoring Datasets with Prediction Benchmarks,

    R. Sergazinov, E. Chun, V . Rogovchenko, N. Fernandes, N. Kasman, and I. Gaynanova, “GlucoBench: Curated List of Continuous Glucose Monitoring Datasets with Prediction Benchmarks,” 2024. Version Number: 1

    work page 2024

  7. [7]

    Presenting diadata for research on type 1 diabetes,

    B. Cinar and M. Maleshkova, “Presenting diadata for research on type 1 diabetes,” 2025

    work page 2025

  8. [8]

    Diabetes classification applica- tion with efficient missing and outliers data handling algorithms,

    H. Torkey, E. Ibrahim, E. E.-D. Hemdan, A. El-Sayed, and M. A. Shouman, “Diabetes classification applica- tion with efficient missing and outliers data handling algorithms,”Complex &amp; Intelligent Systems, vol. 8, p. 237–253, Apr. 2021

    work page 2021

  9. [9]

    The impact of missing continuous blood glucose samples on machine learning models for predicting postprandial hypoglycemia: An experimental analysis,

    N. U. Rehman, I. Contreras, A. Beneyto, and J. Vehi, “The impact of missing continuous blood glucose samples on machine learning models for predicting postprandial hypoglycemia: An experimental analysis,”Mathematics, vol. 12, p. 1567, May 2024

    work page 2024

  10. [10]

    Assessing the accuracy of cgm metrics: The role of missing data and imputation strategies,

    S. L. Cichosz, T. Kronborg, S. Hangaard, P. Vestergaard, and M. H. Jensen, “Assessing the accuracy of cgm metrics: The role of missing data and imputation strategies,” Feb. 2025

    work page 2025

  11. [11]

    Application of the outlier detection method for web-based blood glucose level monitoring system,

    R. A. Nurhaliza, M. Q. H. Octava, F. M. Hilmy, U. Farooq, and G. Alfian, “Application of the outlier detection method for web-based blood glucose level monitoring system,”Bulletin of Electrical Engineering and Informatics, vol. 13, p. 2809–2816, Aug. 2024

    work page 2024

  12. [12]

    Heart rate and cgm feature representation diabetes detection from heart rate: Learning joint features of heart rate and continuous glucose monitors yields better representations,

    H. Rashtian, S. S. Torbaghan, S. Rahili, M. Snyder, and N. Aghaeepour, “Heart rate and cgm feature representation diabetes detection from heart rate: Learning joint features of heart rate and continuous glucose monitors yields better representations,”IEEE Access, vol. 9, p. 83234–83240, 2021

    work page 2021

  13. [13]

    Analyzing the performance of transformers for the prediction of the blood glucose level considering imputation and smoothing,

    E. Acuna, R. Aparicio, and V . Palomino, “Analyzing the performance of transformers for the prediction of the blood glucose level considering imputation and smoothing,”Big Data and Cognitive Computing, vol. 7, p. 41, Feb. 2023

    work page 2023

  14. [14]

    Predicting Glycaemia in Type 1 Diabetes Patients: Experiments in Feature Engineering and Data Imputation.,

    J. Jeon, P. J. Leimbigler, G. Baruah, M. H. Li, Y . Fossat, and A. J. Whitehead, “Predicting Glycaemia in Type 1 Diabetes Patients: Experiments in Feature Engineering and Data Imputation.,”Journal of healthcare informatics research, vol. 4, pp. 71–90, Mar. 2020. Place: Switzerland

    work page 2020

  15. [15]

    Feature- Based Machine Learning Model for Real-Time Hypoglycemia Prediction,

    D. Dave, D. J. DeSalvo, B. Haridas, S. McKay, A. Shenoy, C. J. Koh, M. Lawley, and M. Erraguntla, “Feature- Based Machine Learning Model for Real-Time Hypoglycemia Prediction,”Journal of Diabetes Science and Technology, vol. 15, pp. 842–855, July 2021

    work page 2021

  16. [16]

    Prediction of Glucose Concentration in Children with Type 1 Diabetes Using Neural Networks: An Edge Computing Application,

    F. D’Antoni, L. Petrosino, F. Sgarro, A. Pagano, L. V ollero, V . Piemonte, and M. Merone, “Prediction of Glucose Concentration in Children with Type 1 Diabetes Using Neural Networks: An Edge Computing Application,” Bioengineering, vol. 9, no. 5, 2022

    work page 2022

  17. [17]

    Transfer Learning in Hypoglycemia Classification,

    B. Cinar, F. Grensing, L. Van Den Boom, and M. Maleshkova, “Transfer Learning in Hypoglycemia Classification,” inArtificial Intelligence in Healthcare(X. Xie, I. Styles, G. Powathil, and M. Ceccarelli, eds.), vol. 14975, pp. 98–109, Cham: Springer Nature Switzerland, 2024. Series Title: Lecture Notes in Computer Science

    work page 2024

  18. [18]

    Deep learning-based hypoglycemia classification across multiple prediction horizons,

    B. Cinar, J. D. Onwuchekwa, and M. Maleshkova, “Deep learning-based hypoglycemia classification across multiple prediction horizons,” 2025

    work page 2025

  19. [19]

    F. Hüni, J. Garcia-Tirado, and K. Riesen,LSTM Networks and Graph Neural Networks for Predicting Events of Hypoglycemia, p. 52–61. Springer Nature Switzerland, 2025. 10 Benchmarking ResNet for Short-Term Hypoglycemia Classification with DiaData

    work page 2025

  20. [20]

    Intelligent real-time hypoglycemia prediction for type 1 diabetes,

    N. A. Mahdy, A. F. Elnokrashy, D. A. Hammad, and W. A. Mohamed, “Intelligent real-time hypoglycemia prediction for type 1 diabetes,” in2024 Intelligent Methods, Systems, and Applications (IMSA), p. 325–329, IEEE, July 2024

    work page 2024

  21. [21]

    Enhanced Predictive Monitoring of Hypoglycemia in Type 1 Diabetes Using Shapelet-Based Analysis of Wearable Sensor Data,

    J. I. Daniel Onwuchekwa, D. Claus, C. Weber, and M. Maleshkova, “Enhanced Predictive Monitoring of Hypoglycemia in Type 1 Diabetes Using Shapelet-Based Analysis of Wearable Sensor Data,” in2024 IEEE 12th International Conference on Serious Games and Applications for Health (SeGAH), (Funchal, Portugal), pp. 1–8, IEEE, Aug. 2024

    work page 2024

  22. [22]

    T1diabetesgranada: a longitudinal multi-modal dataset of type 1 diabetes mellitus,

    C. Rodriguez-Leon, M. D. Aviles-Perez, O. Banos, M. Quesada-Charneco, P. J. Lopez-Ibarra Lozano, C. Villalonga, and M. Munoz-Torres, “T1diabetesgranada: a longitudinal multi-modal dataset of type 1 diabetes mellitus,” Scientific Data, vol. 10, Dec. 2023

    work page 2023

  23. [23]

    Diatrend: A dataset from advanced diabetes technology to enable development of novel analytic solutions,

    T. Prioleau, A. Bartolome, R. Comi, and C. Stanger, “Diatrend: A dataset from advanced diabetes technology to enable development of novel analytic solutions,”Scientific Data, vol. 10, Aug. 2023

    work page 2023

  24. [24]

    Hupa-ucm diabetes dataset,

    J. Alvarado, “Hupa-ucm diabetes dataset,” 2024

    work page 2024

  25. [25]

    Diabetes datasets-shanghait1dm and shanghait2dm,

    J. Zhu, “Diabetes datasets-shanghait1dm and shanghait2dm,” 2022

    work page 2022

  26. [26]

    Dataset - diabetes adolescents time series with heart rate,

    ICT Innovaties Zorg, “Dataset - diabetes adolescents time series with heart rate,” 2025. Accessed: 2025-04-23

    work page 2025

  27. [27]

    Diabetes datasets - public data archive

    Jaeb Center for Health Research, “Diabetes datasets - public data archive.” https://public.jaeb.org/ datasets/diabetes, 2025. Accessed: 2025-04-23

    work page 2025

  28. [28]

    Accuracy of the third generation of a 14-day continuous glucose monitoring system,

    S. Alva, R. Brazg, K. Castorino, M. Kipnes, D. R. Liljenquist, and H. Liu, “Accuracy of the third generation of a 14-day continuous glucose monitoring system,”Diabetes Therapy, vol. 14, p. 767–776, Mar. 2023

    work page 2023

  29. [29]

    Defining continuous glucose monitor time in range in a large, community-based cohort without diabetes,

    N. L. Spartano, N. Sultana, H. Lin, H. Cheng, S. Lu, D. Fei, J. M. Murabito, M. E. Walker, H. A. Wolpert, and D. W. Steenkamp, “Defining continuous glucose monitor time in range in a large, community-based cohort without diabetes,”The Journal of Clinical Endocrinology &amp; Metabolism, vol. 110, p. 1128–1134, Sept. 2024

    work page 2024

  30. [30]

    Imputing missing multi-sensor data in the healthcare domain: A systematic review,

    V . Gupta, F. Grensing, B. Cinar, and M. Maleshkova, “Imputing missing multi-sensor data in the healthcare domain: A systematic review,”Image Vis. Comput., p. 105797, Oct. 2025

    work page 2025

  31. [31]

    Filling Missing Data Using Interpolation Methods: Study on the Effect of Fitting Distribution,

    M. Noor, A. Yahaya, N. Ramli, and A. M. M. Al Bakri, “Filling Missing Data Using Interpolation Methods: Study on the Effect of Fitting Distribution,”Key Engineering Materials, vol. 594–595, pp. 889–895, Dec. 2013. Publisher: Trans Tech Publications, Ltd

    work page 2013

  32. [32]

    Time to Hypoglycemia Prediction for Personalized Diabetes Care and Management,

    J. I. Daniel Onwuchekwa, B. Cinar, L. Van Den Boom, and M. Maleshkova, “Time to Hypoglycemia Prediction for Personalized Diabetes Care and Management,” in2024 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), (Orlando, FL, USA), pp. 1–7, IEEE, July 2024

    work page 2024

  33. [33]

    F. A. Pujol, T. Ramírez, and H. Mora,Deep Learning Techniques for Predicting Hypoglycemic Events in Diabetic Patients, p. 57–62. Springer Nature Switzerland, 2024

    work page 2024

  34. [34]

    Generalization of a deep learning model for continuous glucose monitoring–based hypoglycemia prediction: Algorithm development and validation study,

    J. Shao, Y . Pan, W.-B. Kou, H. Feng, Y . Zhao, K. Zhou, and S. Zhong, “Generalization of a deep learning model for continuous glucose monitoring–based hypoglycemia prediction: Algorithm development and validation study,” JMIR Medical Informatics, vol. 12, p. e56909–e56909, May 2024

    work page 2024

  35. [35]

    Deep learning for time series classification: a review,

    H. I. Fawaz, G. Forestier, J. Weber, L. Idoumghar, and P.-A. Muller, “Deep learning for time series classification: a review,”Data Mining and Knowledge Discovery, vol. 33, pp. 917–963, Mar. 2019. Publisher: Springer Science and Business Media LLC

    work page 2019

  36. [36]

    Time Series Classification from Scratch with Deep Neural Networks: A Strong Baseline,

    Z. Wang, W. Yan, and T. Oates, “Time Series Classification from Scratch with Deep Neural Networks: A Strong Baseline,” 2016

    work page 2016

  37. [37]

    Evaluating imputation techniques for short-term gaps in heart rate data,

    V . Gupta and M. Maleshkova, “Evaluating imputation techniques for short-term gaps in heart rate data,” 2025

    work page 2025

  38. [38]

    DiaData: An integrated large dataset for type 1 diabetes and hypoglycemia research,

    B. Cinar and M. Maleshkova, “DiaData: An integrated large dataset for type 1 diabetes and hypoglycemia research,” 2025. 11

    work page 2025