byluuu/gluco-tsfm-benchmark

Overview

This dataset aggregates multiple open-access CGM cohorts for blood glucose forecasting.

Included sub-datasets and their original download links

• BIG_IDEA_LAB: https://doi.org/10.13026/zthx-5212
• D1NAMO: https://zenodo.org/records/5651217
• HUPA-UCM: https://data.mendeley.com/datasets/3hbcscwz44/1
• Colas2019: https://pubmed.ncbi.nlm.nih.gov/31851681/
• ShanghaiT2DM: https://doi.org/10.6084/m9.figshare.c.6310860
• ShanghaiT1DM: https://doi.org/10.6084/m9.figshare.c.6310860
• UCHTT1DM: https://github.com/fisiologiacuantitativauc/UC_HT_T1DM (Undistributable, excluded from GlucoFM Bench HF dataset, please download from their Github)
• Bris-T1D Open: https://data.bris.ac.uk/data/dataset/33z5jc8fa6tob21ptrugzqog08
• T1DM-UOM: https://zenodo.org/records/15806142
• CGMacros: https://doi.org/10.13026/3z8q-x658
• AZT1D: https://data.mendeley.com/datasets/gk9m674wcx/1
• Hall2018: https://pubmed.ncbi.nlm.nih.gov/30040822/

Data format

Each row corresponds to one participant time series:

• dataset: sub-dataset name
• subject_id: participant identifier
• timestamp: numeric timestamp (e.g., Unix seconds or seconds since start)
• BGvalues: glucose values in mg/dL

Splits

Participant-level split:

• Train: 80%
• Test: 20% No participant appears in more than one split.

Usage

from datasets import load_dataset
ds = load_dataset("byluuu/gluco-tsfm-benchmark")

Dataset info:

  features:
  - name: dataset
    dtype: string
  - name: subject_id
    dtype: string
  - name: timestamp
    list: float64
  - name: BGvalue
    list: float64
  splits:
  - name: train
    num_bytes: 37175699
    num_examples: 549
  - name: test
    num_bytes: 9312019
    num_examples: 549
  download_size: 18556367
  dataset_size: 46487718
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*

License and Redistribution

This benchmark dataset aggregates multiple open-access datasets that permit redistribution under their respective licenses. Users must comply with the terms of each original dataset license.

The included datasets and their licenses are summarized below:

Sub-dataset	Access Type	License (if known)	Source / Notes
Hall_2018	Open Access	CC BY 4.0 / journal open data	Dataset downloadable via publication
D1NAMO	Open Access	CC BY-SA 4.0	Zenodo record shows CC BY-SA 4.0
Colas_2019	Open Access	CC BY 4.0 / journal open data	Published open access
BIG IDEAs	Open Access	ODC-By v1.0	Publication linked
ShanghaiT1DM	Open Access	CC BY 4.0	Open
ShanghaiT2DM	Open Access	CC BY 4.0	Open
UCHTT1DM (excluded)	Open Access but no derivatives	CC BY-NC-ND 4.0	Hosted on GitHub
HUPA-UCM	Open Access	CC BY 4.0	Mendeley Data (usually open)
CGMacros	Open Access	CC BY-NC-SA 4.0	Newly released open
T1D-UOM	Open Access	CC BY 4.0	Zenodo usually CC BY
BrisT1D-Open	Open Access	CC BY 4.0	Open
AZT1D	Open Access	CC BY 4.0	Open

Redistribution statement (important for compliance)

This repository redistributes only datasets whose licenses explicitly permit reuse and redistribution. For datasets that restrict redistribution, we provide preprocessing scripts and links to the original sources instead of hosting the raw data.

All dataset content remains subject to the original license terms specified above.

Citation

If you use this benchmark dataset, please cite the following:

@article{GlucoFMBench2025,
  title   = {GlucoFM-Bench: Benchmarking Time-Series Foundation Models for Blood Glucose Forecasting},
  author  = {Lu, Baiying and Liang, Zhaohui and Pontius, Ryan and Tang, Shengpu and Prioleau Temiloluwa},
  journal = {Under Submission},
  year    = {2026}
}

Original data sources

Please also cite the original datasets that compose this benchmark:

@article{Hall2018,
   author = {Heather Hall and Dalia Perelman and Alessandra Breschi and Patricia Limcaoco and Ryan Kellogg and Tracey McLaughlin and Michael Snyder},
   doi = {10.1371/JOURNAL.PBIO.2005143},
   issn = {1545-7885},
   issue = {7},
   journal = {PLoS biology},
   keywords = {Adult,Aged,Blood Glucose / metabolism*,Blood Glucose Self-Monitoring*,Carbohydrate Metabolism,Cohort Studies,Dalia Perelman,Diabetes Mellitus,Extramural,Female,Heather Hall,Homeostasis,Humans,Hyperglycemia / blood,Hyperglycemia / diagnosis,Insulin / metabolism,Internet,MEDLINE,Male,Meals,Michael Snyder,Middle Aged,N.I.H.,NCBI,NIH,NLM,National Center for Biotechnology Information,National Institutes of Health,National Library of Medicine,Non-P.H.S.,PMC6057684,PubMed Abstract,Research Support,Type 2 / blood,Type 2 / diagnosis,U.S. Gov't,doi:10.1371/journal.pbio.2005143,pmid:30040822},
   month = {7},
   pmid = {30040822},
   publisher = {PLoS Biol},
   title = {Glucotypes reveal new patterns of glucose dysregulation},
   volume = {16},
   url = {https://pubmed.ncbi.nlm.nih.gov/30040822/},
   year = {2018}
}
@article{D1NAMO,
   author = {Fabien Dubosson and Jean Eudes Ranvier and Stefano Bromuri and Jean Paul Calbimonte and Juan Ruiz and Michael Schumacher},
   doi = {10.1016/J.IMU.2018.09.003},
   issn = {2352-9148},
   journal = {Informatics in Medicine Unlocked},
   keywords = {Accelerometers,Annotated food pictures,Diabetes management,ECG,Glucose,Wearable devices},
   month = {1},
   pages = {92-100},
   publisher = {Elsevier},
   title = {The open D1NAMO dataset: A multi-modal dataset for research on non-invasive type 1 diabetes management},
   volume = {13},
   url = {https://www.sciencedirect.com/science/article/pii/S2352914818301059?via%3Dihub},
   year = {2018}
}
@article{Cols2019,
   author = {Ana Colás and Luis Vigil and Borja Vargas and David Cuesta-Frau and Manuel Varela},
   doi = {10.1371/JOURNAL.PONE.0225817},
   issn = {1932-6203},
   issue = {12},
   journal = {PloS one},
   keywords = {80 and over,Adolescent,Adult,Aged,Algorithms,Ana Colás,Blood Glucose / analysis*,Datasets as Topic,Diabetes Mellitus,Female,Humans,Luis Vigil,MEDLINE,Male,Manuel Varela,Middle Aged,NCBI,NIH,NLM,National Center for Biotechnology Information,National Institutes of Health,National Library of Medicine,Non-U.S. Gov't,PMC6919578,PubMed Abstract,Research Support,Risk Factors,Type 2 / diagnosis*,Young Adult,doi:10.1371/journal.pone.0225817,pmid:31851681},
   month = {12},
   pmid = {31851681},
   publisher = {PLoS One},
   title = {Detrended Fluctuation Analysis in the prediction of type 2 diabetes mellitus in patients at risk: Model optimization and comparison with other metrics},
   volume = {14},
   url = {https://pubmed.ncbi.nlm.nih.gov/31851681/},
   year = {2019}
}
@article{BIGIDEAs,
   abstract = {Prediabetes affects one in three people and has a 10% annual conversion rate to type 2 diabetes without lifestyle or medical interventions. Management of glycemic health is essential to prevent progression to type 2 diabetes. However, there is currently no commercially-available and noninvasive method for monitoring glycemic health to aid in self-management of prediabetes. There is a critical need for innovative, practical strategies to improve monitoring and management of glycemic health. In this study, using a dataset of 25,000 simultaneous interstitial glucose and noninvasive wearable smartwatch measurements, we demonstrated the feasibility of using noninvasive and widely accessible methods, including smartwatches and food logs recorded over 10 days, to continuously detect personalized glucose deviations and to predict the exact interstitial glucose value in real time with up to 84% and 87% accuracy, respectively. We also establish methods for designing variables using data-driven and domain-driven methods from noninvasive wearables toward interstitial glucose prediction.},
   author = {Brinnae Bent and Peter J. Cho and Maria Henriquez and April Wittmann and Connie Thacker and Mark Feinglos and Matthew J. Crowley and Jessilyn P. Dunn},
   doi = {10.1038/s41746-021-00465-w},
   issn = {2398-6352},
   issue = {1},
   journal = {npj Digital Medicine 2021 4:1},
   keywords = {Biomarkers,Biomedical engineering,Pre,diabetes},
   month = {6},
   pages = {89-},
   publisher = {Nature Publishing Group},
   title = {Engineering digital biomarkers of interstitial glucose from noninvasive smartwatches},
   volume = {4},
   url = {https://www.nature.com/articles/s41746-021-00465-w},
   year = {2021}
}
@article{ShanghaiTDM,
   author = {Qinpei Zhao and Jinhao Zhu and Xuan Shen and Chuwen Lin and Yinjia Zhang and Yuxiang Liang and Baige Cao and Jiangfeng Li and Xiang Liu and Weixiong Rao and Congrong Wang},
   doi = {10.1038/s41597-023-01940-7},
   issn = {2052-4463},
   issue = {1},
   journal = {Scientific Data 2023 10:1},
   keywords = {Data mining,Machine learning,Metabolomics,Public health},
   month = {1},
   pages = {35-},
   pmid = {36653358},
   publisher = {Nature Publishing Group},
   title = {Chinese diabetes datasets for data-driven machine learning},
   volume = {10},
   url = {https://www.nature.com/articles/s41597-023-01940-7},
   year = {2023}
}
@article{Langarica2024,
   author = {Saúl Langarica and Diego De La Vega and Nawel Cariman and Martín Miranda and David C. Andrade and Felipe Núñez and Maria Rodriguez-Fernandez},
   doi = {10.1109/OJEMB.2024.3365290},
   issn = {26441276},
   journal = {IEEE Open Journal of Engineering in Medicine and Biology},
   keywords = {Diabetes,Glucose prediction,deep learning,transfer learning},
   pages = {467-475},
   publisher = {Institute of Electrical and Electronics Engineers Inc.},
   title = {Deep Learning-Based Glucose Prediction Models: A Guide for Practitioners and a Curated Dataset for Improved Diabetes Management},
   volume = {5},
   url = {https://ieeexplore.ieee.org/document/10433750},
   year = {2024}
}
@article{HUPA-UCM,
   author = {J. Ignacio Hidalgo and Jorge Alvarado and Marta Botella and Aranzazu Aramendi and J. Manuel Velasco and Oscar Garnica},
   doi = {10.1016/J.DIB.2024.110559},
   issn = {2352-3409},
   journal = {Data in Brief},
   keywords = {Diabetes,Glucose prediction,Machine learning,T1DM},
   month = {8},
   pages = {110559},
   publisher = {Elsevier},
   title = {HUPA-UCM diabetes dataset},
   volume = {55},
   url = {https://www.sciencedirect.com/science/article/pii/S2352340924005262?via%3Dihub},
   year = {2024}
}
@article{CGMacros,
   author = {Anurag Das and David Kerr and Namino Glantz and Wendy Bevier and Rony Santiago and Ricardo Gutierrez-Osuna and Bobak J. Mortazavi},
   doi = {10.1038/s41597-025-05851-7},
   issn = {2052-4463},
   issue = {1},
   journal = {Scientific Data 2025 12:1},
   keywords = {Biomedical engineering,Metabolic disorders},
   month = {9},
   pages = {1557-},
   pmid = {40998842},
   publisher = {Nature Publishing Group},
   title = {CGMacros: a pilot scientific dataset for personalized nutrition and diet monitoring},
   volume = {12},
   url = {https://www.nature.com/articles/s41597-025-05851-7},
   year = {2025}
}
@article{T1DM-UOM,
   author = {Ashwaq Alsuhaymi and Ahmad Bilal and Daniel Gasca García and Rujiravee Kongdee and Nicole Lubasinski and Hood Thabit and Paul W. Nutter and Simon Harper},
   doi = {10.1038/s41597-025-05695-1},
   issn = {2052-4463},
   issue = {1},
   journal = {Scientific Data 2025 12:1},
   keywords = {Diabetes complications,Type 1 diabetes},
   month = {8},
   pages = {1379-},
   pmid = {40775218},
   publisher = {Nature Publishing Group},
   title = {A Longitudinal Multimodal Dataset of Type 1 Diabetes},
   volume = {12},
   url = {https://www.nature.com/articles/s41597-025-05695-1},
   year = {2025}
}
@article{BrisT1D,
   abstract = {Background: Type 1 diabetes (T1D) has seen a rapid evolution in management technology and forms a useful case study for the future management of other chronic conditions. Further development of this management technology requires an exploration of its real-world use and the potential of additional data streams. To facilitate this, we contribute the BrisT1D Dataset to the growing number of public T1D management datasets. The dataset was developed from a longitudinal study of 24 young adults in the UK who used a smartwatch alongside their usual T1D management. Findings: The BrisT1D dataset features both device data from the T1D management systems and smartwatches used by participants, as well as transcripts of monthly interviews and focus groups conducted during the study. The device data is provided in a processed state, for usability and more rapid analysis, and in a raw state, for in-depth exploration of novel insights captured in the study. Conclusions: This dataset has a range of potential applications. The quantitative elements can support blood glucose prediction, hypoglycaemia prediction, and closed-loop algorithm development. The qualitative elements enable the exploration of user experiences and opinions, as well as broader mixed-methods research into the role of smartwatches in T1D management.},
   author = {Sam Gordon James and Miranda Elaine and Glynis Armstrong and Aisling Ann O'kane and Harry Emerson and Zahraa S Abdallah},
   month = {5},
   title = {BrisT1D Dataset: Young Adults with Type 1 Diabetes in the UK using Smartwatches},
   url = {https://arxiv.org/pdf/2507.17757},
   year = {2025}
}
@article{AZT1D,
   abstract = {High quality real world datasets are essential for advancing data driven approaches in type 1 diabetes (T1D) management, including personalized therapy design, digital twin systems, and glucose prediction models. However, progress in this area has been limited by the scarcity of publicly available datasets that offer detailed and comprehensive patient data. To address this gap, we present AZT1D, a dataset containing data collected from 25 individuals with T1D on automated insulin delivery (AID) systems. AZT1D includes continuous glucose monitoring (CGM) data, insulin pump and insulin administration data, carbohydrate intake, and device mode (regular, sleep, and exercise) obtained over 6 to 8 weeks for each patient. Notably, the dataset provides granular details on bolus insulin delivery (i.e., total dose, bolus type, correction specific amounts) features that are rarely found in existing datasets. By offering rich, naturalistic data, AZT1D supports a wide range of artificial intelligence and machine learning applications aimed at improving clinical decision making and individualized care in T1D.},
   author = {Saman Khamesian and Asiful Arefeen and Bithika M. Thompson and Maria Adela Grando and Hassan Ghasemzadeh},
   doi = {10.17632/gk9m674wcx.1},
   keywords = {cs.LG},
   month = {5},
   title = {AZT1D: A Real-World Dataset for Type 1 Diabetes},
   url = {http://arxiv.org/abs/2506.14789 http://dx.doi.org/10.17632/gk9m674wcx.1},
   year = {2025}
}