CoChat

Self-Supervised Contrastive Learning for Medical Time Series: A Systematic Review

Open access

The DOI matches its Crossref record; 128 of the 131 references it cites were found in a registry or an index, and 3 are not checked yet.

Abstract

Medical time series are sequential data collected over time that measures health-related signals, such as electroencephalography (EEG), electrocardiography (ECG), and intensive care unit (ICU) readings. Analyzing medical time series and identifying the latent patterns and trends that lead to uncovering highly valuable insights for enhancing diagnosis, treatment, risk assessment, and disease progression. However, data mining in medical time series is heavily limited by the sample annotation which is time-consuming and labor-intensive, and expert-depending. To mitigate this challenge, the emerging self-supervised contrastive learning, which has shown great success since 2020, is a promising solution. Contrastive learning aims to learn representative embeddings by contrasting positive and negative samples without the requirement for explicit labels. Here, we conducted a systematic review of how contrastive learning alleviates the label scarcity in medical time series based on PRISMA standards. We searched the studies in five scientific databases (IEEE, ACM, Scopus, Google Scholar, and PubMed) and retrieved 1908 papers based on the inclusion criteria. After applying excluding criteria, and screening at title, abstract, and full text levels, we carefully reviewed 43 papers in this area. Specifically, this paper outlines the pipeline of contrastive learning, including pre-training, fine-tuning, and testing. We provide a comprehensive summary of the various augmentations applied to medical time series data, the architectures of pre-training encoders, the types of fine-tuning classifiers and clusters, and the popular contrastive loss functions. Moreover, we present an overview of the different data types used in medical time series, highlight the medical applications of interest, and provide a comprehensive table of 51 public datasets that have been utilized in this field. In addition, this paper will provide a discussion on the promising future scopes such as providing guidance for effective augmentation design, developing a unified framework for analyzing hierarchical time series, and investigating methods for processing multimodal data. Despite being in its early stages, self-supervised contrastive learning has shown great potential in overcoming the need for expert-created annotations in the research of medical time series.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.3390/s23094221

Cites

128 of 131 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Pipeline of self-supervised contrastive learning which is composed of three stages. (a) The pre-training receives unlabelled time series sample xi as anchor sample, the augmented sample xi^(′) as positive sample (Section 3.4) while a different sample xj as negative sample. The hi, hi^(′) and hj denotes learned embedding of the original sample xi, positive pair xi^(′), and negative pair xj, respectively. A contrastive loss (Section 3.7) is calculated based on the distance among embeddings of samples, which is used to update the encoder through backpropagation. (b) The well-trained encoder will be inherited by the fine-tuning stage which receives a labeled sample and makes a prediction through a downstream classifier. A standard supervised loss function (e.g., cross-entropy) will be used to update the encoder and/or classifier. (c) The testing stage makes predictions from the learned embedding htest of an unseen test sample xtest.
Figure 1 Pipeline of self-supervised contrastive learning which is composed of three stages. (a) The pre-training receives unlabelled time series sample xi as anchor sample, the augmented sample xi^(′) as positive sample (Section 3.4) while a different sample xj as negative sample. The hi, hi^(′) and hj denotes learned embedding of the original sample xi, positive pair xi^(′), and negative pair xj, respectively. A contrastive loss (Section 3.7) is calculated based on the distance among embeddings of samples, which is used to update the encoder through backpropagation. (b) The well-trained encoder will be inherited by the fine-tuning stage which receives a labeled sample and makes a prediction through a downstream classifier. A standard supervised loss function (e.g., cross-entropy) will be used to update the encoder and/or classifier. (c) The testing stage makes predictions from the learned embedding htest of an unseen test sample xtest.CC BY · PubMed Central
Figure 2 PRISMA diagram of the literature review process. We retrieved 43 publications among the 2102 papers that are collected from the five academic platforms.
Figure 2 PRISMA diagram of the literature review process. We retrieved 43 publications among the 2102 papers that are collected from the five academic platforms.CC BY · PubMed Central
Figure 3 Types of medical time series in the reviewed papers. The majority of studies have focused on EEG, ECG, and ICU data, and one potential reason for this trend is the availability of large-scale public datasets in these fields. In contrast, other physiological signals may not have as many large-scale datasets available, making it more challenging to develop and validate machine learning models using those signals.
Figure 3 Types of medical time series in the reviewed papers. The majority of studies have focused on EEG, ECG, and ICU data, and one potential reason for this trend is the availability of large-scale public datasets in these fields. In contrast, other physiological signals may not have as many large-scale datasets available, making it more challenging to develop and validate machine learning models using those signals.CC BY · PubMed Central
Figure 4 Applications of the reviewed papers. Consistent with the distribution of data types, the healthcare applications identified in this review predominantly focus on cardiovascular disease detection, sleep status monitoring, ICU-related scenarios, and neurological disorder diagnosis.
Figure 4 Applications of the reviewed papers. Consistent with the distribution of data types, the healthcare applications identified in this review predominantly focus on cardiovascular disease detection, sleep status monitoring, ICU-related scenarios, and neurological disorder diagnosis.CC BY · PubMed Central
Figure 5 Visualization of the commonly-used augmentations for time series. In each subfigure, we present both the original sample and the augmented sample. Detailed descriptions in Section 3.4.
Figure 5 Visualization of the commonly-used augmentations for time series. In each subfigure, we present both the original sample and the augmented sample. Detailed descriptions in Section 3.4.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 131 references

Research works printed but not linked: 14

  1. Spathis, D., Perez-Pozuelo, I., Brage, S., Wareham, N.J., and Mascolo, C. (2020). Learning generalizable physiological representations from large-scale wearable data. arXiv.
  2. Pan, L., Feng, Z., and Peng, S. (2022). A review of machine learning approaches, challenges and prospects for computational tumor pathology. arXiv.
  3. Kingma, D.P., and Welling, M. (2013). Auto-encoding variational bayes. arXiv.
  4. Wagh, N., Wei, J., Rawal, S., Berry, B., Barnard, L., Brinkmann, B., Worrell, G., Jones, D., and Varatharajah, Y. (2021, January 6–7). Domain-guided Self-supervision of EEG Data Improves Downstream Classification Performance and Generalizability. Proceedings of the Machine Learning for Health, PMLR, Virtual.
  5. Ho, T.K.K., and Armanfard, N. (2022). Self-Supervised Learning for Anomalous Channel Detection in EEG Graphs: Application to Seizure Analysis. arXiv.
  6. Yang, C., Xiao, D., Westover, M.B., and Sun, J. (2021). Self-supervised eeg representation learning for automatic sleep staging. arXiv.
  7. Cheng, J.Y., Goh, H., Dogrusoz, K., Tuzel, O., and Azemi, E. (2020). Subject-aware contrastive learning for biosignals. arXiv.
  8. Huijben, I.A., Nijdam, A.A., Overeem, S., van Gilst, M.M., and van Sloun, R.J. (2022). SOM-CPC: Unsupervised Contrastive Learning with Self-Organizing Maps for Structured Representations of High-Rate Time Series. arXiv.
  9. Zhang, X., Zhao, Z., Tsiligkaridis, T., and Zitnik, M. (2022). Self-supervised contrastive pre-training for time series via time-frequency consistency. arXiv.
  10. Weatherhead, A., Greer, R., Moga, M.A., Mazwi, M., Eytan, D., Goldenberg, A., and Tonekaboni, S. (2022, January 7–8). Learning Unsupervised Representations for ICU Timeseries. Proceedings of the Conference on Health, Inference, and Learning, PMLR, Virtual.
  11. Edinburgh, T., Smielewski, P., Czosnyka, M., Cabeleira, M., Eglen, S.J., and Ercole, A. (2021). Intracranial Pressure and Neuromonitoring XVII, Springer.
  12. Wever, F., Keller, T.A., Symul, L., and Garcia, V. (2021). As easy as APC: Overcoming missing data and class imbalance in time series with self-supervised learning. arXiv.
  13. van den Oord, A., Li, Y., and Vinyals, O. (2018). Representation learning with contrastive predictive coding. arXiv.
  14. Rocha, B., Filos, D., Mendes, L., Vogiatzis, I., Perantoni, E., Kaimakamis, E., Natsiavas, P., Oliveira, A., Jácome, C., and Marques, A. A respiratory sound database for the development of automated classification. Proceedings of the International Conference on Biomedical and Health Informatics.

Cited by

See all 108

Metadata

Authors
Liu, Z., Alavi, A., Li, M., & Zhang, X.
Affiliations
  • RMIT University (AU)
  • Coles, Melbourne, VIC 3123, Australia
  • University of North Carolina at Charlotte (US)
Corresponding author
  • Alavi, Azadeh
  • Zhang, Xiang
Venue
Sensors
Publisher
MDPI AG
Published
2023-04-23
Volume
23
Issue
9
Pages
4221
ISSN
1424-8220 (ISSN-L)
OpenAlex
W4366829097
Cited by
90
References
128
Citations by year
2026: 35 · 2025: 26 · 2024: 25 · 2023: 4
FWCI
1.88
Type
journal article
Language
en
Open access
Locations
  • Sensors (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Figshare (submittedVersion, cc-by)
Funders
National Science Foundation (2245894)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
  • Physical Sciences › Computer Science › Signal Processing › Time Series Analysis and Forecasting
  • Health Sciences › Medicine › Cardiology and Cardiovascular Medicine › ECG Monitoring and Analysis
Keywords
self-supervised contrastive learning, medical time series, limited labeled data, data augmentation, contrastive loss, EEG, ECG, intensive care monitoring, hierarchical time series, multimodal data, medical datasets
MeSH terms
Learning, Time Factors, Databases, Factual, Data Mining

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Self-Supervised Contrastive Learning for Medical Time Series: A Systematic Review. https://app.cochat.ai/papers/10.3390/s23094221
The paper
Liu, Z., Alavi, A., Li, M., & Zhang, X. (2023). Self-Supervised Contrastive Learning for Medical Time Series: A Systematic Review. Sensors. https://doi.org/10.3390/s23094221

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.