EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Interpretable non-invasive glucose monitoring: an attention-based deep learning framework for visualizing hemodynamic correlates in PPG signals

Rizal Syamsul, Kim Dong-Seong

Frontiers in Bioengineering and Biotechnology · 03.07.2026 · 14

DOI: 10.3389/fbioe.2026.1843361 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Frontiers in Bioengineering and Biotechnology, online erschienen am 03.07.2026. Verfasst von Rizal Syamsul und 1 weiteren Autor:innen. Beteiligte Einrichtung: ICT-CRC, Kumoh National Institute of Technology, Gumi, Republic of Korea. Die Arbeit nennt 15 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Diabetes mellitus necessitates frequent blood glucose monitoring, yet invasive finger-prick methods remain a barrier to adherence. Although Photoplethysmography (PPG) offers a promising noninvasive alternative, current deep learning approaches largely operate as “black boxes,” obscuring the physiological features driving their predictions and limiting clinical trust.

Methods

To address this interpretability gap, we propose a novel Attention - Guided Convolutional - Recurrent Neural Network (AG-CRNN). Unlike standard deep learning models that treat the entire signal uniformly, our architecture integrates a Temporal Attention Mechanism that dynamically assigns importance weights to specific segments of the PPG waveform. This allows the model to learn and emphasize hemodynamic fluctuations associated with glucose variability, while suppressing motion artifacts. We validated the framework using a dataset of synchronized PPG and glucose measurements from 23 subjects.

Results

The proposed attention-based model achieved a Mean Absolute Error (MAE) of 11.59 mg/dL, significantly outperforming standard LSTM and GRU baselines. Crucially, our Explainable AI (XAI) analysis revealed that the model preferentially focuses on the systolic decay and diastolic decay regions. This visualization aligns with the established physiological evidence linking hyperglycemia to vascular stiffness and wave reflection timing.

Conclusion

This study presents a high-accuracy glucose estimation framework that provides visual explanations of its decision-making process. By demonstrating that the model leverages genuine morphological correlates rather than spurious noise, we bridged the gap between AI performance and clinical plausibility.

Schlagwörter

Stichwörter der Autor:innen

blood glucose estimationdeep learninggated recurrent unit (GRU)long short-term memory (LSTM)noninvasive monitoringphotoplethysmography (PPG)recurrent neural networks (RNN)

Autor:innen und Institutionen

  • ICT-CRC, Kumoh National Institute of Technology, Gumi, Republic of Korea.
  • School of Electrical Engineering, Telkom University, Bandung, Indonesia.
  • NSLab Co. Ltd, Gumi, Republic of Korea.

Zitieren

APA

Syamsul, R., & Dong-Seong, K. (2026). Interpretable non-invasive glucose monitoring: an attention-based deep learning framework for visualizing hemodynamic correlates in PPG signals. Frontiers in Bioengineering and Biotechnology, 14. https://doi.org/10.3389/fbioe.2026.1843361

Vancouver

Syamsul R, Dong-Seong K. Interpretable non-invasive glucose monitoring: an attention-based deep learning framework for visualizing hemodynamic correlates in PPG signals. Frontiers in Bioengineering and Biotechnology. 2026;14. doi:10.3389/fbioe.2026.1843361

BibTeX
@article{Syamsul2026,
  author = {Syamsul, Rizal and Dong-Seong, Kim},
  title = {Interpretable non-invasive glucose monitoring: an attention-based deep learning framework for visualizing hemodynamic correlates in PPG signals},
  journal = {Frontiers in Bioengineering and Biotechnology},
  year = {2026},
  volume = {14},
  doi = {10.3389/fbioe.2026.1843361}
}
Kostenlos starten →