Transfer Learning for Pediatric Glucose Forecasting

Alain Ryser, Chuhao Feng, Tobias Scheithauer, Marc Pfister, Marie-Anne Burckhardt, Sara Bachmann, Alexander Marx, Julia E. Vogt
Proceedings of the 4th Machine Learning for Health Symposium, PMLR 259:839-860, 2025.

Abstract

Effective blood glucose forecasting is crucial for detecting events such as hypo- or hyperglycemia in people with diabetes, yet remains challenging in domains with only small, heterogeneous datasets, such as in the pediatric field. We present GluTFT, a novel transfer learning approach that allows leveraging models pretrained on publicly available adult diabetes datasets for pediatric glucose forecasting. We systematically evaluate multiple transfer learning strategies, including zero-shot prediction and fine-tuning across the entire dataset as well as specific subgroups of participants. Our extensive experiments reveal that GluTFT excels on the pretraining datasets and significantly outperforms baseline methods when fine-tuned. To validate the clinical relevance of our approach, we evaluate Parkes Error Grids, demonstrating the quality of GluTFT’s blood glucose forecasts and its potential for enhancing clinical decision-making for pediatric diabetes.

Cite this Paper


BibTeX
@InProceedings{pmlr-v259-ryser25a, title = {Transfer Learning for Pediatric Glucose Forecasting}, author = {Ryser, Alain and Feng, Chuhao and Scheithauer, Tobias and Pfister, Marc and Burckhardt, Marie-Anne and Bachmann, Sara and Marx, Alexander and Vogt, Julia E.}, booktitle = {Proceedings of the 4th Machine Learning for Health Symposium}, pages = {839--860}, year = {2025}, editor = {Hegselmann, Stefan and Zhou, Helen and Healey, Elizabeth and Chang, Trenton and Ellington, Caleb and Mhasawade, Vishwali and Tonekaboni, Sana and Argaw, Peniel and Zhang, Haoran}, volume = {259}, series = {Proceedings of Machine Learning Research}, month = {15--16 Dec}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v259/main/assets/ryser25a/ryser25a.pdf}, url = {https://proceedings.mlr.press/v259/ryser25a.html}, abstract = {Effective blood glucose forecasting is crucial for detecting events such as hypo- or hyperglycemia in people with diabetes, yet remains challenging in domains with only small, heterogeneous datasets, such as in the pediatric field. We present GluTFT, a novel transfer learning approach that allows leveraging models pretrained on publicly available adult diabetes datasets for pediatric glucose forecasting. We systematically evaluate multiple transfer learning strategies, including zero-shot prediction and fine-tuning across the entire dataset as well as specific subgroups of participants. Our extensive experiments reveal that GluTFT excels on the pretraining datasets and significantly outperforms baseline methods when fine-tuned. To validate the clinical relevance of our approach, we evaluate Parkes Error Grids, demonstrating the quality of GluTFT’s blood glucose forecasts and its potential for enhancing clinical decision-making for pediatric diabetes.} }
Endnote
%0 Conference Paper %T Transfer Learning for Pediatric Glucose Forecasting %A Alain Ryser %A Chuhao Feng %A Tobias Scheithauer %A Marc Pfister %A Marie-Anne Burckhardt %A Sara Bachmann %A Alexander Marx %A Julia E. Vogt %B Proceedings of the 4th Machine Learning for Health Symposium %C Proceedings of Machine Learning Research %D 2025 %E Stefan Hegselmann %E Helen Zhou %E Elizabeth Healey %E Trenton Chang %E Caleb Ellington %E Vishwali Mhasawade %E Sana Tonekaboni %E Peniel Argaw %E Haoran Zhang %F pmlr-v259-ryser25a %I PMLR %P 839--860 %U https://proceedings.mlr.press/v259/ryser25a.html %V 259 %X Effective blood glucose forecasting is crucial for detecting events such as hypo- or hyperglycemia in people with diabetes, yet remains challenging in domains with only small, heterogeneous datasets, such as in the pediatric field. We present GluTFT, a novel transfer learning approach that allows leveraging models pretrained on publicly available adult diabetes datasets for pediatric glucose forecasting. We systematically evaluate multiple transfer learning strategies, including zero-shot prediction and fine-tuning across the entire dataset as well as specific subgroups of participants. Our extensive experiments reveal that GluTFT excels on the pretraining datasets and significantly outperforms baseline methods when fine-tuned. To validate the clinical relevance of our approach, we evaluate Parkes Error Grids, demonstrating the quality of GluTFT’s blood glucose forecasts and its potential for enhancing clinical decision-making for pediatric diabetes.
APA
Ryser, A., Feng, C., Scheithauer, T., Pfister, M., Burckhardt, M., Bachmann, S., Marx, A. & Vogt, J.E.. (2025). Transfer Learning for Pediatric Glucose Forecasting. Proceedings of the 4th Machine Learning for Health Symposium, in Proceedings of Machine Learning Research 259:839-860 Available from https://proceedings.mlr.press/v259/ryser25a.html.

Related Material