Interpreting Deep Glucose Predictive Models for Diabetic People Using RETAIN

Maxime De Bois; Mounîm A. El Yacoubi; Mehdi Ammi

doi:10.1007/978-3-030-59830-3_59

Interpreting Deep Glucose Predictive Models for Diabetic People Using RETAIN

Maxime De Bois
, Mounîm A. El Yacoubi
, Mehdi Ammi

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Progress in the biomedical field through the use of deep learning is hindered by the lack of interpretability of the models. In this paper, we study the RETAIN architecture for the forecasting of future glucose values for diabetic people. Thanks to its two-level attention mechanism, the RETAIN model is interpretable while remaining as efficient as standard neural networks. We evaluate the model on a real-world type-2 diabetic population and we compare it to a random forest model and a LSTM-based recurrent neural network. Our results show that the RETAIN model outperforms the former and equals the latter on common accuracy metrics and clinical acceptability metrics, thereby proving its legitimacy in the context of glucose level forecasting. Furthermore, we propose tools to take advantage of the RETAIN interpretable nature. As informative for the patients as for the practitioners, it can enhance the understanding of the predictions made by the model and improve the design of future glucose predictive models.

Original language	English
Title of host publication	Pattern Recognition and Artificial Intelligence - International Conference, ICPRAI 2020, Proceedings
Editors	Yue Lu, Nicole Vincent, Pong Chi Yuen, Wei-Shi Zheng, Farida Cheriet, Ching Y. Suen
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	685-694
Number of pages	10
ISBN (Print)	9783030598297
DOIs	https://doi.org/10.1007/978-3-030-59830-3_59
Publication status	Published - 1 Jan 2020
Event	2nd International Conference on Pattern Recognition and Artificial Intelligence, ICPRAI 2020 - Zhongshan, China Duration: 19 Oct 2020 → 23 Oct 2020

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	12068 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	2nd International Conference on Pattern Recognition and Artificial Intelligence, ICPRAI 2020
Country/Territory	China
City	Zhongshan
Period	19/10/20 → 23/10/20

Keywords

Attention
Deep learning
Diabetes
Glucose prediction
Interpretability
Neural networks

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1007/978-3-030-59830-3_59

Cite this

De Bois, M., El Yacoubi, M. A., & Ammi, M. (2020). Interpreting Deep Glucose Predictive Models for Diabetic People Using RETAIN. In Y. Lu, N. Vincent, P. C. Yuen, W.-S. Zheng, F. Cheriet, & C. Y. Suen (Eds.), Pattern Recognition and Artificial Intelligence - International Conference, ICPRAI 2020, Proceedings (pp. 685-694). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12068 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-030-59830-3_59

De Bois, Maxime ; El Yacoubi, Mounîm A. ; Ammi, Mehdi. / Interpreting Deep Glucose Predictive Models for Diabetic People Using RETAIN. Pattern Recognition and Artificial Intelligence - International Conference, ICPRAI 2020, Proceedings. editor / Yue Lu ; Nicole Vincent ; Pong Chi Yuen ; Wei-Shi Zheng ; Farida Cheriet ; Ching Y. Suen. Springer Science and Business Media Deutschland GmbH, 2020. pp. 685-694 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{cb6b92246f304b7ca73ce2b01fbaf25e,

title = "Interpreting Deep Glucose Predictive Models for Diabetic People Using RETAIN",

abstract = "Progress in the biomedical field through the use of deep learning is hindered by the lack of interpretability of the models. In this paper, we study the RETAIN architecture for the forecasting of future glucose values for diabetic people. Thanks to its two-level attention mechanism, the RETAIN model is interpretable while remaining as efficient as standard neural networks. We evaluate the model on a real-world type-2 diabetic population and we compare it to a random forest model and a LSTM-based recurrent neural network. Our results show that the RETAIN model outperforms the former and equals the latter on common accuracy metrics and clinical acceptability metrics, thereby proving its legitimacy in the context of glucose level forecasting. Furthermore, we propose tools to take advantage of the RETAIN interpretable nature. As informative for the patients as for the practitioners, it can enhance the understanding of the predictions made by the model and improve the design of future glucose predictive models.",

keywords = "Attention, Deep learning, Diabetes, Glucose prediction, Interpretability, Neural networks",

author = "\{De Bois\}, Maxime and \{El Yacoubi\}, \{Moun{\^i}m A.\} and Mehdi Ammi",

note = "Publisher Copyright: {\textcopyright} 2020, Springer Nature Switzerland AG.; 2nd International Conference on Pattern Recognition and Artificial Intelligence, ICPRAI 2020 ; Conference date: 19-10-2020 Through 23-10-2020",

year = "2020",

month = jan,

day = "1",

doi = "10.1007/978-3-030-59830-3\_59",

language = "English",

isbn = "9783030598297",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "685--694",

editor = "Yue Lu and Nicole Vincent and Yuen, \{Pong Chi\} and Wei-Shi Zheng and Farida Cheriet and Suen, \{Ching Y.\}",

booktitle = "Pattern Recognition and Artificial Intelligence - International Conference, ICPRAI 2020, Proceedings",

}

De Bois, M, El Yacoubi, MA & Ammi, M 2020, Interpreting Deep Glucose Predictive Models for Diabetic People Using RETAIN. in Y Lu, N Vincent, PC Yuen, W-S Zheng, F Cheriet & CY Suen (eds), Pattern Recognition and Artificial Intelligence - International Conference, ICPRAI 2020, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12068 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 685-694, 2nd International Conference on Pattern Recognition and Artificial Intelligence, ICPRAI 2020, Zhongshan, China, 19/10/20. https://doi.org/10.1007/978-3-030-59830-3_59

Interpreting Deep Glucose Predictive Models for Diabetic People Using RETAIN. / De Bois, Maxime; El Yacoubi, Mounîm A.; Ammi, Mehdi.
Pattern Recognition and Artificial Intelligence - International Conference, ICPRAI 2020, Proceedings. ed. / Yue Lu; Nicole Vincent; Pong Chi Yuen; Wei-Shi Zheng; Farida Cheriet; Ching Y. Suen. Springer Science and Business Media Deutschland GmbH, 2020. p. 685-694 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12068 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Interpreting Deep Glucose Predictive Models for Diabetic People Using RETAIN

AU - De Bois, Maxime

AU - El Yacoubi, Mounîm A.

AU - Ammi, Mehdi

PY - 2020/1/1

Y1 - 2020/1/1

N2 - Progress in the biomedical field through the use of deep learning is hindered by the lack of interpretability of the models. In this paper, we study the RETAIN architecture for the forecasting of future glucose values for diabetic people. Thanks to its two-level attention mechanism, the RETAIN model is interpretable while remaining as efficient as standard neural networks. We evaluate the model on a real-world type-2 diabetic population and we compare it to a random forest model and a LSTM-based recurrent neural network. Our results show that the RETAIN model outperforms the former and equals the latter on common accuracy metrics and clinical acceptability metrics, thereby proving its legitimacy in the context of glucose level forecasting. Furthermore, we propose tools to take advantage of the RETAIN interpretable nature. As informative for the patients as for the practitioners, it can enhance the understanding of the predictions made by the model and improve the design of future glucose predictive models.

AB - Progress in the biomedical field through the use of deep learning is hindered by the lack of interpretability of the models. In this paper, we study the RETAIN architecture for the forecasting of future glucose values for diabetic people. Thanks to its two-level attention mechanism, the RETAIN model is interpretable while remaining as efficient as standard neural networks. We evaluate the model on a real-world type-2 diabetic population and we compare it to a random forest model and a LSTM-based recurrent neural network. Our results show that the RETAIN model outperforms the former and equals the latter on common accuracy metrics and clinical acceptability metrics, thereby proving its legitimacy in the context of glucose level forecasting. Furthermore, we propose tools to take advantage of the RETAIN interpretable nature. As informative for the patients as for the practitioners, it can enhance the understanding of the predictions made by the model and improve the design of future glucose predictive models.

KW - Attention

KW - Deep learning

KW - Diabetes

KW - Glucose prediction

KW - Interpretability

KW - Neural networks

U2 - 10.1007/978-3-030-59830-3_59

DO - 10.1007/978-3-030-59830-3_59

M3 - Conference contribution

AN - SCOPUS:85092900522

SN - 9783030598297

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 685

EP - 694

BT - Pattern Recognition and Artificial Intelligence - International Conference, ICPRAI 2020, Proceedings

A2 - Lu, Yue

A2 - Vincent, Nicole

A2 - Yuen, Pong Chi

A2 - Zheng, Wei-Shi

A2 - Cheriet, Farida

A2 - Suen, Ching Y.

PB - Springer Science and Business Media Deutschland GmbH

T2 - 2nd International Conference on Pattern Recognition and Artificial Intelligence, ICPRAI 2020

Y2 - 19 October 2020 through 23 October 2020

ER -

De Bois M, El Yacoubi MA, Ammi M. Interpreting Deep Glucose Predictive Models for Diabetic People Using RETAIN. In Lu Y, Vincent N, Yuen PC, Zheng WS, Cheriet F, Suen CY, editors, Pattern Recognition and Artificial Intelligence - International Conference, ICPRAI 2020, Proceedings. Springer Science and Business Media Deutschland GmbH. 2020. p. 685-694. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-59830-3_59

Interpreting Deep Glucose Predictive Models for Diabetic People Using RETAIN

Abstract

Publication series

Conference

Keywords

UN SDGs

Access to Document

Fingerprint

Cite this