Predicting Glycaemia in Type 1 Diabetes Patients: Experiments in Feature Engineering and Data Imputation

Jeon, Jouhyun; Leimbigler, Peter; Baruah, Gaurav; Li, Michael H.; Fossat, Yan; Whitehead, Alfred J.

doi:10.1007/s41666-019-00063-2

Cited by 20 publications

(11 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While a significant body of research exists in modeling glucose homeostasis 13 , the mathematical models that currently exist for humans or animals include a multitude of variables and physiological parameters. In the case of machine learning-based models, the result is individualized to each subject 14 . These features limit their use as a diagnostic tool in clinical settings where simplicity, interpretability, and universality are paramount.…”

Section: Discussionmentioning

confidence: 99%

Homeostasis as a proportional–integral control system

Morra¹,

Palanica²,

Fossat³

2020

npj Digit. Med.

Self Cite

View full text Add to dashboard Cite

According to medical guidelines, the distinction between "healthy" and "unhealthy" patients is commonly based on single, discrete values taken at an isolated point in time (e.g., blood pressure or core temperature). Perhaps a more robust and insightful diagnosis can be obtained by studying the functional interdependence of such indicators and the homeostasis that controls them. This requires quasi-continuous measurements and a procedure to map the data onto a parsimonious control model with a degree of universality. The current research illustrates this approach using glucose homeostasis as a target. Data were obtained from 41 healthy subjects wearing over-the-counter glucose monitors, and projected onto a simple proportional-integral (PI) controller, widely used in engineering applications. The indicators quantifying the control function are clustered for the great majority of subjects, while a few outliers exhibit less responsive homeostasis. Practical implications for healthcare and education are further discussed.

show abstract

Section: Discussionmentioning

confidence: 99%

Homeostasis as a proportional–integral control system

Morra¹,

Palanica²,

Fossat³

2020

npj Digit. Med.

Self Cite

View full text Add to dashboard Cite

show abstract

“…While a significant body of research exists in modelling glucose homeostasis [14], the mathematical models that currently exist for humans or animals include a multitude of variables and physiological parameters. In the case of machine learning-based models, the result is individualised to each subject [15]. These features limit their translation to clinical settings where simplicity, interpretability, and universality are paramount.…”

Section: Discussionmentioning

confidence: 99%

I'm a doctor, not a mathematician! Homeostasis as a proportional-integral control system

Morra¹,

Palanica²,

Fossat³

2019

Preprint

Self Cite

View full text Add to dashboard Cite

The distinction between "healthy" and "unhealthy" patients is commonly based on single, discrete values taken at an isolated point in time (e.g., blood pressure or core temperature). Perhaps a more robust and insightful diagnosis can be obtained by studying the functional interdependence of such indicators and the homeostasis that controls them. This requires quasi-continuous measurements and a procedure to map the data onto a parsimonious control model with a degree of universality. The current research illustrates this approach using glucose homeostasis as a target. Data were obtained from 41 healthy subjects wearing over-the-counter glucose monitors, and projected onto a simple proportionalintegral (PI) controller, widely used in engineering applications. The indicators quantifying the control function are clustered for the great majority of subjects, while a few outliers exhibit less efficient homeostasis. Practical implications for healthcare and education are further discussed.

show abstract

“…This randomization encourages diversity within the forest, allowing the final prediction, computed as the average of the individual decisions, to be more accurate. More efficient than traditional decision trees, randomized forests are increasingly being used for the task of blood glucose prediction 23,12,22,11 . In this study, we used a forest of 100 trees.…”

Section: Reference Models Based On Decision Treesmentioning

confidence: 99%

“…This method differs from random forests where trees are created simultaneously. Like random forests, models based on gradient boosting (e.g., GBM, XGBoost) are also increasingly used in the field of blood glucose prediction : 23,12,22 . As for the DT and RF models, we have optimized the minimum number of samples required to create new branches to 250 and 2000 for the IDIAB and OhioT1DM datasets respectively.…”

Section: Reference Models Based On Decision Treesmentioning

confidence: 99%

See 1 more Smart Citation

Enhancing the Interpretability of Deep Models in Healthcare Through Attention: Application to Glucose Forecasting for Diabetic People

Bois

Yacoubi

Ammi

2021

Int. J. Patt. Recogn. Artif. Intell.

View full text Add to dashboard Cite

The adoption of deep learning in healthcare is hindered by their “black box” nature. In this paper, we explore the RETAIN architecture for the task of glucose forecasting for diabetic people. By using a two-level attention mechanism, the recurrent-neural-network-based RETAIN model is interpretable. We evaluate the RETAIN model on the type-2 IDIAB and the type-1 OhioT1DM datasets by comparing its statistical and clinical performances against two deep models and three models based on decision trees. We show that the RETAIN model offers a very good compromise between accuracy and interpretability, being almost as accurate as the LSTM and FCN models while remaining interpretable. We show the usefulness of its interpretable nature by analyzing the contribution of each variable to the final prediction. It revealed that signal values older than 1[Formula: see text]h are not used by the RETAIN model for 30[Formula: see text]min ahead of time prediction of glucose. Also, we show how the RETAIN model changes its behavior upon the arrival of an event such as carbohydrate intakes or insulin infusions. In particular, it showed that the patient’s state before the event is particularly important for the prediction. Overall the RETAIN model, thanks to its interpretability, seems to be a very promising model for regression or classification tasks in healthcare.

show abstract

Predicting Glycaemia in Type 1 Diabetes Patients: Experiments in Feature Engineering and Data Imputation

Cited by 20 publications

References 21 publications

Homeostasis as a proportional–integral control system

Homeostasis as a proportional–integral control system

I'm a doctor, not a mathematician! Homeostasis as a proportional-integral control system

Enhancing the Interpretability of Deep Models in Healthcare Through Attention: Application to Glucose Forecasting for Diabetic People

Contact Info

Product

Resources

About