Detection of Meals and Physical Activity Events From Free-Living Data of People With Diabetes
Authors' conclusion
The meal and exercise times detected by the RNN models can be used to warn people for entering meal and exercise information to hybrid closed-loop automated insulin delivery systems. Reliable accuracy for event detection necessitates powerful ML and large data sets. The use of additional sensors and algorithms for detecting these events and their characteristics provides a more accurate alternative.
Study context
- Population
- People with diabetes; free-living continuous glucose monitoring (CGM) data, pump insulin data and manual meal and exercise entries
- Outcome
- Detection accuracy of meal and exercise events
Methodological detailsmachine learning, recurrent neural networks, original and imputed data sets
- Machine learning
- Recurrent neural networks
- Original and imputed data sets
Performance
As reported in the abstract
Accuracy for identifying state of event:
- Meal
- Exercise
- Both
- Neither
| Test or model | Accuracy, no meal or exercise | Detection accuracy, meals | Detection accuracy, exercise | Detection accuracy, concurrent meal-exercise |
|---|---|---|---|---|
| LSTM with 1D Convolution | 94.58% | 98.05% | 93.42% | 55.56% |
LSTM with 1D Convolution
Accuracy, no meal or exercise94.58%
Detection accuracy, meals98.05%
Detection accuracy, exercise93.42%
Detection accuracy, concurrent meal-exercise55.56%
- Accuracy of RNN models for identifying state of event, various users
- 89.9% to 95.7%
Original publication
Detection of Meals and Physical Activity Events From Free-Living Data of People With Diabetes. J Diabetes Sci Technol. Published 15 June 2022. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →