Prediction of Hypoglycemia From Continuous Glucose Monitoring in Insulin-Treated Patients With Type 2 Diabetes Using Transfer Learning on Type 1 Diabetes Data: A Deep Transfer Learning Approach
Authors' conclusion
The transfer learned CNN model showed promising performance in predicting hypoglycemic episodes and with slightly better results than a non-transfer learned CNN model.
Study context
- Population
- Insulin-treated patients with T2D; Continuous glucose monitoring (CGM) data from 226 patients with T1D and 180 patients with T2D334,711 CGM samples validated
- Outcome
- Prediction of hypoglycemia (three consecutive CGM values below 3.9 mmol/L (70 mg/dL)) one hour after a one-hour CGM sample, AUC
- Follow-up
- One hour after the sample
Methodological detailsdeep transfer learning, convolutional neural network pre-trained on T1D data
- Deep transfer learning
- Convolutional neural network pre-trained on T1D data
Performance
As reported in the abstract
Prediction of hypoglycaemia (3 consecutive CGM values <3.9 mmol/L) 1 h after sample; external validation on separate T2D data set · 1 h
- External validation one-hour CGM samples
- 334 711
- Samples labeled hypoglycemic
- 15 695 (4.69%)
- AUC
- 0.941
- PPV at specificity of 95%
- 40.49%
- Sensitivity at specificity of 95%
- 69.16%
Further findings
Comparison with non-transfer learned convolutional neural network (CNN)
- Slightly better results than a non-transfer learned CNN model
Original publication
Prediction of Hypoglycemia From Continuous Glucose Monitoring in Insulin-Treated Patients With Type 2 Diabetes Using Transfer Learning on Type 1 Diabetes Data: A Deep Transfer Learning Approach. J Diabetes Sci Technol. Published 28 November 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →