Blood Glucose Forecasting Algorithm
Our new forecasting algorithm uses Deep Learning to anticipate blood glucose (BG) levels in People with Diabetes. We explain below the results achieved through our analysis of a publicly available dataset of People with Diabetes.
Dataset
The OhioT1DM Dataset was developed for blood glucose level prediction research. The dataset consists of 8 weeks of continuous glucose monitoring via Medtronic sensors and self-reported life-event data for 12 people with Type 1 Diabetes. The patient demographic was five females aged 20-60, and seven males aged 20-80.
Accuracy, Sensitivity and Specificity of BG Forecasting
We verified the model’s performance using sensitivity (true positive rate), specificity (true negative rate) and accuracy metrics. The table below catalogs our results.
Dataset
OhioT1DM
Number of Patients
12
Sensitivity
(%)
83.6
Specificity
(%)
96.1
Accuracy
(%)
93.6
To illustrate the accuracy of our algorithm, conveyed in the graph below are the blood glucose levels anticipated by our ML algorithm overlaid with the actual BG values of the People with Diabetes in the dataset.
Our Vision
This breakthrough has been achieved through Bio Conscious Technologies' innovative use of ML algorithms, which accurately anticipate high-risk diabetic events CGM data. We believe this breakthrough represents a significant advancement of Diabetes technology and has the potential to revolutionize the way professionals monitor and treat People with Diabetes. Our vision is to achieve modularity through ML model chaining to provide multiple decision support solutions.
