Fig. 6

A Personalizing a digital twin using data from one person to train and validate a passive digital twin, such as the one presented herein, and making the digital twin active. This personalization can be in the form of input parameters, such as age and height, parameters estimated and validated on time-series data, such as meal response glucose, and input parameters representing activities, such as energy intake or topiramate dosage. B Using the digital twin to predict and compare scenarios with different lifestyles and/or treatments. In this example, the digital twin is used to predict two scenarios. In scenario 1, the digital twin simulates an increase in energy intake for 40 years (from 40 to 80 years of age) and a resulting increase in BMI—from overweight to obese levels (BMI over 25 and 30 kg/m², respectively)—and an increase in fasting plasma glucose—from prediabetic to diabetic levels (fasting glucose above 5.6 and 7 mmol/l, respectively). In Scenario 2, the digital twin simulates a decrease in energy intake with a weight-loss drug such as topiramate, resulting in a decrease to healthy levels of BMI and fasting plasma glucose. C Following the chosen lifestyle and getting continuous feedback by zooming in on 4 weeks of the predicted fasting plasma glucose (solid line) and comparing with data (blue squares) collected by the user. Zooming in even more and looking at meal response glucose before and after the 4 weeks, one can see that the glucose curve is higher before (left box) compared to after (right box), indicating an improvement in meal response glucose levels as well