*6.1.3 Insulin injection guidance*

AI can be used to provide personalized recommendations for insulin dosage besides meal planning based on glucose levels, physical activity, and other factors. The CGM sensors provide in real-time, every 1–5 minutes, the current blood glucose concentration, and its rate-of-change, two key pieces of information for improving the determination of exogenous insulin administration and the prediction of forthcoming adverse events such as hypo−/hyper-glycemia.

The most popular rtCGM sensors are minimally invasive electrochemical sensors that measure interstitial glucose concentration by a small transcutaneous electrode placed under the skin of the abdomen, or the arm. Some insulin pumps can be integrated with rtCGM sensors into the so-called closed-loop system in which a control algorithm automatically adjusts the insulin dose based on the glucose concentration measured [58]. A recent randomized controlled trial using an automated AI-based decision support system for insulin showed statistically no difference in the percentage of time spent within the target glucose range with no adverse events reported in patients on remote AI arm versus three adverse events in patients on remote adjustments by physicians' arm [59].

### *6.1.4 Glycemic adverse events detection*

Particularly, pediatric and geriatric patients are at risk of severe hypo- and hyperglycemic events. Many noninvasive techniques using AI-based algorithms are being proposed and tested to detect glycemic events. Scientists have developed an AI system that will detect hypoglycemia or low glucose through data collected *via* CGM, which is employed for detecting low glucose levels using a noninvasive wearable sensor.

Glycemic events, using ECG signals collected through noninvasive devices, are also being tested [60]. ECG-based glucose detection can be more practical for diabetic patients with comorbidities who are more familiar with ECG monitoring for other clinical monitoring [61]. It could also be more favorable for prediabetics who might be more aware due to commercial use in fitness or sports applications. Such AI assistants also provide statistics and communicate with the care provider in the case of an emergency.
