AI Tool Helps People With Type 1 Diabetes Estimate Carbs From Photos

Researchers at the AIDCARE (AI-assisted Diabetes Care) platform have developed a system that uses artificial intelligence to analyze meal photos and help with insulin dosing decisions. The technology works in two main steps: first, it identifies what foods are on your plate, then it estimates how much of each food you have.

Once the system identifies the foods and portion sizes, it calculates the total carbohydrate content. That estimate is combined with your personal clinical information—your insulin-to-carbohydrate ratio and insulin sensitivity factor—to generate a suggested bolus amount. Importantly, you must review and confirm any recommendation before acting on it.

The research team tested the system on a dataset of 40 different food categories. The artificial intelligence engine responsible for identifying foods achieved 94.91% accuracy in validation testing, meaning it correctly identified foods in the image about 95 times out of 100.

The portion-estimation component was also evaluated. On average, it missed the actual portion size by about 12 grams, with some variation. This level of performance is reasonable for a support tool, though no system is perfect, and visual estimation of portion size remains inherently challenging.

This technology is explicitly designed as decision support, not as an autonomous insulin calculator. The system generates advisory bolus guidance, but you retain control—you must confirm or correct the recommendation before using it to dose insulin.

For people with Type 1 diabetes who count carbohydrates or use insulin-to-carbohydrate ratios, a tool that quickly and accurately estimates meal composition could reduce mental load and help you make faster, more confident dosing choices. It may be especially helpful when you're eating unfamiliar foods or dining away from home.

This research describes a working prototype tested on a curated dataset of 40 food types. Real-world use would involve testing on a wider range of foods, lighting conditions, plate styles, and meal types. The system is part of a larger AI-assisted diabetes care platform, suggesting that broader clinical evaluation may be underway or planned.

If you're interested in tools like this, it's worth staying informed about developments in your diabetes technology—and remembering that any new tool is meant to complement, not replace, your knowledge of your own body and your work with your healthcare team.

Source: Nutrients. Evidence type: PubMed indexed literature. Type1Cure is an information and intelligence hub, not a medical advice service. This article summarizes published research and does not provide diagnosis, treatment, or personal medical guidance. Always talk to your own care team before changing anything about your Type 1 diabetes management.