TOPLINE: 

An artificial intelligence (AI)-powered smartphone application achieves 99% accuracy in classifying gluten-free compatibility of packaged foods, with results closely aligned with expert dietitian evaluations. 

METHODOLOGY: 

• Patients with celiac disease require strict adherence to a gluten-free diet, yet inconsistent labeling and limited access to dietary expertise can hinder compliance.
• Researchers evaluated whether an AI-powered smartphone application could accurately classify the gluten-free compatibility of packaged foods using images of product labels.
• Investigators prospectively curated photographs of packaged food products from a supermarket in Los Angeles in September 2024; images captured the product front, full ingredient list, and allergen statements under typical in-store lighting conditions.
• A custom AI-powered smartphone application that applies image recognition and natural language processing or large language models was used to extract and interpret label information and classify gluten-free compatibility at both the product and ingredient levels.
• Two gastroenterology-focused registered dietitians independently reviewed all products and ingredients while blinded to the application’s classifications.

TAKEAWAY: 

• In total, 100 packaged food products containing 1884 individual ingredients were analyzed.
• Expert dietitians identified 64% of products and 95.9% of ingredients as compatible with a gluten-free diet. However, only 37.5% of the products classified as gluten-free by experts had an appropriate gluten-free label.
• The AI-powered application achieved 99% accuracy at the product level (sensitivity, 100%; specificity, 98.4%; F1 score, 98.6%) and 99.8% accuracy at the ingredient level (sensitivity, 96.3%; specificity, 99.9%; F1 score, 97.5%).
• The application failed to capture 24 ingredients through image recognition, but this did not affect overall product-level classification accuracy.

IN PRACTICE:

“With rigorous evaluation and validation, such tools may offer practical support for individuals with [celiac disease] by reducing reliance on labeling, enhancing dietary adherence, and complementing existing strategies for gluten avoidance in everyday settings,” the authors wrote.

SOURCE:

This study was led by Jamil S. Samaan, MD, Karsh Division of Gastroenterology and Hepatology, Cedars-Sinai Medical Center, Los Angeles. It was published online in the American Journal of Gastroenterology.

LIMITATIONS:

This study was conducted in a single retail setting under ideal conditions, which may not reflect real-world variability in lighting or user behavior. It included only packaged foods with ingredient labels, excluding fresh or prepared items that often lack standardized disclosures. The application was not benchmarked against other gluten-identification tools and did not account for cross-contamination risks during manufacturing or storage. 

DISCLOSURES:

This study received no external funding or support. The authors reported no relevant conflicts of interest.

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.