Using Artificial Intelligence for Food Identification and Dietary Assessment within Restaurant and Institutional...

Recent advancements in artificial intelligence (AI) have led to greater usage of machine and deep learning approaches to assess dietary intake. However, a comprehensive review of existing approaches used in institutional settings is lacking. The aim of this scoping review is to explore current AI methods being utilized for assessing food selections and dietary intake within these settings and summarize emerging applications of AI methods.

Literature searches were conducted across four research databases including PubMed, SCOPUS, IEEE Xplore, and Academic Search Premier. Peer-reviewed studies were selected based on the utilization of AI methods to assess food offerings, food selections, or dietary intake collected from human populations within restaurant and institutional settings (i.e., hospitals, schools, or work cafeterias).

After title/abstract and full-text screening, 23 studies met inclusion criteria. Most were conducted in the United States (n=7). Study settings included hospitals (n=7), restaurants (n=6), cafeterias (n=6), and schools (n=4). The majority of studies relied on image-based data (n=21), while some used wearable sensors for collecting dietary data (n=2). The studies utilized AI for a variety of purposes including policy evaluation (n=6), dietary assessment (n=11), and optimization of operations such as billing systems (n=6). Main challenges for image-based AI methods included dataset variability, overlapping food items, and light reflections.

AI advancements offer the opportunity to more objectively and efficiently measure dietary intake and provide nutritional information, especially within large-scale settings. This demonstrates AI as an emerging tool for public health applications including assessing dietary intake and evaluating policies.