Applying AI & Phenotypes to Obesity Care
- Apr 14, 2025
- 4 min read
By Leon Katz, MD
The application of Artificial Intelligence (AI) and Precision Medicine (PM) is having a monumental effect on how patients are getting their care now and in the near future. Numerous models of AI have been introduced, too many for us to discuss in this article.
Before starting a discussion about the use of AI for precision medicine as applied to the treatment of obesity, let’s take a look at the scope of the problem. In 2022, there were approximately 890 million obese adults worldwide.1 With the enormous prevalence of the disease – estimated at 20% of women and 14% of men worldwide by 2030 – the global obesity epidemic is causing tremendous strain on healthcare systems worldwide.2 The goal of this article is to discuss the potential use of AI to improve precision medicine as applied to the treatment of obesity and to explore opportunities to improve care and reduce healthcare utilization.
Discussing AI and precision medicine requires a few basic definitions. AI is sometimes defined as a system’s ability to correctly interpret external data, learn from such data, and use those learnings to achieve specific goals and tasks through flexible adaptation.3
Precision medicine has been defined as any medical intervention targeted to population subgroups categorized on common genetic patterns, lifestyles, drug responses and/or environmental and cultural factors.4
The application of AI in medicine has been used extensively in the field of radiology where AI can assist a radiologist in recognition of subtle abnormalities that could easily be missed even by a highly trained professional.
The current standard of care for obesity management utilizes a step wise approach with behavior modification as a first step, followed by anti-obesity medication and then bariatric surgery for people with severe obesity or comorbidities.
The challenge is how to recognize which treatment modality would provide the best results depending on the patient’s etiology of obesity. Clinical data from obesity research is very difficult to interpret because many variables are based on questionnaires, which can introduce bias and lack of consistency.
Dr. Andres Acosta of Mayo Clinic and his colleagues looked to tackle these challenges by classifying obesity based on four pathophysiological and behavioral phenotypes. Their study demonstrated that a phenotype-guided approach showed a 1.75 fold greater weight loss after 12 months compared to non-phenotype guided therapy, which translated to a mean weight loss of 15.9% compared to 9%.5
Additional research has been done to identify and review AI applications in obesity research and care. Dr. Ruopeng An of Washington University Brown School and colleagues reviewed 46 studies that used various AI models and found that the combination of artificial intelligence and precision medicine allowed for recognition of clinically significant patterns of obesity or relationships between specific covariates and weight outcomes.6
AI can assist at every step of obesity management. For example, the initial integration of AI would include the assessment of data in the medical chart, blood work and anthropomorphic data. From there, aggregation of the data would assist with predictive models of future disease burden for the individual patient. This data would be used to create a personalized eating plan that takes into account food availability, preference of the individual and the unique cultural factors involved in choosing food. Traditional methods of food intake depend on people keeping track of individual foods and recording data on paper or in a digital app. By using images of foods consumed, AI may be able to create a more accurate assessment of portion size and caloric intake.
Furthermore, exercise and physical activity represent an excellent example of integrating data from wearables and allowing AI to have a comprehensive understanding of the optimal time of the day for achieving the greatest benefit from physical activity. The response of the individual to different types of exercise, the duration that they can achieve, and the pleasure they derive from specific exercise can provide necessary data so that AI can further improve recommended exercise regimens.
Additionally, a smart chatbot can assist the person by providing motivational messages, but unlike current simplistic models that simply send a message of encouragement, an AI driven chatbot can take the discussion to a higher level by providing real-time responses to the person’s unique thoughts and feelings. The virtuous cycle repeats as the improvement in comorbidities leads to further refinement of recommended treatment. In these ways, the future of AI in the treatment of obesity is exciting and encouraging.
With continued research, we will have an improved understanding of the interaction of genes that contribute to different forms of obesity. New anti-obesity medications will be developed and have improved specificity for differing obesity phenotypes. With the help of AI, aggregated data will provide personalized treatment modalities.
Before concluding, consideration of the importance of explainable artificial intelligence is critical. The future of personalized developments in the medical treatment of obesity depend on excellent AI models. Yet, one of the biggest challenges with AI is the difficulty explaining the manner in which the application derived the recommendation, even for the computer scientists who programmed the original AI code. Use of AI by clinicians and the acceptance of its use by patients will depend on our ability to understand how the data and recommendations were derived.7
There will always be a need for Human In The Loop (HITL) protocols. This approach demonstrates the importance of having human oversight in applying AI for medical management. HITL will help ensure that clinical standards, patient needs, and ethical considerations help provide the safest and most effective treatment for human beings.
Written by Dr. Leon Katz, Medical Weight Loss Specialist & Diplomat, American Board of Obesity Medicine
Comments