Artificial Intelligence in MNT

Artificial Intelligence in MNT
By Sophia Condic, MS, RDN
Today’s Dietitian
Vol. 25 No. 8 P. 38

Limited research suggests RDs may be able to use this innovative technology in the near future to improve patient care.

Artificial intelligence (AI) is becoming the newest celebrity in dietetics. Nutrition professionals and the greater medical community at large are excited about this technology and are learning more about it every day. However, it may take some time before clinicians in various specialties can harness all it has to offer to fully integrate it into daily practice and make a considerable impact on patient care.

IBM says AI “leverages computers and machines to mimic the problem-solving and decision-making capabilities of the human mind.”1 In other words, AI learns from the experts and makes an educated guess on an outcome.1

AI research in the health care field has shown that technology and humans can work together to diagnose, treat, and educate those who have heart failure, diabetes, renal disease, disordered eating patterns and eating disorders, and overweight or obesity. Current AI research has been applied in radiology, pathology, and surgery to help physicians diagnose diseases and identify tumors through new algorithms, thereby leading to less invasive surgery through the da Vinci surgical AI system.2

To date, no AI research has been done in the field of MNT. However, there have been some studies conducted over the past couple of years on AI applications for heart failure, diabetes, and chronic kidney disease (CKD) in inpatient and outpatient settings that may help prepare RDs for the future of AI in dietetics.

AI in Inpatient Settings

How RDs Can Integrate AI in MNT
According to the Agency for Healthcare Research and Quality, some of the most common conditions responsible for hospital admissions in 2018 were heart failure, diabetes with complications, and acute and unspecified renal failure.3 These are some of the more common diseases RDs treat in inpatient settings. During their initial assessments of such patients, RDs have to access anthropometric data. Research suggests dietitians may be able to obtain this information easier than they do currently through a subset of AI called machine learning (ML).

Heart Failure
ML uses algorithms to analyze large amounts of data, learn from the insights, and make informed decisions. It imitates the way humans learn, gradually improving over time.4-6 This subset of AI can be important in relation to heart failure since research is showing that “ML methods can handle temporal, large-volume, and multimodality data (eg, sound, language, tabular EHR, imaging, and metabolomic data).”5 This type of data can be used to understand a diagnosis and then personalize treatment options for patients with heart failure in inpatient settings. In the past, AI had been used for selecting patients who may respond to cardiac resynchronization therapy and collecting continuous data through implanted sensors.5,7 However, new research is showing that AI can benefit those with heart failure in other ways by including other health professionals, such as RDs who are part of a health care team in an inpatient setting.

Although there hasn’t been much research showing how RDs can use ML or how ML can make their jobs easier in inpatient settings, Averbuch and colleagues stated that in relation to heart failure, “ML algorithms can be used to provide recommendations to clinicians regarding optimal sequencing and dosing of evidence-based therapies.”6 Even though the researchers didn’t specify the care RDs provide through MNT, future research can focus on how these algorithms may be able to generate nutrition recommendations specific for patients with heart failure in inpatient settings. For example, in the future, researchers can conduct studies that implement ML algorithms to generate personalized MNT recommendations for patients with heart failure. From there, researchers can test whether the algorithms produced an accurate result and if these algorithms, in turn, can work as an additional resource for RDs.

Diabetes
In addition to heart failure, researchers are focusing on AI applications for patients with diabetes in inpatient settings. One AI technique used extensively in diabetes care is case-based reasoning (CBR), an experience-based approach to solving new problems. According to a study by Ellahham, the goal of CBR is to develop better solutions for managing diabetes based on data from past experiences. For example, the 4 Diabetes Support System, an example of CBR used in diabetes care, can “automatically detect problems in control of blood glucose, propose solutions to the detected problems, and remember the effective and ineffective solutions for individual patients.” CBR also has been used to optimize and personalize insulin therapy for a wide range of meal situations.8

Researchers also are studying AI-based clinical decision support (CDS) tools to improve diabetes management. These tools, with the use of ML, can make short- and long-term predictions about HbA1c after patients start insulin therapy and can “help identify clinical variables that can influence a patient’s HbA1c response.”8 Although CBR and CDS tools may be able to improve how efficiently RDs assess patients with diabetes, there isn’t enough evidence currently to determine whether both of these systems can produce accurate and consistent results or whether they may replace RDs. However, CBR and CDS tools may be used as a resource in MNT for patients with diabetes in inpatient settings. RDs can use CBR to help personalize meal plans for patients with diabetes, while CDS tools can provide the data needed to track patients’ HbA1c. This will enable RDs to educate patients on managing their diabetes and preventing future hospitalizations.8

Chronic Kidney Disease
Dietitians also have to monitor patients with renal failure so patients can avoid hospitalizations. Individuals with renal failure also may be on dialysis and at risk of comorbidities such as CVD, dyslipidemia, and anemia.9 Despite having preexisting conditions, a study by Chen and colleagues showed that AI can be used to confirm the efficacy of an at-home renal care program, which might be a step toward integrating AI in MNT.

The purpose of the study was to evaluate a hospital-to-home (H2H) program involving 120 participants with stage 3 to 5 CKD. They were admitted to a Chinese hospital from March 2019 to March 2021.10 Half of the participants were placed in the experimental group that incorporated the internet-based H2H program.10 The group had access to online resources from an RD, such as recipes and nutrition education handouts.10 The remaining half of the participants were in the control group and didn’t have access to the H2H program.10

Despite the differences between the two groups, all of the participants had their renal status monitored with the use of AI. Everyone had CT images done on their kidneys before and after the study.10 At first, the CT images weren’t clear; however, once the wavelength transform denoising algorithm was incorporated into the imaging system through AI, the “results of CT perfusion imaging [became] more conclusive to analysis and comparison.”10 Because of AI, the researchers could evaluate participants’ renal status throughout the study and determine that the H2H program can be used to monitor patients who have CKD.10

AI in Outpatient Settings

How RDs Can Integrate AI in MNT
Just as some patients visit with RDs in the hospital, others may see them in outpatient settings. According to Sarah Klemm, RDN, CD, LDN, manager of nutrition information services for the Academy of Nutrition and Dietetics in Chicago, some of the more common reasons patients see RDs in the outpatient setting include diabetes management, disordered eating patterns and eating disorders, and weight loss.11 Although there isn’t enough research to determine whether or not AI will replace RDs in outpatient settings, there’s been some evidence showing that AI presented through mobile apps and algorithms may encourage patients to reach their nutrition goals more effectively than before. This research has been prevalent in diabetes management.

According to a study by Ellahham, mobile apps that use AI technology enable people with diabetes to take better control of their condition through self-management.8 For example, the One Drop Mobile app that was tested on 1,288 participants with diabetes over four months reminded them to take their medication, set goals, and view data about their health.8 After the intervention, Ellahham reported that there was “a 1.07% to 1.27% absolute reduction in HbA1c during a median four months of using the app.”8 Furthermore, another intervention program called FareWell, which was accessible through a mobile app, was tested on 118 adults with diabetes who had an HbA1c greater than 6.5% for 12 weeks.8 The goal of this program through the app was to encourage participants to consider a plant-based diet and regular exercise.8 At the end of the study, 28% of participants had an HbA1c lower than 6.5%, and more than 86% continued to use the app after the study.8

In addition to mobile apps, researchers are testing algorithms to determine whether they may benefit patients with diabetes in outpatient settings. Joshua and colleagues examined the Smart Plate Health to Eat, a technological innovation that enables patients with diabetes to determine the type of food, weight, and nutrients contained in the foods they eat to help them meet their nutrition goals. The study included 50 types of food with a total of 30,800 foods using the YOLOv5s algorithm that identified, analyzed nutrient values, and estimated the weight of the foods studied.12 Researchers found that the YOLOv5s algorithm “showed good identification accuracy in the analysis of four types of food, including rice, braised quail eggs in soy sauce, spicy beef soup, and dried radish, with accuracy for weight and nutrition.”12 Although researchers are still testing food and nutrient identification algorithms like YOLOv5s, there may be a future for them for people with diabetes in outpatient settings.12 In the future, RDs may be able to recommend programs with these algorithms to their patients so they can have an additional resource as they keep their nutrition goals on track.

Mobile apps and algorithms also can help individuals who have disordered eating patterns and eating disorders. Braun-Trocchio and colleagues studied the HALO app used to generate a body composition scan on 98 participants.13 The goal of the study was to better understand body image issues and compare the results with the participants’ perceived body, ideal body, and a body their partner would find attractive through a visual analog scale.13 After participants used the app and completed a questionnaire related to their body image, the researchers could determine there was a mean difference between the app and the participants’ perspectives about body image.13 Although more research is needed to establish whether mobile apps, such as the HALO app, will be applicable to RDs in outpatient settings, there’s a possibility that dietitians could still screen patients at risk of disordered eating patterns and eating disorders, which could alert them to intervene and make personalized recommendations and referrals.

Nevertheless, there are other AI tools RDs can use to help patients with disordered eating patterns and eating disorders. Petrauskas and colleagues examined a CDS system (CDSS) to predict the risk of eating disorders in 83 participants with dementia.14 The CDSS was 90.36% accurate for predicting eating disorders, leading researchers to conclude that “compared to the resident geriatricians, the CDSS is more accurate in diagnosing … eating disorders in dementia.”14 While more research will confirm whether CDSS can run independently to assess risk factors for eating disorders, this AI tool may be implemented in MNT in the future. RDs who work in outpatient settings may be able to use CDSS after a nutrition assessment to quickly determine whether a client is at risk of an eating disorder. From there, dietitians can make additional referrals to provide patients with the care they need.

But while the use of AI tools to provide MNT looks promising for clients with diabetes, disordered eating patterns, and eating disorders, there isn’t enough evidence to determine whether they can be used to monitor weight management or weight loss goals.

For example, Oh and colleagues reviewed nine studies to investigate whether Chatbot (voice-activated applications like Amazon’s Alexa or Apple’s Siri) had an impact on changing health behaviors such as physical activity, healthful eating, and weight management.15 The research team found there were no definitive conclusions on whether Chatbot could affect weight management behaviors.15

Moreover, Chew and colleagues reviewed 66 studies to determine whether AI could be used to help participants lose weight.16 The studies were separated into three categories based on what type of AI application was used in each study.16 Most studies used machine perception, which focused on using AI to recognize foods and behaviors.16 Other studies used AI to predict health behaviors, such as weight loss and emotional eating, or collect personalized data to provide feedback to a participant.16 However, after analyzing the efficacy of the AI applications used in the studies, Chew and colleagues said that “Only six studies reported average weight losses (2.4 to 4.7%), of which two were statistically significant.”16

Despite these findings that there isn’t enough research yet to use AI applications to establish weight management and weight loss goals, RDs can do their part by learning more about AI in the health care field to understand how it may impact MNT in the future.

What the Future Holds
Like any up-and-coming innovative technology, AI is making great inroads in the medical field overall and trying to find its place in dietetics. And while currently, there haven’t been many studies demonstrating how dietitians can implement AI in MNT, research suggests the possibilities.

For example, future research may examine how AI can generate assessment plans, PES statements, and personalized goals based on clients’ data. In addition, as more studies emerge, there will be more opportunities for RDs to join research teams to study and understand the impact of AI in MNT. When Chen and colleagues added an RD to their research team, they could provide nutrition education materials to participants in their experimental group and therefore now understand how MNT and AI can work together to improve health outcomes.10

More research will determine how AI may change the workflow for RDs or have the potential to replace them in the future.

However, as dietitians wait for more AI research to surface, they can learn more about the technology in the health care field through peer-reviewed literature and zero in on studies that include RDs on the research team so they can educate patients on new AI applications and help them reach their health goals in new and innovative ways.

AI has been and will continue to be a hot topic among health care professionals, and this trend will keep growing as researchers learn more about the efficiency of AI in assessing and monitoring patients.

— Sophia Condic, MS, RDN, is a Michigan-based dietitian who’s currently working in a community clinic. She earned a Bachelor of Integrative Studies at Oakland University and completed the Coordinated Program in Dietetics and a Master of Science in Clinical Dietetics at Grand Valley State University. She has previously written for Today’s Dietitian and for the Dietitians in Business and Communications Practice Group through the Academy of Nutrition and Dietetics.

References
1. What is artificial intelligence (AI)? IBM website. https://www.ibm.com/topics/artificial-intelligence. Accessed April 17, 2023.

2. Liu P, Lu L, Zhang J, Tong-tong H, Song-xiang L, Zhe-wei Y. Application of artificial intelligence in medicine: an overview. Curr Med Sci. 2021;41(6):1105-1115.

3. Most frequent principal diagnoses for inpatient stays in U.S. hospitals, 2018. Agency for Healthcare Research and Quality. https://hcup-us.ahrq.gov/reports/statbriefs/sb277-Top-Reasons-Hospital-Stays-2018.jsp. Published July 13, 2021. Accessed April 18, 2023.

4. What is machine learning? IBM website. https://www.ibm.com/topics/machine-learning. Accessed June 20, 2023.

5. Averbuch T, Sullivan K, Sauer A, et al. Applications of artificial intelligence and machine learning in heart failure. Eur Heart J Digit Health. 2022;3(2):311-322.

6. Artificial intelligence (AI) vs. machine learning (ML). Google Cloud website. https://cloud.google.com/learn/artificial-intelligence-vs-machine-learning. Accessed August 9, 2023.

7. Bachtiger P, Plymen CM, Pabari PA, et al. Artificial intelligence, data sensors and interconnectivity: future opportunities for heart failure. Card Fail Rev. 2020;6:e11.

8. Ellahham S. Artificial intelligence: the future for diabetes care. Am J Med. 2020;133(8):P895-900.

9. Treat complications & comorbidities. National Institute of Diabetes and Digestive and Kidney Diseases website. https://www.niddk.nih.gov/health-information/professionals/clinical-tools-patient-management/kidney-disease/identify-manage-patients/manage-ckd/treat-complications-comorbidities. Accessed July 4, 2023.

10. Chen X, Huang X, Yin M. Implementation of hospital-to-home model for nutritional nursing management of patients with chronic kidney disease using artificial intelligence algorithm with CT internet. Contrast Media Mol Imaging. 2022;2022:1183988.

11. Klemm S. 10 reasons to see an RDN. Academy of Nutrition and Dietetics website. https://www.eatright.org/health/wellness/healthful-habits/10-reasons-to-see-an-rdn. Published August 2, 2022. Accessed April 18, 2023.

12. Joshua SR, Shin S, Lee J, Kim SK. Health to eat: a smart plate with food recognition, classification, and weight measurement for type-2 diabetic mellitus patients’ nutrition control. Sensors. 2023;23(3):1656-1674.

13. Braun-Trocchio R, Brandner C, Willis J, Graybeal A. Quantifying body image through smartphone-based artificial intelligence: a new methodological approach. J Acad Nutr Diet. 122(9):A44.

14. Petrauskas V, Jasinevicius R, Damuleviciene G, et al. Explainable artificial intelligence-based decision support system for assessing the nutrition-related geriatric syndromes. Appl Sci. 2021;11(24):11763-11782.

15. Oh YJ, Zhang J, Fang M, Fukuoka Y. A systematic review of artificial intelligence chatbots for promoting physical activity, healthy diet, and weight loss. Int J Behav Nutr Phys Act. 2021;18(1):160.

16. Chew HSJ, Ang WHD, Lau Y. The potential of artificial intelligence in enhancing adult weight loss: a scoping review. Public Health Nutr. 2021;24(8):1993-2020.