The technology developed at the USP has already been tested by nutritionists, who agreed with 89% of the characteristics of the prepared menus
Agência FAPESP – Researchers at the Food Research Center (FoRC) have used an artificial intelligence (AI) technique to develop a computational tool that, in the future, will allow nutritionists to generate personalized meal plans automatically for your patients The tool has already been tested by nutritionists, who shared 89% of the characteristics of the prepared food plans. The methodology and evaluation were described in the published article in the Journal of Food Composition and Analysis.
Among the numerous artificial intelligence techniques analysed, the researchers chose one applied to computer games, called the Finite State Machine (MSF). Through it, the computer performs a certain action until the desired result is achieved.
“There are a number of rules and possibilities for food pairings. The tool generates different food plans based on these rules, managing to understand which are the correct pairings. This prevents, for example, the system from suggesting rice and coffee beans in morning,” explains nutritionist Kristy Soraya Coelho, a researcher at the FoRC, a FAPESP research, innovation and dissemination center based at the Faculty of Pharmaceutical Sciences of the University of São Paulo (FCF-USP).
In the decision making process of the machine, the chemical composition of foods and yours forms of preparation (cooked, roasted, grilled, with sauce, presence or absence of seasoning, among others), the seasonality and the sensory characteristics of the food (colour, taste and texture).
The patient’s preferences, restrictions and nutritional recommendations are entered into the tool by the nutritionist. The machine then goes through all the variables and possibilities of foods and preparations and makes choices based on them, creating the food plan.
“The tool understands, for example, that you need to have a drink for breakfast, which can be milk, coffee, tea, coffee with milk, among others, as well as other foods, such as bread, biscuits, cakes or fruit ; to then make choices and define meals If the patient, for example, does not consume tomatoes, this food alone or even in preparations will not be part of the meals in the meal plan, such as in a salad, in a sauce or in a a mixed dish,” explains Coelho, coordinator of the project.
During the generation of food plans, the tool estimates the energy requirement, considering the correct distribution of macronutrients (60% carbohydrates, 15% proteins and 25% lipids) and micronutrients (vitamins and minerals), according to the recommendations of the dietary intakes of reference (DRI).
“If necessary, you can replace foods and preparations, change portions and the tool will adjust the meal plan,” he explains.
human competence
One of the differentials of the tool is following a decision tree representing the expertise of the nutritionist in the choice of foods and preparations. “We work with the representation of the nutritionist’s knowledge. We apply the artificial intelligence technique to translate his expertise into the decision-making process and the elaboration of food plans that are more faithful to the patient’s needs and preferences. This type of system, called a specialist, it is unique on the market,” says the researcher.
Another difference is that the tool follows the four phases of the Nutrition Care Process (NCP) defined by the Academy of Nutrition and Dietetics (Academy of Nutrition and Dietetics): assessment of nutritional status, nutritional diagnosis, nutritional intervention, monitoring and follow-up up . It is a systematized process that aims to optimize the nutritionist’s work routine.
“The tool represents a great help in the decision-making process of the nutritionist, as it will optimize the work of the query. However, the final decision on the dietary prescription will always be up to him”, she underlines. Today, according to the researcher, the tools available on the market perform isolated calculations and analyses. “These are information systems that store, process and retrieve the data collected, but they don’t make deductions, that’s up to the nutritionist,” Coelho points out.
Furthermore, the rules of the tool have been optimized with the food habits of Brazilians in mind. The way this software has been designed can be optimized for other countries food habits and include local databases.
national data
The USP tool uses the Brazilian Food Composition Table (TBCA) database, which provides the chemical composition and energy value of foods and preparations, including raw and cooked foods, convenience processed foods, and preparations. The TBCA data relate to foods and preparations usually consumed by the Brazilian population, with adequate and standardized methods.
“This is information obtained from the laboratories of the FCF-USP and from studies at various universities in Brazil, which consider not only the values of fresh foods, but also simple preparations (roasted, boiled, fried, grilled) and recipes. estimate the composition food chemistry through calculations,” he explains.
successful test
When tested, the tool was based on a version of the TBCA that contained data for 2,182 foods, totaling more than 1,800 preparations. Fifteen healthy fictitious patients were created, and based on this, seven weekly meal plans were generated, for a total of 105 plans.
The results were then evaluated by 18 nutritionists with specialization in clinical nutrition or with at least three years of experience in the field. On a scale of zero to ten, they should indicate how closely they match the characteristics of the plans created by the tool.
“The approval rating was 89%, a percentage considered very good for a knowledge-based tool, as not even a human expert can get results 100% correct. The meal plans generated by the system were able to accommodate preferences, nutritional recommendations, and appropriateness in terms of portions and meals,” says Coelho.
With the success of the initial tests, other studies are being developed with different profiles of people, from pregnant women to the elderly.
Precision nutrition
“We will soon start a more advanced phase of the project, which concerns precision nutrition o personalized nutrition, a field that further considers the specificities of patients to increase the assertiveness of the tool. In this clinical study we will consider not only nutritional evaluation but also biochemical tests and genetic aspects,” he adds.
In the future, in addition to information on foods and preparations specific to the regions of Brazil and regional dietary habits, the researchers intend to include resources that allow estimating the cost of meal plans and providing a shopping list for the patient. / This text was originally published by Agência FAPESP.
Source: Terra
Rose James is a Gossipify movie and series reviewer known for her in-depth analysis and unique perspective on the latest releases. With a background in film studies, she provides engaging and informative reviews, and keeps readers up to date with industry trends and emerging talents.
