A system for AI-supported nutritional evaluation and personalized recommendation of vitamin B12 and iron intake

Abstract

An AI-integrated system for personalized assessment and optimization of vitamin B intake 12 and iron, including: a multimodal input aggregator configured to receive and preprocess nutritional data, biometric signals, clinical laboratory data, and optionally genomic information; a predictive absorption modeling unit configured to estimate the individual bioavailability of vitamin B 12 and estimates iron based on physiological, nutritional and contextual factors; a deficiency scoring unit configured to analyze longitudinal data and generate a real-time risk score that indicates the probability of vitamin B deficiency 12 and iron; an adaptive recommendation generator configured to provide personalized nutritional and supplementation suggestions based on estimated absorption and risk values; and a closed-loop feedback system configured to dynamically refine recommendations based on user compliance, biometric feedback, and clinical outcomes.

Description

TECHNICAL FIELD OF THE INVENTION

[0001] The present invention relates to nutritional science, artificial intelligence, and digital health technologies. More specifically, it relates to a novel, hardware- and software-integrated system for assessing, predicting, and dynamically personalizing vitamin B intake. 12 and iron. It uses multimodal inputs, AI-powered nutrient intake models, and real-time adjustment based on biometric and clinical data streams. BACKGROUND OF THE INVENTION

[0002] A vitamin B deficiency 12Iron is among the world's most common and clinically significant micronutrient deficiencies. These deficiencies often go undetected in early stages and can lead to serious health consequences—including iron deficiency anemia, neurological dysfunction, fatigue, developmental delays in children, and complications during pregnancy. Despite widespread awareness of their importance, current methods for detecting and treating these deficiencies are inadequate, outdated, and largely unpersonalized.

[0003] Traditional dietary assessment methods rely heavily on self-report instruments such as food frequency questionnaires (FFQs), 24-hour dietary recalls, or manually completed food diaries. These approaches are highly susceptible to recall bias, inaccurate portion estimates, and incomplete nutrient profiles. They also fail to account for critical contextual factors such as preparation methods, food matrix effects, or nutrient interactions—all factors that influence the actual availability and absorption of nutrients. As a result, even individuals who meet their recommended intake on paper may suffer from functional deficiencies, such as poor bioavailability or absorption inhibition.

[0004] Furthermore, most current dietary recommendations are based on generalized guidelines that do not take into account individual differences in age, metabolism, genetics, digestive health, medication use, and lifestyle. For example, a vegan may consume sufficient amounts of plant-based iron, but without concomitant vitamin C intake or in the presence of phytate inhibitors, absorption may be severely impaired. Similarly, individuals taking metformin or proton pump inhibitors long-term may experience iron deficiency despite adequate vitamin B intake. 12 -Intake suffer from poor absorption. However, such subtleties are rarely taken into account in conventional health tools or advice platforms.

[0005] Another key shortcoming of existing nutrition systems is their reactive nature. Nutrient deficiencies are often only diagnosed after symptoms or complications have already occurred—in many cases, by which time irreversible damage has already occurred. This problem is exacerbated by fragmented healthcare systems in which food diaries, lab results, biometric data from wearables, and medical histories are stored in isolated, disconnected systems. This lack of integration prevents both professionals and those affected from gaining a holistic overview of a person's nutritional status.

[0006] Furthermore, most digital health applications primarily focus on calorie counting, weight loss, or general well-being—while neglecting a differentiated, personalized approach to individual micronutrients. Even recommended dietary supplements are often prescribed according to a "one-size-fits-all" principle, without analyzing actual deficiency risk, absorption capacity, or individual responses to supplementation. Continuous follow-up or adjustment based on real-world results rarely occurs, which can lead to ineffective or even counterproductive strategies.

[0007] With advances in artificial intelligence, wearable sensors, digital health records, and personalized medicine, there is a clear—and urgently needed—opportunity to develop a new generation of tools that continuously, adaptively, and individually monitor and optimize micronutrient intake and effects. However, to date, no comprehensive, AI-powered platform exists that intelligently integrates multimodal data to optimize vitamin B intake. 12 and iron taking into account individual absorption mechanisms and clinical feedback.

[0008] The present invention closes this critical gap through a novel, AI-based system that combines nutritional data, biometric parameters, laboratory values, genetic information and behavioral models to provide personalized, context-aware recommendations for vitamin B intake. 12and iron. Through proactive risk assessment, tailored action recommendations, and ongoing adjustments based on user feedback and results, this system can fundamentally change the supply of these essential nutrients—both at the individual and population levels. SUMMARY OF THE INVENTION

[0009] The present invention relates to a novel, AI-integrated hardware and software system consisting of specialized computing units, sensor units and an AI analysis module, which is used to evaluate, monitor and personalize the intake of vitamin B 12 and iron is configured in real time.

[0010] The system includes a multimodal input unit for capturing nutritional data (e.g., food images), biometric signals (e.g., heart rate variability, intestinal activity), and external medical data (e.g., laboratory values, genomic data). It also includes an AI-based prediction unit for modeling vitamin B bioavailability. 12 and iron based on contextual, real-time physiological data, a risk analysis component for assessing deficiency probabilities taking into account intake, absorption forecasts, and long-term course, as well as a personalized recommendation system that generates adaptive dietary and supplementation suggestions. The system is complemented by a clinical user interface and a data synchronization layer for remote monitoring and physician decision support.

[0011] The main innovations are the real-time integration of absorption forecasts, multimodal personalization, the use of federated learning for privacy-friendly model updates, and an adaptive feedback system for recommendations. DETAILED DESCRIPTION OF THE INVENTION

[0012] The present invention discloses an advanced, AI-integrated system specifically designed for comprehensive, personalized, and real-time assessment and optimization of vitamin B intake. 12and iron. The system utilizes multimodal data fusion, predictive modeling, and adaptive feedback to overcome the significant deficiencies of existing nutrition tools, which typically lack personalization, real-time responsiveness, and precise modeling of nutrient bioavailability. The invention provides a novel framework in which hardware and software components interact seamlessly to continuously collect, analyze, and interpret biological, nutritional, and environmental data. This generates precise and dynamic recommendations tailored to an individual's physiological and lifestyle context.

[0013] The system is based on the Multimodal Input Aggregator (MIA), a sophisticated data platform for integrating heterogeneous data types. The MIA processes nutrition-related inputs through various mechanisms, including manually entered food diaries, barcode scans of packaged foods, and innovative image-based food recognition using convolutional neural networks (CNNs). The image processing module is trained on extensive food databases and enables the precise identification of food items, portion sizes, and preparation methods. This significantly improves the accuracy of nutrient assessment compared to traditional text protocols. In parallel, the MIA communicates with wearable biometric sensors to continuously collect physiological data such as heart rate variability (HRV), body temperature, gastrointestinal motility patterns, and sleep quality.These biometric indicators provide crucial context about a person's metabolic and digestive health—key factors for nutrient absorption and utilization.

[0014] The system also supports integration with clinical laboratory information systems via standardized application programming interfaces (APIs), enabling laboratory values ​​such as serum vitamin B 12 , ferritin, homocysteine, mean corpuscular volume (MCV), and other relevant blood parameters can be imported automatically. The integration of these objective, validated laboratory values ​​provides a solid medical foundation for the system's predictions. Optionally, genomic data can also be included, such as polymorphisms in genes like MTHFR or transcobalamin, which influence micronutrient metabolism. This enables even deeper personalization based on genetic predispositions.

[0015] After acquisition, the various data sets undergo extensive preprocessing and normalization to harmonize them structurally and temporally. This prepared database forms the foundation for the system's central analysis module: the Predictive Absorption Modeling Unit (PAME).

[0016] The PAME represents a technological breakthrough in the field of nutritional assessment, as it uses advanced machine learning techniques to estimate the bioavailability of vitamin B 12and iron individually. In contrast to traditional systems that only consider reported intake levels, PAME integrates physiological and contextual factors influencing absorption. These include, for example, concurrently ingested nutrients (vitamin C promotes iron absorption), intestinal motility and pH derived from biometric data, hydration status, medication use, and lifestyle factors such as alcohol consumption or smoking. By synthesizing these variables, PAME provides a dynamic, precise estimate of the actual physiologically available amount of vitamin B 12 and iron, thus enabling an optimized nutritional and supplementation strategy.

[0017] In addition to the absorption model, the system includes a Deficiency Risk Scoring Unit (DRSU), which uses advanced time-series machine learning models, particularly long-short-term memory (LSTM) networks, to analyze patterns in nutrient intake, absorption predictions, biometric trends, and laboratory values. The DRSU continuously generates updated risk scores that reflect the probability of a person currently having a vitamin B deficiency. 12 or iron deficiency or is at risk of developing it. This proactive risk assessment enables early intervention and can prevent serious health consequences.

[0018] To convert the analysis results into usable instructions, the invention includes an Adaptive Recommendation Generator (ARG). This generates personalized recommendations for food selection and supplementation – including suggestions for specific foods, preparation forms (e.g., oral tablets, sublingual preparations, or intramuscular injections for vitamin B 12 ), dosages, and timing of administration to maximize efficacy and tolerability. The ARG uses reinforcement learning to continuously improve its recommendations based on real-time feedback, compliance data, symptom assessments, and biometric signals. For example, if gastric discomfort is reported after iron intake, an alternative formulation or administration schedule can be suggested.

[0019] A central feature of the system is its Closed-Loop Feedback System (CLFS) architecture, which allows for continuous learning and optimization. The system monitors adherence to recommendations, physiological responses, and clinical outcomes to dynamically adapt future suggestions. This ensures that the individualized nutrition plan remains up-to-date and effective. The feedback promotes user loyalty, increases treatment success, and reduces the risk of recurrence.

[0020] Given the sensitive nature of health and biometric data, the system uses a federated learning framework to ensure user privacy and data security. This decentralized learning paradigm enables AI models to be trained collaboratively on multiple user devices without the need to transfer raw data to central servers. Data exchanged between devices and the cloud is end-to-end encrypted, ensuring compliance with data protection regulations such as HIPAA and GDPR. At the same time, the system can continuously improve its predictive accuracy and generalization capability through collective intelligence.

[0021] Furthermore, the invention includes a clinical dashboard and a remote access interface for healthcare professionals. These components provide comprehensive tools for patient monitoring and targeted intervention. The dashboard visualizes nutrient trends, deficiency risk alerts, and adherence data in an intuitive format, supporting timely clinical decision-making. The interface enables integration with existing electronic health records (EHR) via interoperability standards such as HL7 and FHIR, enabling seamless data exchange and coordinated care between different healthcare providers. Through secure remote access, professionals can adjust treatment plans, prescribe supplements, and communicate directly with patients—significantly expanding the scope and effectiveness of nutritional care.

[0022] The invention presents a groundbreaking system that combines multimodal data fusion, AI-based absorption prediction modeling, deficiency scoring systems, and adaptive feedback mechanisms in a unified platform. It offers a highly personalized and dynamic approach to managing micronutrient intake—with a particular focus on vitamin B. 12 and iron, two nutrients of significant importance to public health. Through early detection of deficiencies and context-aware nutritional optimization, this system contributes to improving individual health outcomes and makes a groundbreaking contribution to the advancement of personalized nutritional medicine.

Claims

[1] An AI-integrated system for personalized assessment and optimization of vitamin B intake 12 and iron, including: a multimodal input aggregator configured to receive and preprocess nutritional data, biometric signals, clinical laboratory data, and optionally genomic information; a predictive absorption modeling unit configured to estimate the individual bioavailability of vitamin B 12 and estimates iron based on physiological, nutritional and contextual factors; a deficiency scoring unit configured to analyze longitudinal data and generate a real-time risk score that indicates the probability of vitamin B deficiency 12 and iron; an adaptive recommendation generator configured to provide personalized nutritional and supplementation suggestions based on estimated absorption and risk values; and a closed-loop feedback system configured to dynamically refine recommendations based on user compliance, biometric feedback, and clinical outcomes. [2] The system of claim 1, wherein the multimodal input aggregator comprises modules for food recognition using convolutional neural networks and barcode scanning to estimate food intake. [3] The system of claim 1, wherein the biometric signals comprise data from wearable devices that measure heart rate variability, gastrointestinal motility, body temperature, and sleep quality. [4] The system of claim 1, wherein the clinical laboratory data comprises serum levels of vitamin B 12, ferritin, homocysteine ​​and mean corpuscular volume, imported via standardized APIs. [5] The system of claim 1, wherein the predictive absorption modeling unit considers physiological variables, including intestinal pH, motility, hydration status, drug effects, and concurrently ingested nutrients, to estimate nutrient bioavailability. [6] The system of claim 1, wherein the deficiency scoring unit uses machine learning models including long-short-term memory networks to analyze temporal trends in nutrient intake, biometric data, and laboratory values.