Personalized Glucose Ranges for Dietary Guidance
Solution Overview
Problem
Determining healthy food choices is challenging due to individual variations in response to different foods, influenced by factors like age, sex, weight, and microbiome, making it difficult to provide personalized nutritional guidance.
Innovation Solution
Generating personalized glucose ranges based on glucose traces, machine learning techniques, and non-glucose factors such as lipid profiles and health risks, which can be updated over time, along with providing personalized food guidance to help users make healthier choices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If personalized glucose ranges are generated using machine learning techniques and multiple data sources, then the accuracy and personalization of nutritional guidance is improved, but the device complexity and computational requirements increase
Solution Approach 1:
The system segments the complex task of personalized nutritional guidance into distinct functional modules: a data ingestion service that collects glucose data from CGM devices and non-glucose data from various sources, a machine learning service that processes this segmented data to generate personalized glucose ranges, and a guidance service that delivers recommendations. This segmentation allows each module to specialize in specific functions, improving overall accuracy while managing complexity through modular architecture.
Solution Approach 2:
The patent introduces intermediary components including a data ingestion service that acts as a mediator between multiple data sources (CGM devices, user inputs, health records) and the machine learning model. This intermediary layer standardizes data formats, validates inputs, and prepares data for processing, thereby improving measurement precision without requiring the core machine learning system to directly handle the complexity of diverse data sources.
2Adaptability or versatility
If multiple data sources including CGM devices and non-glucose factors are integrated, then the personalization and effectiveness of food guidance is improved, but the difficulty of detecting and measuring increases
Solution Approach 1:
The data ingestion service is designed as a universal interface that can accept and process multiple types of data sources through standardized methods. It handles glucose data from CGM devices, non-glucose factors (dietary information, physical activity, sleep patterns), and user profile data using a unified data collection framework. This multi-functional approach enables comprehensive personalization while simplifying the detection and measurement process by providing a single entry point for diverse data types.
Solution Approach 2:
The system dynamically adjusts measurement parameters based on the type of data being collected. For glucose data from CGM devices, it uses continuous monitoring parameters; for non-glucose factors like dietary intake, it uses discrete survey-based parameters; for health records, it uses structured medical data parameters. This parameter adaptation allows the system to optimize data collection methods for each specific data source, improving personalization while managing the complexity of detecting and measuring diverse variables.
3Measurement precision
If glucose ranges are updated over time based on new data, then the relevance and accuracy of recommendations is improved, but the loss of time for data processing increases
Solution Approach 1:
The system implements periodic updates of personalized glucose ranges at scheduled intervals (e.g., weekly or monthly) in addition to continuous monitoring. During these periodic recalibration events, the machine learning model reprocesses accumulated data to generate updated glucose ranges. This periodic action ensures accuracy improves over time while preventing excessive real-time processing, as the system only performs intensive updates at scheduled moments rather than continuously.
Solution Approach 2:
The system performs preliminary data processing and validation during data ingestion, preparing data for future model training before it is actually needed. By pre-processing and organizing data as it arrives from various sources, the system reduces the computational burden during actual glucose range updates. This preliminary action allows faster, more efficient recalibration when periodic updates occur, reducing the time loss associated with reprocessing raw data.
Data Source
AI summary
Techniques are disclosed herein for generating personalized glucose ranges. In some examples, during a calibration period, a user may consume foods that they normally consume, as well as some standardized meals. Personalized upper limits and lower limits to the glucose range can be determined (e.g., using machine learning techniques) that account for the shape of glucose trace for the particular user, the food consumed during the calibration period, as well as other non-glucose factors such as questionnaires, the user's lipid profile, obesity or other health risks, and the like. In some cases, food recommendations may also be provided to the user (e.g., using machine learning techniques). The personalized glucose range and other information may be presented to the user on a display.