A meal pairing system

Abstract

The present application relates to the technical field of diabetes dietary intervention, and discloses a dietary collocation system, which comprises an image acquisition module, an image preprocessing module, an AI recognition module, a dietary collocation module, a user interaction module and a multi-dimensional food material database. The present application solves the problem that the existing diabetes dietary intervention is based on the western mode and does not combine Chinese dietary culture and traditional Chinese medicine theory, realizes the synchronous recognition of food types, heat, GI value and traditional Chinese medicine cold, heat, warm and cool, and five flavor attributes, and can collocate personalized diet according to the user's constitution, is suitable for the dietary control of pre-diabetes and diabetes patients, helps the collaborative blood sugar control of traditional Chinese medicine and western medicine, is convenient to operate, has high recognition accuracy, can be applied to medical scenes such as outpatient service, physical examination center and community health center, and can also be used for daily personal dietary guidance of pre-diabetes and diabetes patients.

Description

Technical Field

[0001] This invention relates to the field of dietary intervention technology for diabetes, and more particularly to a dietary matching system, belonging to the technical field of combining artificial intelligence and traditional Chinese medicine. Background Technology

[0002] Diabetes is a common chronic metabolic disease, and dietary intervention is one of the core means of diabetes prevention and treatment. Current research on diabetes dietary intervention mainly focuses on Western dietary patterns, such as the Mediterranean diet and low-carb diets. It has not yet developed healthy dietary intervention strategies suitable for the Chinese cultural context and meeting the dietary needs of prediabetic and diabetic patients. These strategies have low compatibility with the dietary habits and culture of Chinese residents, resulting in poor patient adherence.

[0003] Traditional Chinese medicine has a long-standing theory of "medicine and food sharing the same origin." Food, like medicine, possesses "four natures" (cold, hot, warm, cool) and "five flavors" (sour, bitter, sweet, pungent, salty). The Yellow Emperor's Inner Classic clearly proposes the theory of the five flavors and five organs entering the meridians: "sour enters the liver, pungent enters the lungs, bitter enters the heart, salty enters the kidneys, and sweet enters the spleen." This theory is an important basis for dietary regulation in traditional Chinese medicine. It allows for targeted selection and combination of ingredients based on individual constitution, achieving "dietary therapy based on syndrome differentiation."

[0004] Current applications of artificial intelligence technology in the field of food recognition mainly focus on the types, calories, and nutrients of food, as well as the calculation of glycemic index (GI) for specific populations. They do not incorporate the five flavors and the attributes of cold, hot, warm, and cool in traditional Chinese medicine, and therefore cannot achieve personalized dietary matching based on TCM constitution. Furthermore, the existing food image recognition models have a single dimension of data annotation and do not integrate diabetes-adaptive attributes with TCM constitution attributes, so the recognition results cannot directly serve the TCM dietary conditioning of diabetic patients.

[0005] Furthermore, existing dietary intervention tools are mostly general-purpose and lack specific dietary guidelines designed for the physiological characteristics of diabetic patients. They fail to balance the blood sugar control requirements of Western medicine with the constitution-regulating needs of Traditional Chinese Medicine (TCM), making it difficult to achieve the goal of coordinated blood sugar control through both TCM and Western medicine. Therefore, there is an urgent need to develop a diabetic dietary matching system that combines the TCM theory of five flavors and five internal organs with artificial intelligence technology to address the aforementioned shortcomings of existing technologies. Summary of the Invention

[0006] The purpose of this invention is to provide a dietary matching system to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides a dietary matching system, including an image acquisition module, an image preprocessing module, an AI recognition module, a dietary matching module, a user interaction module, and a multi-dimensional food database, wherein the modules are interconnected.

[0008] The image acquisition module is used to acquire food image information; The image preprocessing module performs standardization and data enhancement on food images; The multidimensional food database stores basic food information, diabetes compatibility attributes, and traditional Chinese medicine (TCM) properties of four natures and five flavors. Basic food information includes name, category, calories, and nutritional components. Diabetes compatibility attributes include GI value, dietary fiber content, and compatibility rating. TCM properties of four natures and five flavors include cold, hot, warm, and cool properties and five flavor types. The AI ​​recognition module is based on a deep learning model constructed using CNN and YOLO algorithms to achieve feature extraction and real-time multi-attribute recognition of food images. The dietary matching module incorporates the traditional Chinese medicine five-flavor combination rules and diabetes dietary management rules, and generates personalized dietary matching plans by combining the recognition results of the AI ​​recognition module, the user's physique and blood sugar indicators. The user interaction module is used to realize user information input, recognition result display and dietary plan output.

[0009] Preferably, the image acquisition module is a high-definition camera, and the resolution of the acquired food images is not less than 1920×1080 pixels. It supports the acquisition of static images and dynamic video frames, with an acquisition frame rate of 15-30fps.

[0010] Preferably, the standardization processing of the image preprocessing module includes adjusting the food image to a uniform size of 640×640 pixels and converting it to JPEG format. Data enhancement includes adjusting the brightness by ±20%, adjusting the contrast by ±20%, rotating the image randomly from 0 to 360°, and flipping it horizontally / vertically.

[0011] Preferably, during the deep learning model training process of the AI ​​recognition module, the labeled food image dataset is divided into a training set, a validation set, and a test set according to 80%:10%:10%, and the model is optimized by cross-validation. The model's food recognition accuracy is ≥95%, precision is ≥94%, and recall is ≥94%.

[0012] Preferably, in the multidimensional food database, the five flavor types are divided into five categories: sour, bitter, sweet, pungent, and salty, corresponding to the five internal organs and meridians: sour enters the liver, bitter enters the heart, sweet enters the spleen, pungent enters the lungs, and salty enters the kidneys; the four nature attributes are divided into five levels: cold, cool, warm, hot, and neutral; the diabetes suitability rating is marked as 1 / 0, where 1 indicates suitable for diabetic patients and 0 indicates unsuitable. Suitable foods are low-GI foods (GI < 55) and high-dietary-fiber foods (dietary fiber content ≥ 2g / 100g).

[0013] Preferably, the dietary matching module includes the TCM five-flavor matching rules, which include the balanced ratio of the five flavors and the matching rules between body constitution and the four natures and five flavors. The diabetes dietary management rules include the control of daily calorie intake, the carbohydrate energy ratio of ≤50%, and the low glycemic load dietary matching rules.

[0014] Preferably, the user interaction module supports the input of user information including age, gender, weight, blood glucose level, and TCM constitution type. The TCM constitution types cover Yin deficiency, Yang deficiency, Qi deficiency, phlegm-dampness, damp-heat, blood stasis, Qi stagnation, special constitution, and balanced constitution. The dietary plan output format includes a text-based combination list, a nutrition facts table, and cooking suggestions.

[0015] Preferably, the AI ​​recognition module uses LabelImg and Labelme tools to annotate food images during model training. The annotation content includes the bounding box position, name, category, GI value, five flavor types, hot / cold / warm / cool attributes, and adaptation rating of the food. The annotation file is in TXT format and corresponds one-to-one with the corresponding food image.

[0016] The dietary planning method of the dietary planning system includes the following steps: S1: Users enter their personal information through the user interaction module, including their TCM constitution type, blood sugar level, weight, and dietary preferences; S2: The image acquisition module acquires image information of the food to be identified and transmits it to the image preprocessing module; S3: The image preprocessing module performs standardization and data enhancement on the food image and then transmits it to the AI ​​recognition module; S4: The AI ​​recognition module extracts food image features through a deep learning model, matches them with a multi-dimensional food database, and outputs multi-attribute recognition results for the food. S5: The dietary matching module combines the recognition results and user personal information to generate personalized dietary matching plans based on the four natures and five flavors matching rules of traditional Chinese medicine and the dietary management rules for diabetes. S6: The user interaction module displays the recognition results and outputs dietary matching plans, supporting the modification and saving of the plans.

[0017] Preferably, the system is applied in medical units such as outpatient clinics, physical examination centers, and community health centers, and is suitable for dietary guidance and intervention for people with prediabetes, type 2 diabetes patients, and healthy people who need to control blood sugar and lose weight.

[0018] Compared with the prior art, the beneficial effects of the present invention are: the dietary matching system: This invention combines the theory of five flavors and five internal organs in traditional Chinese medicine, the properties of cold, heat, warmth and coolness, with deep learning technology in artificial intelligence to construct a diabetes dietary matching system suitable for the context of Chinese dietary culture. It solves the problem that existing dietary interventions that borrow from Western models are not compatible with the dietary habits of Chinese residents, and achieves the goal of dietary intervention for blood sugar control through the synergy of traditional Chinese and Western medicine. The multidimensional food database constructed by this invention integrates basic food information, diabetes compatibility attributes, and the five flavors, cold, hot, warm and cool attributes of traditional Chinese medicine, realizing the unified storage and matching of multidimensional food information, and providing comprehensive data support for AI recognition and dietary matching. The AI ​​recognition module of this invention is based on the CNN+YOLOv8 deep learning model, which realizes the synchronous real-time recognition of food type, calories, GI value and the five flavors and cold, hot, warm and cool attributes of traditional Chinese medicine. It has a high recognition accuracy and can handle recognition scenarios with complex backgrounds and multiple overlapping foods, adapting to the needs of actual dietary scenarios. The dietary matching module of this invention has a built-in dual matching rule, which takes into account both the blood glucose control requirements of Western medicine for diabetes and the body conditioning needs of traditional Chinese medicine. It can generate personalized dietary plans based on the user's body constitution and blood glucose indicators, thereby improving the scientific nature and pertinence of dietary intervention and effectively improving the patient's dietary compliance. The system of this invention is easy to operate, the image acquisition and recognition process is fast, and the dietary plan output format is intuitive. It can be implemented in medical scenarios such as outpatient clinics and community health centers, and can also be used for patients' home dietary management. It has a wide range of applications and helps to implement the national "Rational Diet Action", reduce the incidence of diabetes, and alleviate the social and economic burden. Attached Figure Description

[0019] Figure 1 This is a diagram illustrating the overall architecture of the dietary planning system. Figure 2 Flowchart for processing food image datasets; Figure 3 The model architecture and workflow diagram for the AI ​​recognition module; Figure 4 This is a flowchart of the overall workflow of the dietary planning system. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention have been clearly and completely described. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figures 1-4The present invention provides a technical solution: a dietary matching system, including an image acquisition module, an image preprocessing module, an AI recognition module, a dietary matching module, a user interaction module, and a multi-dimensional food database, wherein the modules are interconnected. The image acquisition module is used to acquire food image information. It adopts a high-definition camera, supports the acquisition of static images and dynamic video frames, and the resolution of the acquired food images is not less than 1920×1080 pixels. The acquisition frame rate is 15-30fps, which can clearly capture the appearance features of food and meet the subsequent recognition needs. The image preprocessing module performs standardization and data augmentation on food images. Standardization includes adjusting the food images to a uniform size of 640×640 pixels and converting them to JPEG format. Data augmentation includes adjusting the brightness by ±20%, the contrast by ±20%, rotating the images randomly from 0 to 360°, and flipping them horizontally / vertically. Data augmentation enhances the diversity of the dataset and optimizes the model training effect. The multidimensional food database serves as the core data support for the system, storing basic food information, diabetes compatibility attributes, and traditional Chinese medicine (TCM) five flavors and four properties attributes. Basic food information includes name, category (staple food, vegetables, fruits, meat, snacks, etc.), calories, and nutritional components such as protein / carbohydrate / fat. Diabetes compatibility attributes include GI value, dietary fiber content, and compatibility rating (1 indicates suitable, 0 indicates unsuitable). TCM five flavors and four properties attributes include five flavor types (sour, bitter, sweet, pungent, salty) and cold / hot / warm / cool attributes (cold, cool, warm, hot, neutral). The five flavor types correspond to the five internal organs and meridians: sour enters the liver, bitter enters the heart, sweet enters the spleen, pungent enters the lungs, and salty enters the kidneys. In the diabetes compatibility rating, category 1 foods are low-GI foods (GI < 55) and high-fiber foods (dietary fiber content ≥ 2g / 100g), meeting the dietary management requirements for diabetic patients. The AI ​​recognition module is based on a deep learning model constructed using CNN (Convolutional Neural Network) and YOLO (YouOnlyLookOnce) algorithms. Specifically, ResNet50 is used as the backbone network of the CNN for food image feature extraction, and YOLOv8 algorithm is combined to achieve real-time target detection and multi-attribute recognition of food. During model training, LabelImg and Labelme tools are used to annotate food images. The annotation content includes the bounding box position, name, category, GI value, five flavor types, hot / cold / warm / cool attributes, and suitability rating of the food. The annotation files are in TXT format and correspond one-to-one with the corresponding food images. The annotated food image dataset is divided into training, validation, and test sets according to 80%:10%:10%. Cross-validation is used to optimize the model. The model's food recognition accuracy is ≥95%, precision is ≥94%, and recall is ≥94%, which can efficiently handle image recognition scenarios with complex backgrounds and multiple overlapping foods. The dietary matching module incorporates traditional Chinese medicine (TCM) rules on the compatibility of five flavors and diabetic dietary management rules. The TCM rules on the compatibility of five flavors include balanced proportions and matching of body constitution with the five flavors (e.g., warm and hot foods are suitable for those with Yang deficiency, and cool and cold foods are suitable for those with Yin deficiency). The diabetic dietary management rules include daily calorie intake control, carbohydrate energy ratio ≤50%, and low glycemic load dietary matching rules. This module combines the recognition results of the AI ​​recognition module, the user's body constitution, and blood glucose indicators to generate personalized dietary matching plans through algorithm calculations, taking into account both TCM body constitution conditioning and Western medicine blood glucose management. The user interaction module serves as the system's human-computer interaction entry point. It supports the input of user information including age, gender, weight, fasting / postprandial blood glucose levels, and TCM constitution type (balanced, qi deficiency, yang deficiency, yin deficiency, phlegm-dampness, damp-heat, qi stagnation, blood stasis, and special constitution). It is also used to display food identification results and output dietary matching plans. The output plan includes graphic and textual matching lists, nutrition facts tables, and cooking suggestions. Users can also personalize and save the plans locally.

[0022] This invention also provides a method for combining meals using the above-mentioned meal combination system, comprising the following steps: S1: Users enter personal information through the user interaction module, including body type, blood sugar level, weight and dietary preferences. The system verifies and stores the entered information. S2: The image acquisition module acquires image information of the food to be identified. It can acquire images of a single food or multiple mixed foods and transmit the image information to the image preprocessing module in real time. S3: The image preprocessing module performs standardization and data enhancement on food images, unifies image size and format, improves the recognizability of image features, and transmits the processed images to the AI ​​recognition module. S4: The AI ​​recognition module extracts high-dimensional features from food images through a deep learning model, matches them with data in a multi-dimensional food database, and quickly outputs multi-attribute recognition results such as the food's name, category, calories, GI value, five flavor types, hot / cold / warm / cool attributes, and suitability rating. S5: The dietary matching module combines the above identification results and user personal information, and generates a personalized dietary matching plan that includes food pairing, consumption amount and eating time through algorithm calculation based on the five flavor matching rules of traditional Chinese medicine and the dietary management rules of diabetes. This ensures that the plan not only meets the blood sugar management requirements of diabetes, but also adapts to the user's traditional Chinese medicine constitution. S6: The user interaction module displays the multi-attribute recognition results of food and outputs dietary matching plans in the form of a combination of pictures and text. Users can modify the plans according to their own dietary preferences, and the plan can be saved and printed.

[0023] The present invention also provides applications of the above-mentioned dietary planning system. The system is applied in medical units such as outpatient clinics, physical examination centers, and community health centers. It is suitable for dietary guidance and intervention for people with prediabetes, type 2 diabetes patients, and healthy people who need to control blood sugar and lose weight. It can be used as an auxiliary tool for dietary guidance by medical staff, and also as a tool for home dietary management for patients.

[0024] This embodiment also provides a dietary matching method for the above system. Taking a prediabetic individual (male, 45 years old, weight 75kg, phlegm-dampness constitution, fasting blood glucose 6.2mmol / L, dietary preference for vegetarianism) as an example, the specific steps are as follows: S1: This user entered personal information through the system's mobile app: phlegm-dampness constitution, fasting blood glucose 6.2 mmol / L, weight 75 kg, vegetarian preference; S2: The user captures images of a mixed food of celery, brown rice, yam, and winter melon using the APP's high-definition camera. The image acquisition module then transmits the images to the image preprocessing module. S3: The image preprocessing module adjusts the image to JPEG format with 640×640 pixels, performs brightness +10% and horizontal flipping, and then transmits it to the AI ​​recognition module. S4: The AI ​​recognition module completes image feature extraction and database matching within 0.4 seconds, and outputs the recognition results: Celery (pungent, cool, GI25, dietary fiber 1.6g / 100g, compatibility rating 1), Brown rice (sweet, neutral, GI47, dietary fiber 3.5g / 100g, compatibility rating 1), Yam (sweet, neutral, GI51, dietary fiber 0.8g / 100g, compatibility rating 1), Winter melon (sweet, cool, GI16, dietary fiber 0.7g / 100g, compatibility rating 1); S5: The dietary matching module combines the recognition results with user information and generates a personalized dietary plan based on the five flavor matching rules for phlegm-dampness constitution (avoiding sweet and greasy foods, and matching with sweet, bland and cool flavors) and the dietary management rules for diabetes (daily calorie intake of 2250kcal, carbohydrate energy ratio of 45%). Breakfast is brown rice porridge (50g brown rice) + cold celery salad (200g celery), lunch is stir-fried yam and winter melon (100g yam + 200g winter melon) + brown rice (70g brown rice), and dinner is winter melon and kelp soup (200g winter melon + 50g kelp) + steamed yam (100g). The nutritional components, calories and cooking suggestions for each meal are also marked. S6: The APP displays the above recognition results and outputs a dietary plan in the form of pictures and text. Users can change the cold celery salad to stir-fried celery according to their own taste. The system saves the modified plan and provides a printing function.

[0025] The dietary planning system of this embodiment has been tested in a community health center with 100 people with prediabetes for 3 months. The results showed that 89% of the participants said the dietary plan was in line with their own eating habits and physical condition, 78% of the participants' fasting blood glucose dropped to the normal range, and 65% of the participants' physical condition improved, proving that the system of this invention has good practicality and effectiveness.

[0026] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A dietary matching system, characterized in that, It includes an image acquisition module, an image preprocessing module, an AI recognition module, a dietary matching module, a user interaction module, and a multi-dimensional food database, with each module communicating with each other; The image acquisition module is used to acquire food image information; The image preprocessing module performs standardization and data enhancement on food images; The multidimensional food database stores basic food information, diabetes compatibility attributes, and the four natures and five flavors of food in traditional Chinese medicine. The basic food information includes name, category, calories, and nutritional components. The diabetes compatibility attributes include GI value, dietary fiber content, and compatibility rating. The four natures and five flavors of food in traditional Chinese medicine include five flavor types and cold, hot, warm, and cool attributes. The AI ​​recognition module is based on a deep learning model constructed using CNN and YOLO algorithms to achieve feature extraction and real-time multi-attribute recognition of food images. The dietary matching module incorporates the four natures and five flavors compatibility rules of traditional Chinese medicine and the dietary management rules for diabetes. It combines the recognition results of the AI ​​recognition module, the user's physique and blood sugar indicators to generate personalized dietary matching plans. The user interaction module is used to realize user information input, recognition result display and dietary plan output.

2. The dietary matching system according to claim 1, characterized in that, The image acquisition module is a high-definition camera that captures food images with a resolution of no less than 1920×1080 pixels. It supports the acquisition of static images and dynamic video frames at a frame rate of 15-30fps.

3. The dietary matching system according to claim 1, characterized in that, The standardization processing of the image preprocessing module includes adjusting the food image to a uniform size of 640×640 pixels and converting it to JPEG format. Data enhancement includes adjusting the brightness by ±20%, the contrast by ±20%, rotating the image randomly from 0 to 360°, and flipping it horizontally / vertically.

4. The dietary matching system according to claim 1, characterized in that, During the training process of the deep learning model of the AI ​​recognition module, the labeled food image dataset is divided into training set, validation set and test set according to 80%:10%:10%, and the model is optimized by cross-validation. The model has a food recognition accuracy of ≥95%, precision of ≥94% and recall of ≥94%.

5. The dietary matching system according to claim 1, characterized in that, The multidimensional food database categorizes the five flavors into sour, bitter, sweet, pungent, and salty, corresponding to the five internal organs and meridians: sour enters the liver, bitter enters the heart, sweet enters the spleen, pungent enters the lungs, and salty enters the kidneys. The four nature attributes are categorized into cold, cool, warm, hot, and neutral. The diabetes suitability rating is marked as 1 / 0, where 1 indicates suitable for diabetic or prediabetic patients and 0 indicates unsuitable. Suitable foods are low-GI foods (GI < 55) and high-fiber foods (fiber content ≥ 2g / 100g).

6. The dietary matching system according to claim 1, characterized in that, The dietary matching module includes the traditional Chinese medicine five-flavor matching rules, which include the balanced ratio of the five flavors and the matching rules between body constitution and the five flavors. The dietary management rules for diabetes include the control of daily calorie intake, the proportion of carbohydrate energy supply ≤50%, and the dietary matching rules for low glycemic load.

7. The dietary matching system according to claim 1, characterized in that, The user interaction module supports the input of user information including age, gender, weight, blood sugar level, and TCM constitution type. The TCM constitution types cover Yin deficiency, Yang deficiency, Qi deficiency, phlegm-dampness, damp-heat, blood stasis, Qi stagnation, special constitution, and balanced constitution. The dietary plan output format includes a graphic and text-based combination list, a nutrition facts table, and cooking suggestions.

8. The dietary matching system according to claim 1, characterized in that, The AI ​​recognition module uses LabelImg and Labelme tools to annotate food images during model training. The annotations include the bounding box location, name, category, GI value, five flavor types, hot / cold / warm / cool attributes, and adaptation rating of the food. The annotation files are in TXT format and correspond one-to-one with the corresponding food images.

9. The dietary combination method of the dietary combination system according to any one of claims 1-8, characterized in that, Includes the following steps: S1: Users enter their personal information through the user interaction module, including their TCM constitution type, blood sugar level, weight, and dietary preferences; S2: The image acquisition module acquires image information of the food to be identified and transmits it to the image preprocessing module; S3: The image preprocessing module performs standardization and data enhancement on the food image, and then transmits it to the AI ​​recognition module; S4: The AI ​​recognition module extracts food image features through a deep learning model, matches them with a multi-dimensional food database, and outputs multi-attribute recognition results for the food. S5: The dietary matching module combines the recognition results and user personal information to generate personalized dietary matching plans based on the five flavors matching rules of traditional Chinese medicine and the dietary management rules for diabetes. S6: The user interaction module displays the recognition results and outputs dietary matching plans, supporting the modification and saving of the plans.

10. The application of the dietary matching system according to any one of claims 1-8, characterized in that, The system is applied in medical units such as outpatient clinics, physical examination centers, and community health centers, and is suitable for dietary guidance and intervention for people with prediabetes, type 2 diabetes patients, and healthy people who need to control blood sugar and lose weight.

Images