A meal recommendation method and system, an electronic device and a storage medium
By collecting and analyzing oral cavity condition and food characteristics data, and using multispectral imaging and sensor technology combined with machine learning models, dietary recommendations are dynamically adjusted. This solves the problem of lack of multimodal perception and dynamic matching in existing systems, and achieves the accuracy and adaptability of personalized dietary recommendations, meeting the refined needs of oral health management.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NINGBO FOTILE KITCHEN WARE CO LTD
- Filing Date
- 2026-01-27
- Publication Date
- 2026-06-12
AI Technical Summary
Existing dietary recommendation systems lack multimodal perception and dynamic matching capabilities, and cannot effectively combine oral condition and food characteristics, resulting in insufficient accuracy and adaptability of personalized dietary recommendations, and failing to meet the refined needs of oral health management.
By collecting oral cavity status data and food characteristic data, quantitative evaluation is carried out using equipment such as multispectral imaging, miniature pH sensors and piezoresistive thin film sensors. Combined with machine learning models to predict the mechanical and chemical properties of food, dietary recommendations are dynamically adjusted to compensate for nutritional damage, thus achieving multimodal perception and dynamic matching.
It improves the personalization accuracy and adaptability of dietary recommendations, ensures that recommended foods meet oral health needs, avoids mechanical and chemical irritation, optimizes nutrient absorption, and achieves refined oral health management.
Smart Images
Figure CN122201632A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of nutrition management technology, and in particular to a dietary recommendation method, system, electronic device, and storage medium. Background Technology
[0002] With the improvement of living standards and health awareness, people's nutritional awareness has increased. Healthy people often consult nutritionists or look up relevant advice on food websites to plan their daily diet. For people with illnesses, they rely more on the professional knowledge and experience of nutritionists or doctors.
[0003] In the field of oral health management and dietary recommendations, there is currently no intelligent dietary recommendation system directly linked to oral health conditions. Related dietary guidance mainly relies on manual reminders from dentists, lacking systematic and dynamic technological support. Existing dietary recommendation methods heavily depend on subjective user descriptions, such as self-assessment of oral discomfort based on pain levels, lacking objective quantitative indicators. Furthermore, there are no effective means to collect and analyze key parameters such as occlusal force and mucosal damage area. Simultaneously, the lack of integration of real-time biosensor data, such as chewing electromyography signals and oral temperature distribution, results in qualitative assessments of oral health conditions with insufficient accuracy. In addition, traditional systems only use a simple "hard / soft" dichotomy to categorize food ingredients, failing to establish scientific matching models and considering changes in absorption efficiency caused by oral damage. Their ability to adapt nutritional needs to oral health conditions is also significantly inadequate, making it difficult to meet the personalized dietary needs of individuals with abnormal oral health conditions.
[0004] Therefore, it is particularly important to develop an intelligent dietary recommendation system capable of multimodal perception and dynamic matching to meet the refined needs of oral health management and improve the accuracy and adaptability of personalized dietary recommendations. Summary of the Invention
[0005] To address the aforementioned technical problems, this application provides a dietary recommendation method, system, electronic device, and storage medium. By matching dietary recommendation results based on oral health data and food characteristic data, and adjusting the dietary recommendation results through nutritional impairment compensation, this solution addresses the current lack of intelligent dietary recommendation systems capable of multimodal perception and dynamic matching. This addresses the need for refined oral health management and improves the accuracy and adaptability of personalized dietary recommendations.
[0006] The technical solution provided in this application is as follows: On the one hand, this application provides a dietary recommendation method, the dietary recommendation method comprising: Collect current oral cavity condition data and food ingredient characteristic data; Based on the current oral health data and the food ingredient characteristics data, a first dietary recommendation result is determined, which includes a first recommended dose. The nutritional damage coefficient is determined based on the current oral health data. Based on the first recommended dose and the nutritional damage coefficient, determine whether the first recommended dose meets the standard; If the first recommended amount is not met, a second recommended amount is determined based on the nutritional impairment coefficient, the first recommended amount is updated to the second recommended amount, and a second dietary recommendation result is generated. The second dietary recommendation result is sent to the target group.
[0007] In some optional implementations, the food characteristic data includes a target spiciness value, and determining the first dietary recommendation result based on the current oral cavity status data and the food characteristic data includes: The hardness threshold and acid content threshold are predicted respectively for the current oral cavity state data to obtain the hardness threshold and acid content threshold corresponding to the current oral cavity state. The pain index is predicted by analyzing the current oral cavity status data and the target spiciness value. Food ingredients that meet preset conditions are identified as the first recommended food ingredients; the preset conditions are that the hardness value is less than the hardness threshold, the acid content is less than the acid content threshold, and the pain index is less than the pain index threshold. Obtain first preset mapping information, which represents the mapping relationship between ingredients and recommended quantities; Based on the first preset mapping information, obtain the first recommended quantity corresponding to the first recommended ingredient; The first recommended ingredient and the first recommended amount are used as the first dietary recommendation result.
[0008] In some optional implementations, the step of predicting hardness thresholds and acid content thresholds on the current oral cavity state data to obtain the hardness thresholds and acid content thresholds corresponding to the current oral cavity state includes: Obtain a first target model; the first target model is obtained by training a first preset model to predict hardness thresholds based on historical oral cavity state data and the hardness thresholds corresponding to the historical oral cavity state data. Input the current oral cavity state data into the first target model to obtain the hardness threshold corresponding to the current oral cavity state; Obtain a second target model; the second target model is obtained by training a second preset model to predict the acid content threshold based on historical oral cavity status data and the acid content threshold corresponding to the historical oral cavity status data. The current oral cavity state data is input into the second target model to obtain the acid content threshold corresponding to the current oral cavity state.
[0009] In some optional implementations, the current oral cavity status data includes the damaged area and the oral cavity area. The step of predicting the pain index based on the current oral cavity status data and the target spiciness value to obtain the pain index includes: Obtain the third target model; the third target model is obtained by training the third preset model to predict the pain index based on historical oral state data, historical spiciness value and the pain index corresponding to the historical spiciness value. The current oral cavity status data and the target spiciness value are input into the third target model. Based on the third target model, the ratio of the damaged area to the oral cavity area is determined as the target oral cavity damage proportion. Based on the second preset mapping information, the target sensitivity coefficient corresponding to the target oral cavity damage proportion is obtained. Based on the third preset mapping information, the target basic pain index corresponding to the target spiciness value is obtained. The product of the target sensitivity coefficient and the target basic pain index is determined as the pain compensation index. The basic pain index and the pain compensation index are summed to obtain the pain index. The second preset mapping information represents the mapping relationship between the proportion of oral lesions and the sensitivity coefficient; the third preset mapping information represents the mapping relationship between the spiciness value and the basic pain index.
[0010] In some optional implementations, determining the nutritional impairment coefficient based on the current oral condition data includes: Obtain the fourth target model; the fourth target model is obtained by predicting the digestive enzyme activity decay rate based on historical oral cavity state data and the digestive enzyme activity decay rate corresponding to the historical oral cavity state data. The current oral cavity status data is input into the fourth target model to obtain the digestive enzyme activity decay rate; The nutritional damage coefficient is obtained by weighted summation of the proportion of oral lesions and the rate of decline of digestive enzyme activity.
[0011] In some optional implementations, determining whether the first recommended dose is met based on the first recommended dose and the nutritional impairment coefficient includes: The product of the first recommended dose and the nutritional injury coefficient is determined as the nutritional injury amount; The difference between the first recommended dose and the amount of nutritional damage is determined as the actual absorption amount; Obtain the preset nutritional requirements; If the actual absorption is greater than or equal to the nutritional requirement, it is determined that the first recommended amount has been met. If the actual absorption is less than the nutritional requirement, the first recommended amount is determined to be insufficient.
[0012] In some optional implementations, the step of determining a second recommended amount based on the nutritional impairment coefficient when the first recommended amount is not met, updating the first recommended amount to the second recommended amount, and generating a second dietary recommendation result includes: If the first recommended amount is not met, obtain the threshold of the nutritional damage coefficient. The difference between the nutritional damage coefficient threshold and the nutritional damage coefficient is determined as the calculation base; The quotient of the nutritional requirement and the calculation base is determined as the second recommended amount; The first recommended amount in the first dietary recommendation result is updated to the second recommended amount to generate the second dietary recommendation result.
[0013] On the other hand, this application provides a dietary recommendation system, which includes a data acquisition module, a matching processing module, a dynamic compensation module, and a data transmission module. The data acquisition module is used to collect current oral cavity status data and food ingredient characteristic data; The matching processing module is used to determine a first dietary recommendation result based on the current oral cavity status data and the food ingredient characteristic data, wherein the first dietary recommendation result includes a first recommended dose; The dynamic compensation module is used to determine the nutritional damage coefficient based on the current oral condition data; and to determine whether the first recommended amount is up to standard based on the first recommended dose and the nutritional damage coefficient; and to determine the second recommended amount based on the nutritional damage coefficient when the first recommended amount is not up to standard, update the first recommended dose to the second recommended amount, and generate a second dietary recommendation result. The data transmission module is used to send the second dietary recommendation result to the target object.
[0014] On the other hand, this application provides an electronic device including a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the dietary recommendation method as described in any of the above embodiments.
[0015] On the other hand, this application provides a computer-readable storage medium storing at least one instruction or at least one program, which is loaded and executed by a processor to implement the dietary recommendation method as described in any of the above embodiments.
[0016] The dietary recommendation method provided in this application is characterized by comprising: collecting current oral cavity status data and food characteristic data; determining a first dietary recommendation result based on the current oral cavity status data and the food characteristic data, wherein the first dietary recommendation result includes a first recommended dose; determining a nutritional impairment coefficient based on the current oral cavity status data; determining whether the first recommended dose is adequate based on the first recommended dose and the nutritional impairment coefficient; if the first recommended dose is inadequate, determining a second recommended dose based on the nutritional impairment coefficient, updating the first recommended dose to the second recommended dose, and generating a second dietary recommendation result; and sending the second dietary recommendation result to the target object. By matching the dietary recommendation result with oral cavity status data and food characteristic data, and adjusting the dietary recommendation result based on the nutritional impairment coefficient, a multimodal perception and dynamic matching intelligent dietary recommendation system is realized to meet the refined needs of oral health management and improve the accuracy and adaptability of personalized dietary recommendations. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a flowchart of a dietary recommendation method according to an embodiment of the present invention; Figure 2 This is a structural diagram of a dietary recommendation system proposed according to an embodiment of the present invention. Detailed Implementation
[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.
[0020] The term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of this application. In the description of this application, it should be understood that the terms "upper," "lower," "top," "bottom," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. Moreover, the terms "first," "second," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein.
[0021] When a numerical range is disclosed herein, the range is considered continuous and includes the minimum and maximum values of the range, as well as every value between the minimum and maximum values. Furthermore, when the range refers to an integer, it includes every integer between the minimum and maximum values of the range. Additionally, when multiple ranges are provided to describe a feature or characteristic, the ranges may be combined. In other words, unless otherwise specified, all ranges disclosed herein should be understood to include any and all subranges to which they are included. For example, a specified range from “1 to 10” should be considered to include any and all subranges between the minimum value 1 and the maximum value 10. Exemplary subranges of the range 1 to 10 include, but are not limited to, 1 to 6.1, 3.5 to 7.8, 5.5 to 10, etc.
[0022] Because current dietary recommendation methods have significant shortcomings in terms of objectivity of oral condition perception, accuracy of food matching, and dynamics of nutritional adaptation, this application provides a dietary recommendation method, system, electronic device, and storage medium to achieve a multimodal perception and dynamic matching intelligent dietary recommendation system, thereby meeting the refined needs of oral health management and improving the accuracy and adaptability of personalized dietary recommendations.
[0023] Please see Figure 1 , Figure 1 This is a flowchart of a dietary recommendation method according to an embodiment of the present invention. In one aspect, this application provides a dietary recommendation method, which includes: S101. Collect current oral cavity status data and food ingredient characteristic data.
[0024] Optionally, the current oral condition data can include various information, such as ulcer area, oral cavity area, salivary pH, occlusal force distribution, and ulcer healing stage. Multispectral imaging can acquire data such as ulcer area, oral cavity area, and ulcer healing stage to quantify the degree of oral damage and healing process, providing a basis for assessing the risk of food irritation to the oral cavity. The multispectral imager uses a 620 / 850nm dual-wavelength light source, utilizing the difference in reflectance spectrum to distinguish normal mucosa from ulcer areas. Image processing algorithms then identify ulcer boundaries and calculate the ratio of ulcer area to the total oral mucosa area to obtain the ulcer area percentage. A miniature pH sensor can monitor saliva pH in real time, converting the pH of saliva in the oral cavity into an electrical signal for detection, reflecting the oral cavity's acid-base environment and helping to select suitable foods, avoiding excessively acidic or alkaline foods that could irritate the oral mucosa. A piezoresistive thin-film sensor array can measure the distribution of biting force. With a resolution of 0.1N, the piezoresistive thin-film sensor array utilizes the piezoresistive effect; when subjected to biting force, its resistance changes, thereby assessing the patient's chewing ability and recommending foods suitable for chewing function.
[0025] Optionally, the food ingredient properties data include both mechanical and chemical properties. Mechanical properties data include crushing work, viscoelasticity, etc. Viscoelastic parameters are measured using parameters such as hardness and elasticity to match the patient's chewing ability and oral recovery stage. A universal testing machine is used to measure the breaking work of the food by applying a certain force to break it, recording the relationship between force and displacement, and then calculating the breaking work. Viscoelastic parameters are obtained using a dynamic mechanical analyzer. This forms a set of oral state parameters and establishes a quantitative matrix for the hardness and elasticity of the food.
[0026] Optionally, chemical property data include acidity (pH value), capsaicin concentration (Scoville units), citric acid content, etc. These data are used to assess the degree of irritation of food to the oral mucosa, avoid recommending food that may cause damage to the oral mucosa or delay healing, and build a chemical composition database.
[0027] By collecting data from multiple dimensions, subjective descriptions are transformed into objective quantifications, providing input for subsequent algorithms. Furthermore, through models and real-time compensation algorithms, the accuracy and dimensionality of oral health assessment are improved.
[0028] S102. Based on the current oral cavity status data and the food ingredient characteristic data, determine a first dietary recommendation result, which includes a first recommended dose.
[0029] In an optional embodiment, the food characteristic data includes a target spiciness value, and determining the first dietary recommendation result based on the current oral cavity status data and the food characteristic data includes: The hardness threshold and acid content threshold are predicted respectively for the current oral cavity state data to obtain the hardness threshold and acid content threshold corresponding to the current oral cavity state. The pain index is predicted by analyzing the current oral cavity status data and the target spiciness value. Food ingredients that meet preset conditions are identified as the first recommended food ingredients; the preset conditions are that the hardness value is less than the hardness threshold, the acid content is less than the acid content threshold, and the pain index is less than the pain index threshold. Obtain first preset mapping information, which represents the mapping relationship between ingredients and recommended quantities; Based on the first preset mapping information, obtain the first recommended quantity corresponding to the first recommended ingredient; The first recommended ingredient and the first recommended amount are used as the first dietary recommendation result.
[0030] Optionally, a model matching dietary recommendations is trained. Using this trained model, the safety boundaries of the mechanical and chemical properties of food ingredients are determined based on current oral condition data. The system acquires a first target model trained based on historical oral condition data and corresponding hardness thresholds. For example, during the acute phase of an ulcer, the model outputs a suitable hardness threshold based on the patient's chewing ability and food suitability in historical data for that stage, such as recommending liquid foods with a crushing effort <0.5 J / g to avoid increasing the chewing burden with high-hardness foods. Alternatively, the system acquires a second target model trained based on historical oral condition data and corresponding acid content thresholds. Combining information such as saliva pH from the current oral condition data, the model predicts acid content thresholds. For example, when saliva pH <5.5, to prevent acid corrosion of the mucosa, the model's output acid content threshold might be citric acid content <2%, thus selecting ingredients with suitable acidity.
[0031] Optionally, the system invokes a third-objective model trained based on historical oral condition data, historical spiciness values, and corresponding pain indices to derive the pain index. Based on the prediction results, ingredients meeting preset conditions are selected. These preset conditions are: hardness value less than a hardness threshold, acid content less than an acid content threshold, and pain index less than a pain index threshold. For example, in the acute phase of an ulcer, if an ingredient has a crushing effort of 0.4 J / g, a citric acid content of 1.5%, and a pain index of 3, it might be identified as the first recommended ingredient. In this case, the hardness threshold is 0.5 J / g, the acid content threshold is 2%, and the pain index threshold is 4.
[0032] Optionally, the system acquires first preset mapping information, which is based on dietary nutrition guidelines, individual user characteristics, and health status, and represents the mapping relationship between food ingredients and recommended amounts. For example, the recommended daily protein intake for adult women is 55g. If a recommended food, such as tofu, contains 8g of protein per 100g, the corresponding amount can be extracted from the mapping information based on the mapping relationship and the nutritional density of the food. The first recommended food ingredients that meet the conditions and their corresponding first recommended amounts are integrated to form the first dietary recommendation result, providing users with preliminary dietary advice. Subsequent recommendations will be further optimized by incorporating factors such as the nutritional impairment coefficient, focusing on matching oral condition with food characteristics to ensure that the recommended food ingredients are suitable for the current oral condition and meet nutritional needs, achieving initial adaptation between food ingredients and oral condition, and laying the foundation for subsequent dynamic optimization based on the nutritional impairment coefficient.
[0033] In an optional embodiment, the step of predicting a hardness threshold and an acid content threshold from the current oral cavity state data to obtain the hardness threshold and acid content threshold corresponding to the current oral cavity state includes: Obtain a first target model; the first target model is obtained by training a first preset model to predict hardness thresholds based on historical oral cavity state data and the hardness thresholds corresponding to the historical oral cavity state data. Input the current oral cavity state data into the first target model to obtain the hardness threshold corresponding to the current oral cavity state; Obtain a second target model; the second target model is obtained by training a second preset model to predict the acid content threshold based on historical oral cavity status data and the acid content threshold corresponding to the historical oral cavity status data. The current oral cavity state data is input into the second target model to obtain the acid content threshold corresponding to the current oral cavity state.
[0034] Optionally, using historical oral cavity condition data as input features and the corresponding hardness threshold as the output label, a first preset model is trained. Through iterative optimization of parameters and adjustment of feature weights, the model can accurately predict the hardness threshold based on the oral cavity condition. After training, this model becomes the first target model. The system loads the trained first target model; the current oral cavity condition data includes ulcer healing stage, occlusal force distribution, and ulcer area percentage; the first target model outputs the corresponding hardness threshold by comparing the feature similarity between the current data and historical training samples. For example, the output might be "breaking energy < 0.5 J / g," meaning that under the current oral cavity condition, only foods with a hardness below this threshold are recommended, while high-hardness foods should be avoided.
[0035] Optionally, using historical oral condition data as input and the corresponding acid content threshold as the output label, the first preset model is trained, enabling the model to predict safe acid content boundaries based on the current oral pH environment and damage state. After training, this model becomes the "second target model." The system loads the trained second target model; inputting current oral condition data includes saliva pH and ulcer area percentage, the second target model outputs the corresponding acid content threshold by matching the current acidic environment and damage state. For example, the output might be "citric acid content < 2%," meaning that under the current oral condition, foods with a citric acid content exceeding 2% should be avoided to prevent acid corrosion of the ulcer mucosa. The first recommended diet result combines the outputs of the first and second target models, outputting foods that simultaneously meet the threshold boundaries of mechanical and chemical properties.
[0036] The first and second target models transform empirical rules relating oral cavity condition and food safety thresholds into quantifiable machine learning models, resolving the ambiguity of traditional subjective judgments. Through the hardness threshold and acid content threshold output by the models, dual quantitative standards of mechanical and chemical safety are provided for subsequent food selection, ensuring that recommended foods are suitable for the current chewing ability and mucosal tolerance of the oral cavity, thus forming an accurate preliminary recommendation boundary.
[0037] In an optional embodiment, the current oral cavity status data includes the damaged area and the oral cavity area. The step of predicting the pain index based on the current oral cavity status data and the target spiciness value to obtain the pain index includes: Obtain the third target model; the third target model is obtained by training the third preset model to predict the pain index based on historical oral state data, historical spiciness value and the pain index corresponding to the historical spiciness value. The current oral cavity status data and the target spiciness value are input into the third target model. Based on the third target model, the ratio of the damaged area to the oral cavity area is determined as the target oral cavity damage proportion. Based on the second preset mapping information, the target sensitivity coefficient corresponding to the target oral cavity damage proportion is obtained. Based on the third preset mapping information, the target basic pain index corresponding to the target spiciness value is obtained. The product of the target sensitivity coefficient and the target basic pain index is determined as the pain compensation index. The basic pain index and the pain compensation index are summed to obtain the pain index. The second preset mapping information represents the mapping relationship between the proportion of oral lesions and the sensitivity coefficient; the third preset mapping information represents the mapping relationship between the spiciness value and the basic pain index.
[0038] Optionally, the multispectral imager emits a dual-wavelength light source. The 620nm light is absorbed by hemoglobin, while the 850nm light penetrates the tissue. The difference in reflectance spectrum distinguishes between normal mucosa and ulcer areas. The algorithm automatically segments the ulcer boundary, calculates the pixel area, and calculates the ratio with the oral mucosa baseline area to obtain the area proportion.
[0039] Optionally, the third preset model is trained using historical oral condition data and historical spiciness values as input features, and the corresponding pain index, i.e., VAS score, as the output label. By iteratively optimizing the parameters and adjusting the weights of the damage proportion and spiciness value, the model fits the relationship between the degree of damage and spiciness and the pain index. For example, under the same spiciness, the larger the damage area, the higher the predicted pain index; under the same damage, the higher the spiciness, the higher the pain index. After training, this model becomes the third target model, which is embedded in the system for real-time calculation of the pain index.
[0040] Optionally, the two preset mapping information are mapping tables of association rules between the proportion of oral lesions and sensitivity coefficients pre-stored in the system, established based on historical clinical data. Inputting the lesion area and food irritation parameters, the model outputs the pain index of the food on the oral mucosa. The model establishes a dose-response relationship between lesion area and food irritation using multimodal data (occlusal force, spectral imaging, pH value, etc.). For example, when the ulcer area exceeds 10%, the irritation risk of high-acidity foods (pH < 5.5) significantly increases. The model needs to nonlinearly fit this threshold effect; the constructed lesion area-food irritation regression model is a nonlinear regression model, and the R-squared value of the model is... 2 =0.89, R 2 The coefficient of determination measures how well the regression model fits the observed data. It ranges from 0 to 1. The closer the coefficient of determination is to 1, the smaller the deviation between the model's predicted values and the actual observed values, and the better the fit. This indicates that the model can explain 89% of the variable fluctuations and has a good fit.
[0041] Optionally, as shown in Table 1, the third preset mapping information is a mapping table of association rules between spiciness values (Scoville units) and baseline pain index (VAS score) pre-stored in the system. It collects subjective pain scores from a large number of patients after consuming foods with different capsaicin concentrations, establishes a statistical association between the two based on the Scoville units of the foods, and builds upon historical data on the stimuli responses of healthy oral cavity patients to different levels of spiciness. This information is then stored in the system database as the basis for real-time matching rules. The third target model queries the third preset mapping information based on the input target spiciness value and outputs the corresponding target baseline pain index. For example, when the spiciness value is 1000 Scoville units, the baseline pain index is 3 points.
[0042] Table 1
[0043] Optionally, the system dynamically adjusts the pain index of the preset mapping table based on the user's current ulcer detection results. For example, 1000 Scoville units in a healthy oral cavity corresponds to a VAS score of 2. If the user has large-area ulcers, the system will adjust the pain index to 5 points using a "damage area-food irritation regression model." Simultaneously, the user's real-time feedback is incorporated into a historical database, and the mapping table is periodically iterated to improve matching accuracy.
[0044] Optionally, a regression model of damage area versus food irritation (R²) can be used. 2 =0.89) Calculate the sensitivity coefficient. Actual pain index = base pain index * (1 + sensitivity coefficient). If the pain index threshold is 4 points, the food will be blocked due to excessive stimulation risk.
[0045] Optionally, the matching results can be combined with the above model to determine the threshold of mechanical properties of food ingredients based on biting force and ulcer stage. Among mechanically qualified food ingredients, chemically safe food ingredients are screened through a mapping table and acidity rules. Subsequently, nutrient absorption compensation calculations are performed on the remaining food ingredients to finally generate a recommendation list. Taking into account both mechanical and chemical properties improves the matching accuracy from 68% in traditional methods to 93%, and supports gradual adjustment of postoperative diet in 0.5J / g increments, which avoids mechanical damage and prevents chemical stimulation from delaying healing.
[0046] Optionally, mechanical property recommendations and chemical composition parameter recommendations are synergistically matched through cross-validation: mechanical properties ensure the chewability of the food, while chemical composition parameters control the risk of irritation. Together, they constitute a "dual screening mechanism" to ultimately generate a recommended plan that takes into account both oral health and nutritional needs.
[0047] The third objective model is a tool for quantifying the synergistic stimulation effect between the degree of oral injury and the spiciness of food: it trains and solidifies mapping rules through historical data, and combines the current proportion of oral injury, sensitivity coefficient and basic pain index of food spiciness response during real-time calculation, and finally outputs a pain index that reflects the individual's real-time stimulation risk, providing a quantitative standard for chemical safety in food selection.
[0048] S103. Determine the nutritional damage coefficient based on the current oral condition data.
[0049] In an optional embodiment, determining the nutritional impairment coefficient based on the current oral cavity status data includes: Obtain the fourth target model; the fourth target model is obtained by predicting the digestive enzyme activity decay rate based on historical oral cavity state data and the digestive enzyme activity decay rate corresponding to the historical oral cavity state data. The current oral cavity status data is input into the fourth target model to obtain the digestive enzyme activity decay rate; The nutritional damage coefficient is obtained by weighted summation of the proportion of oral lesions and the rate of decline of digestive enzyme activity.
[0050] Optionally, a predictive model is established based on the correlation data between the degree of oral lesions and digestive enzyme activity in historical cases. This model uses historical oral condition data as input features and the corresponding digestive enzyme activity decay rate as the output label to train the fourth preset model. By iteratively optimizing the model parameters and adjusting the weights of ulcer area proportion and saliva pH, the model can accurately predict the digestive enzyme activity decay rate based on oral condition. After training, this model becomes the fourth target model, which is embedded in the system and used to derive the digestive enzyme activity decay rate from current oral condition data in real time.
[0051] Optionally, the current oral cavity status data is input into the fourth target model. Based on the trained association rules, the model outputs the digestive enzyme activity decay rate, that is, the proportion of digestive enzyme function decline, ranging from 0 to 1, where 0 indicates no decay and 1 indicates complete inactivation.
[0052] Optionally, the nutritional damage coefficient can be abstracted as a function of the ulcer area ratio and the digestive enzyme activity decay rate, obtained by weighted summation of the oral lesion ratio and the digestive enzyme activity decay rate, as shown in the formula: In this formula, k is the nutritional damage coefficient, and w_1 and w_2 are weighting coefficients. The sum of w_1 and w_2 is 1, derived from historical clinical data, reflecting the weighting of their influence on nutrient absorption. The k value is calculated by a weighted combination of the ulcer area ratio and digestive enzyme activity. The specific weights and function form are derived from clinical data training. Its core function is to quantify the degree of impact of oral damage on nutrient absorption, providing key parameters for calculating actual absorption. Furthermore, the k value is not preset but dynamically calculated using real-time monitored ulcer area ratio and digestive enzyme activity, combined with a weighted formula. This ensures that each calculation result closely matches the current oral physiological state, achieving precise and personalized nutrient absorption compensation.
[0053] The fourth objective model transforms oral condition into a quantitative tool for the decay rate of digestive enzyme activity. Combined with the weighted calculation of the proportion of oral damage, the final nutritional damage coefficient objectively reflects the total impact of oral damage on nutrient absorption, providing key parameters for subsequent calculation of actual absorption and adjustment of recommended intake.
[0054] S104. Based on the first recommended dose and the nutritional damage coefficient, determine whether the first recommended amount meets the standard.
[0055] In an optional embodiment, determining whether the first recommended dose is met based on the first recommended dose and the nutritional impairment coefficient includes: The product of the first recommended dose and the nutritional injury coefficient is determined as the nutritional injury amount; The difference between the first recommended dose and the amount of nutritional damage is determined as the actual absorption amount; Obtain the preset nutritional requirements; If the actual absorption is greater than or equal to the nutritional requirement, it is determined that the first recommended amount has been met. If the actual absorption is less than the nutritional requirement, the first recommended amount is determined to be insufficient.
[0056] Alternatively, oral damage can lead to a decrease in nutrient absorption efficiency. The actual amount absorbed is quantified by the damage coefficient k to ensure that the effective nutrient intake of the recommended diet meets physiological needs. The actual amount absorbed directly affects the recommended portion size, type, and cooking method of the ingredients, so that the recommended plan shifts from theoretical nutrient requirements to individualized actual absorbable nutrients, thereby improving the accuracy of the dietary plan.
[0057] Optionally, the preset nutritional requirements are based on basic information such as the user's age, gender, weight, height, activity level, and health status. The daily essential nutrient intake is determined with reference to authoritative dietary guidelines such as the Chinese DRIs. The baseline value is adjusted according to the user's health goals such as weight gain, weight loss, and postoperative recovery. This calculation method ensures that the recommended plan not only meets universal nutritional standards but also adapts to the special needs caused by oral damage, providing a scientific basis for compensating for actual absorption. Optionally, the amount of nutrient loss due to oral injury can be calculated according to the formula "Nutritional damage amount = First recommended dose * Nutritional damage coefficient"; and the amount of nutrients that the user can actually absorb can be obtained according to the formula "Actual absorption amount = First recommended dose - Nutritional damage amount", which is the theoretical intake amount minus the damaged part.
[0058] Optionally, the system can call pre-stored personalized nutritional needs data of users. If the actual absorption is greater than or equal to the preset nutritional needs, the first recommended amount meets the user's needs, i.e., it meets the standard; if the actual absorption is less than the preset nutritional needs, the first recommended amount is insufficient and does not meet the standard, so it needs to be adjusted.
[0059] This process uses quantitative calculations to solve the problem that traditional recommendations only consider theoretical intake and ignore absorption loss caused by oral damage. If the actual absorption of the first recommended dose does not meet the requirements, the system will trigger an adjustment mechanism to ensure effective nutrient intake.
[0060] S105. If the first recommended amount is not met, a second recommended amount is determined based on the nutritional impairment coefficient, the first recommended amount is updated to the second recommended amount, and a second dietary recommendation result is generated.
[0061] In an optional embodiment, the step of determining a second recommended amount based on the nutritional impairment coefficient when the first recommended amount is not met, updating the first recommended amount to the second recommended amount, and generating a second dietary recommendation result includes: If the first recommended amount is not met, obtain the threshold of the nutritional damage coefficient. The difference between the nutritional damage coefficient threshold and the nutritional damage coefficient is determined as the calculation base; The quotient of the nutritional requirement and the calculation base is determined as the second recommended amount; The first recommended amount in the first dietary recommendation result is updated to the second recommended amount to generate the second dietary recommendation result.
[0062] Optionally, the nutritional impairment coefficient threshold is pre-stored in the system. This threshold is set based on clinical safety data and is used to limit abnormal recommended intakes caused by overcompensation. The nutritional impairment coefficient threshold is the maximum value of the nutritional impairment coefficient and can be set to 1. If the first recommended intake is not met, the system will work backward to increase the amount of recommended food. For example, if the original recommendation is 100 grams of steamed egg containing 100 micrograms of vitamin B12, since k=0.4, the actual absorption is 60 micrograms. To meet the 200 microgram requirement, the system will adjust the recommended intake to 333 grams, or combine it with other low-irritant, high-B12 foods such as low-fat cheese. The system will replace the first recommended intake in the first dietary recommendation result with the second recommended intake to ensure that the food still meets the safety thresholds for mechanical and chemical properties, forming the final recommendation scheme.
[0063] Optionally, after the ingredients have passed the initial screening, a recommended plan can be generated through multi-dimensional scoring, ranging from 0 to 10 points. Specifically, a mechanical property score is obtained based on the match between the ingredient's crushing ability and the current chewing force, used to adapt to chewing ability; a chemical property score is obtained based on the pain index corresponding to the acidity and capsaicin concentration of the ingredients, used to avoid irritation risks; and a nutritional compensation score is obtained based on the actual absorption rate of the ingredients, used to ensure effective absorption. The scores of the three dimensions are added together. Ingredient combinations with high total scores are recommended first, sorted from highest to lowest.
[0064] Optionally, user feedback after consuming recommended recipes will be recorded and used to optimize the weighting of various dimensions of the rating. For example, if a user reports that a certain ingredient has a high mechanical property score but is actually uncomfortable to chew, the mechanical property score of that type of ingredient will be lowered next time to continuously improve the accuracy of the ranking.
[0065] Optionally, based on actual absorption needs, the system prioritizes recommending foods with high nutrient density and matching mechanical properties. For example, when it is necessary to compensate for protein absorption, tofu with a crushing energy of <0.5J / g is recommended instead of hard meat. Under the premise of meeting absorption needs, highly stimulating foods are avoided. For example, if it is necessary to increase vitamin C intake, the system prioritizes recommending soft-cooked broccoli instead of citrus fruits.
[0066] Optionally, if actual absorption remains insufficient, further optimization based on hardware data can trigger a hardware linkage protocol to adjust the dietary pattern. For example, if the implant pressure sensor detects high-frequency vibration, indicating limited chewing function, a liquid diet mode can be automatically activated, such as replacing solid foods with fruit and vegetable juices to improve absorption efficiency. Furthermore, the intelligent tableware temperature control can adjust the food temperature to below 50°C to avoid high-temperature irritation of the oral mucosa, while simultaneously promoting digestive enzyme activity, indirectly increasing absorption. Compared to the uncompensated system, vitamin B12 absorption efficiency increased by 41%, and the hardware response latency was less than 50ms, meeting real-time control requirements.
[0067] Optionally, when hot food is detected, the k value is temporarily increased by 0.1 to remind the user to eat it after cooling or to automatically adjust the recommended recipe.
[0068] Optionally, the system generates a visual report containing the actual absorption amount and recommended adjustment criteria for users and doctors to refer to; users' dietary feedback will also be incorporated into the model, dynamically updating the k-value calculation rules and continuously optimizing the accuracy of absorption prediction.
[0069] Optionally, hardware devices provide real-time feedback on oral physiological status, directly impacting k-value calculation and actual absorption. The calculated nutrient absorption results drive the hardware to perform adaptive actions, which in turn affect the oral condition, improving absorption efficiency. Together, they serve a dynamic balance between oral health and nutrient absorption, ensuring that the recommended plan conforms to physiological constraints while achieving effective nutrient intake. Furthermore, through quantitative segmentation, collaborative screening, and dynamic iteration, the system addresses the problems of traditional systems such as single-dimensionality, insufficient precision, and imbalances in safety and nutrition. Ultimately, it achieves personalized dietary recommendations that adapt to oral function, ensure effective nutrient intake, and reduce the risk of irritation. The weights of the scoring dimensions can be dynamically adjusted according to oral condition to accommodate individual oral differences, and user feedback will in turn affect the scoring model, continuously optimizing the personalized experience.
[0070] S106. Send the second dietary recommendation result to the target object.
[0071] Optionally, a fully homomorphic encryption (FHE) algorithm is used to encrypt sensitive data. Unlike traditional encryption, which only encrypts data transmission, FHE allows calculations such as ulcer area ratio calculation and pain index derivation to be performed directly in encrypted form. Data processing can be completed without decryption, avoiding the risk of data leakage during the decryption and re-encryption process. Specifically, when ulcer images acquired by the multispectral imager and occlusal force data recorded by the piezoresistive sensor are transmitted to the system, the FHE algorithm encrypts the data in real time, ensuring that the data exists in encrypted form throughout the entire process of data transmission, model calculation, and storage. Only authorized users can decrypt and view the results using a key.
[0072] Optionally, a role-based access control (RBAC) model can be used to hierarchically anonymize data fields. Different data access permissions are assigned based on user roles, and sensitive fields are dynamically masked or obfuscated, while non-sensitive fields remain visible. Specifically, when patients view recommended plans, the system only displays the recipe content, hiding the original oral imaging data; when doctors make diagnoses, they can view the complete images but cannot obtain patient identification information; system administrators can only access data transmission logs and cannot access specific medical content, achieving the goal of data usability without visibility.
[0073] Encrypted data is processed using the FHE algorithm, allowing it to be directly used in core processes such as multimodal oral condition analysis and food matching model calculations. The data value is fully preserved, ensuring that system functions are not affected by encryption. At the same time, FHE encryption meets data transmission encryption standards, and combined with RBAC dynamic desensitization, it achieves the pseudonymization and de-identification processing required by GDPR, complying with international medical data security standards, with zero vulnerabilities.
[0074] Please see Figure 2 , Figure 2 This is a structural diagram of a dietary recommendation system according to an embodiment of the present invention. On the other hand, this application provides a dietary recommendation system, which includes a data acquisition module 201, a matching processing module 202, a dynamic compensation module 203, and a data transmission module 204. The data acquisition module 201 is used to collect current oral cavity status data and food ingredient characteristic data; The matching processing module 202 is used to determine a first dietary recommendation result based on the current oral cavity status data and the food ingredient characteristic data, wherein the first dietary recommendation result includes a first recommended dose; The dynamic compensation module 203 is used to determine the nutritional damage coefficient based on the current oral condition data; and to determine whether the first recommended amount is up to standard based on the first recommended dose and the nutritional damage coefficient; and to determine the second recommended amount based on the nutritional damage coefficient when the first recommended amount is not up to standard, update the first recommended dose to the second recommended amount, and generate a second dietary recommendation result. The data transmission module 204 is used to send the second dietary recommendation result to the target object.
[0075] The dietary recommendation method provided in this application is characterized by comprising: collecting current oral cavity status data and food characteristic data; determining a first dietary recommendation result based on the current oral cavity status data and the food characteristic data, wherein the first dietary recommendation result includes a first recommended dose; determining a nutritional impairment coefficient based on the current oral cavity status data; determining whether the first recommended amount is met based on the first recommended dose and the nutritional impairment coefficient; determining a second recommended amount based on the nutritional impairment coefficient if the first recommended amount is not met, updating the first recommended dose to the second recommended amount, generating a second dietary recommendation result; and sending the second dietary recommendation result to the target object. The dietary recommendation method provided in this application has the following beneficial effects: (1) By using the damage area-food irritation regression model, the relationship between ulcer area and food mechanical and chemical stimulation is quantified, providing a scientific basis for subsequent matching, breaking through the subjective limitations of oral state perception, and achieving objective quantification and accurate assessment. (2) By constructing a precise matching model between food ingredients and oral condition, cross-validation of mechanical and chemical properties can be achieved, avoiding oral damage caused by single-dimensional assessment and improving the safety and suitability of recommendations. (3) The recommended amount is adjusted in a targeted manner through the nutrient absorption compensation algorithm, and the implant pressure sensor and smart tableware temperature control data are received to realize the hardware closed-loop linkage to correct the recommendation scheme in real time. The pain feeling and digestion feedback of the user after eating are included in the model training, so that the recommendation accuracy is continuously improved with use, and multi-dimensional collaborative adaptation ensures effective nutrient intake. (4) Homomorphic encryption is used to implement ciphertext operations, and dynamic desensitization strategy based on RBAC is used to ensure compliance of the entire data process and balance security and availability.
[0076] In an optional embodiment, this application provides an electronic device including a processor and a memory, the memory storing at least one instruction or at least one program, the at least one instruction or at least one program being loaded and executed by the processor to implement the dietary recommendation method as described in any of the above embodiments.
[0077] In an optional embodiment, this application provides a computer-readable storage medium storing at least one instruction or at least one program segment, which is loaded and executed by a processor to implement the dietary recommendation method as described in any of the above embodiments.
[0078] The above description is only an optional embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.
Claims
1. A dietary recommendation method, characterized in that, The dietary recommendation methods include: Collect current oral cavity condition data and food ingredient characteristic data; Based on the current oral health data and the food ingredient characteristics data, a first dietary recommendation result is determined, which includes a first recommended dose. The nutritional damage coefficient is determined based on the current oral health data. Based on the first recommended dose and the nutritional damage coefficient, determine whether the first recommended dose meets the standard; If the first recommended amount is not met, a second recommended amount is determined based on the nutritional impairment coefficient, the first recommended amount is updated to the second recommended amount, and a second dietary recommendation result is generated. The second dietary recommendation result is sent to the target group.
2. The dietary recommendation method according to claim 1, characterized in that, The food ingredient characteristic data includes a target spiciness value. Determining the first dietary recommendation result based on the current oral cavity status data and the food ingredient characteristic data includes: The hardness threshold and acid content threshold are predicted respectively for the current oral cavity state data to obtain the hardness threshold and acid content threshold corresponding to the current oral cavity state. The pain index is predicted by analyzing the current oral cavity status data and the target spiciness value. Food ingredients that meet preset conditions are identified as the first recommended food ingredients; the preset conditions are that the hardness value is less than the hardness threshold, the acid content is less than the acid content threshold, and the pain index is less than the pain index threshold. Obtain first preset mapping information, which represents the mapping relationship between ingredients and recommended quantities; Based on the first preset mapping information, obtain the first recommended quantity corresponding to the first recommended ingredient; The first recommended ingredient and the first recommended amount are used as the first dietary recommendation result.
3. The dietary recommendation method according to claim 2, characterized in that, The step of predicting hardness threshold and acid content threshold based on the current oral cavity state data to obtain the hardness threshold and acid content threshold corresponding to the current oral cavity state includes: Obtain a first target model; the first target model is obtained by training a first preset model to predict hardness thresholds based on historical oral cavity state data and the hardness thresholds corresponding to the historical oral cavity state data. Input the current oral cavity state data into the first target model to obtain the hardness threshold corresponding to the current oral cavity state; Obtain a second target model; the second target model is obtained by training a second preset model to predict the acid content threshold based on historical oral cavity status data and the acid content threshold corresponding to the historical oral cavity status data. The current oral cavity state data is input into the second target model to obtain the acid content threshold corresponding to the current oral cavity state.
4. The dietary recommendation method according to claim 2, characterized in that, The current oral cavity status data includes the damaged area and the oral cavity area. The step of predicting the pain index based on the current oral cavity status data and the target spiciness value to obtain the pain index includes: Obtain the third target model; the third target model is obtained by training the third preset model to predict the pain index based on historical oral state data, historical spiciness value and the pain index corresponding to the historical spiciness value. The current oral cavity status data and the target spiciness value are input into the third target model. Based on the third target model, the ratio of the damaged area to the oral cavity area is determined as the target oral cavity damage proportion. Based on the second preset mapping information, the target sensitivity coefficient corresponding to the target oral cavity damage proportion is obtained. Based on the third preset mapping information, the target basic pain index corresponding to the target spiciness value is obtained. The product of the target sensitivity coefficient and the target basic pain index is determined as the pain compensation index. The basic pain index and the pain compensation index are summed to obtain the pain index. The second preset mapping information represents the mapping relationship between the proportion of oral lesions and the sensitivity coefficient; the third preset mapping information represents the mapping relationship between the spiciness value and the basic pain index.
5. The dietary recommendation method according to claim 4, characterized in that, Determining the nutritional impairment coefficient based on the current oral health data includes: Obtain the fourth target model; the fourth target model is obtained by predicting the digestive enzyme activity decay rate based on historical oral cavity state data and the digestive enzyme activity decay rate corresponding to the historical oral cavity state data. The current oral cavity status data is input into the fourth target model to obtain the digestive enzyme activity decay rate; The nutritional damage coefficient is obtained by weighted summation of the proportion of oral lesions and the rate of decline of digestive enzyme activity.
6. The dietary recommendation method according to claim 5, characterized in that, The step of determining whether the first recommended dose is met based on the first recommended dose and the nutritional injury coefficient includes: The product of the first recommended dose and the nutritional injury coefficient is determined as the nutritional injury amount; The difference between the first recommended dose and the amount of nutritional damage is determined as the actual absorption amount; Obtain the preset nutritional requirements; If the actual absorption is greater than or equal to the nutritional requirement, it is determined that the first recommended amount has been met. If the actual absorption is less than the nutritional requirement, the first recommended amount is determined to be insufficient.
7. The dietary recommendation method according to claim 6, characterized in that, The step of determining a second recommended amount based on the nutritional impairment coefficient when the first recommended amount is not met, updating the first recommended amount to the second recommended amount, and generating a second dietary recommendation result includes: If the first recommended amount is not met, obtain the threshold of the nutritional damage coefficient. The difference between the nutritional damage coefficient threshold and the nutritional damage coefficient is determined as the calculation base; The quotient of the nutritional requirement and the calculation base is determined as the second recommended amount; The first recommended amount in the first dietary recommendation result is updated to the second recommended amount to generate the second dietary recommendation result.
8. A dietary recommendation system, characterized in that, The dietary recommendation system includes a data acquisition module, a matching processing module, a dynamic compensation module, and a data transmission module; The data acquisition module is used to collect current oral cavity status data and food ingredient characteristic data; The matching processing module is used to determine a first dietary recommendation result based on the current oral cavity status data and the food ingredient characteristic data, wherein the first dietary recommendation result includes a first recommended dose; The dynamic compensation module is used to determine the nutritional damage coefficient based on the current oral cavity status data; And for determining whether the first recommended dose is met based on the first recommended dose and the nutritional damage coefficient; And, when the first recommended amount is not met, to determine a second recommended amount based on the nutritional impairment coefficient, update the first recommended amount to the second recommended amount, and generate a second dietary recommendation result; The data transmission module is used to send the second dietary recommendation result to the target object.
9. An electronic device, characterized in that, The electronic device includes a processor and a memory, the memory storing at least one instruction or at least one program, the at least one instruction or at least one program being loaded and executed by the processor to implement the dietary recommendation method as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one instruction or at least one program, which is loaded and executed by a processor to implement the dietary recommendation method as described in any one of claims 1-7.