Clinical nutrition virtual ward management system

Abstract

The invention relates to the field of intelligent management, and provides a clinical nutrition virtual ward management system which comprises the following steps: acquiring first nutrition consultation data before a to-be-evaluated nutrition scheme is used and second nutrition consultation data after the to-be-evaluated nutrition scheme is used; carrying out semantic coding on the nutrition scheme to be evaluated, the first nutrition consultation data and the second nutrition consultation data by adopting an AI-based natural language processing and analysis technology; and according to the difference change between the first nutrition consultation data and the second nutrition consultation data and the semantic interaction fusion feature between the to-be-evaluated nutrition schemes, intelligently judging whether the to-be-evaluated nutrition schemes are valid or not. Through the mode, the nutrition consultation data of the patient can be automatically processed and analyzed, and manual intervention is reduced, so that the intelligent level of clinical nutrition management is improved, and personalized nutrition suggestions can be provided for the patient.

Description

Technical Field

[0001] The present application relates to the field of intelligent management, and more specifically, to a clinical nutrition virtual ward management system. Background Art

[0002] As an important part of clinical medicine, clinical nutrition plays an increasingly important role in disease prevention, treatment and rehabilitation. Clinical nutrition management refers to the process of professional nutritional assessment, diagnosis and treatment of patients in a clinical medical environment, aiming to ensure that patients receive appropriate nutritional support to promote health, treat diseases and prevent malnutrition.

[0003] In traditional clinical nutrition practice, doctors and nutritionists need to determine whether a specific nutrition plan has effectively improved the patient's nutritional status. However, this evaluation process relies on manual analysis of the patient's medical records and nutrition consultation data, which is not only time-consuming but also may be affected by subjective judgment and is prone to omissions and misjudgments.

[0004] Therefore, a clinical nutrition virtual ward management system is desired. Summary of the invention

[0005] In response to the shortcomings in the prior art, the present application provides a clinical nutrition virtual ward management system, which obtains the first nutrition consultation data before the nutrition plan to be evaluated is used and the second nutrition consultation data after the nutrition plan to be evaluated is used, and uses AI-based natural language processing and analysis technology to semantically encode the nutrition plan to be evaluated, the first nutrition consultation data, and the second nutrition consultation data, so as to intelligently judge whether the nutrition plan to be evaluated is effective based on the differential changes between the first nutrition consultation data and the second nutrition consultation data and the semantic interaction fusion characteristics between the nutrition plan to be evaluated. In this way, the patient's nutrition consultation data can be automatically processed and analyzed, reducing manual intervention, thereby improving the intelligence level of clinical nutrition management and helping to provide patients with personalized nutrition advice.

[0006] According to one aspect of the present application, a clinical nutrition virtual ward management system is provided, which includes:

[0007] A first nutrition consultation data acquisition module, used to acquire first nutrition consultation data before the nutrition plan to be evaluated is used;

[0008] A second nutrition consultation data acquisition module, used to acquire second nutrition consultation data after using the nutrition plan to be evaluated;

[0009] A semantic coding module for the nutritional scheme to be evaluated, used for semantically coding the nutritional scheme to be evaluated to obtain a semantic coding feature vector of the nutritional scheme to be evaluated;

[0010] A first and second nutrition consultation data semantic encoding module, for performing semantic encoding on the first nutrition consultation data and the second nutrition consultation data respectively to obtain a first nutrition consultation data semantic encoding feature vector and a second nutrition consultation data semantic encoding feature vector;

[0011] A differential feature calculation module, used for calculating a differential feature vector between the first nutrition consultation data semantic coding feature vector and the second nutrition consultation data semantic coding feature vector as a nutrition consultation change semantic coding feature vector;

[0012] A feature dynamic interaction integration module is used to input the semantic coding feature vector of the nutrition plan to be evaluated and the semantic coding feature vector of the nutrition consultation change into a vector dynamic interaction combination module based on a switch mechanism to obtain a nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector as a nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation feature;

[0013] The classification result generation module is used to obtain a classification result based on the semantic interaction fusion representation feature of the nutrition plan to be evaluated-nutrition consultation change, and the classification result is used to indicate whether the nutrition plan to be evaluated is effective.

[0014] In the above-mentioned clinical nutrition virtual ward management system, the semantic encoding module of the nutrition plan to be evaluated is used to: perform word segmentation processing on the nutrition plan to be evaluated and input the result into the converter-based semantic encoder of the nutrition plan to be evaluated to obtain the semantic encoding feature vector of the nutrition plan to be evaluated.

[0015] In the above-mentioned clinical nutrition virtual ward management system, the feature dynamic interactive integration module includes: a feature joint processing unit, which is used to input the semantic coding feature vector of the nutrition plan to be evaluated and the semantic coding feature vector of the nutrition consultation change into the feature joint module for cascade processing to obtain the nutrition plan to be evaluated-nutrition consultation change semantic coding joint feature vector; a gated response calculation unit, which is used to input the semantic coding joint feature vector of the nutrition plan to be evaluated-nutrition consultation change into the gated response function to obtain the nutrition plan to be evaluated-nutrition consultation change semantic coding information fusion response gated value; a multi-scale semantic fusion unit, which is used to calculate the position-weighted sum of the semantic coding feature vector of the nutrition plan to be evaluated and the semantic coding feature vector of the nutrition consultation change using the semantic coding information fusion response gated value of the nutrition plan to be evaluated-nutrition consultation change as a weight to obtain the semantic interactive fusion representation vector of the nutrition plan to be evaluated-nutrition consultation change.

[0016] In the above-mentioned clinical nutrition virtual ward management system, the gated response calculation unit includes: a biased feature vector calculation subunit, which is used to calculate the matrix product of the joint feature vector of the semantic coding of the nutrition plan to be evaluated and the nutrition consultation change and the parameter matrix, and then add the obtained feature vector and the bias vector by position to obtain the joint biased feature vector of the semantic coding of the nutrition plan to be evaluated and the nutrition consultation change; a subunit for calculating the gating value response of the nutrition plan to be evaluated, which is used to input the joint biased feature vector of the semantic coding of the nutrition plan to be evaluated and the nutrition consultation change into a sigmoid activation function to obtain the information fusion response gating value of the semantic coding of the nutrition plan to be evaluated and the nutrition consultation change.

[0017] In the above-mentioned clinical nutrition virtual ward management system, the multi-scale semantic fusion unit includes: a semantic coding gating adjustment subunit of the nutrition scheme to be evaluated, which is used to calculate the positional product between the semantic coding feature vector of the nutrition scheme to be evaluated and the gating value of the semantic coding information fusion response of the nutrition scheme to be evaluated-nutrition consultation changes to obtain the semantic coding gating adjustment vector of the nutrition scheme to be evaluated; a nutrition consultation change semantic coding gating adjustment subunit, which is used to calculate the positional product between the semantic coding feature vector of the nutrition consultation changes and one minus the gating value of the semantic coding information fusion response of the nutrition scheme to be evaluated-nutrition consultation changes to obtain the semantic coding gating adjustment vector of the nutrition consultation changes; a semantic interaction fusion subunit, which is used to add the semantic coding gating adjustment vector of the nutrition scheme to be evaluated and the semantic coding gating adjustment vector of the nutrition consultation changes by position to obtain the semantic interaction fusion representation vector of the nutrition scheme to be evaluated-nutrition consultation changes.

[0018] In the above-mentioned clinical nutrition virtual ward management system, the classification result generation module is used to: input the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation changes into the classifier to obtain a classification result, and the classification result is used to indicate whether the nutrition plan to be evaluated is effective.

[0019] The above-mentioned clinical nutrition virtual ward management system also includes a training module for training the converter-based semantic encoder of the nutrition plan to be evaluated, the switch mechanism-based vector dynamic interaction joint module and the classifier.

[0020] In the above-mentioned clinical nutrition virtual ward management system, the training module includes: a training data acquisition unit, used to acquire training data, the training data including a training nutrition scheme to be evaluated, a first training nutrition consultation data before the training nutrition scheme to be evaluated is used, a second training nutrition consultation data after the training nutrition scheme to be evaluated is used, and a true and effective result, the true and effective result is a true value of whether the training nutrition scheme to be evaluated is effective; a training nutrition scheme to be evaluated semantic encoding unit, used to input the training nutrition scheme to be evaluated semantic encoder based on the converter after word segmentation processing to obtain a semantic encoding feature vector of the training nutrition scheme to be evaluated; a training nutrition consultation data semantic encoding unit, used to semantically encode the first training nutrition consultation data and the second training nutrition consultation data respectively to obtain a first training nutrition consultation data semantic encoding feature vector and a second training nutrition consultation data semantic encoding feature vector; a training camp A nutrition consultation change semantic encoding unit, used to calculate the difference feature vector between the first training nutrition consultation data semantic encoding feature vector and the second training nutrition consultation data semantic encoding feature vector as a training nutrition consultation change semantic encoding feature vector; a training data feature dynamic interaction integration unit, used to input the training nutrition plan to be evaluated semantic encoding feature vector and the training nutrition consultation change semantic encoding feature vector into the vector dynamic interaction joint module based on the switch mechanism to obtain a training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector; a classification loss unit, used to input the training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector into the classifier to obtain a classification loss function value; a model training unit, used to train the converter-based nutrition plan to be evaluated semantic encoder, the vector dynamic interaction joint module based on the switch mechanism and the classifier based on the classification loss function value and through back propagation of gradient descent.

[0021] This application has significant technical effects due to the adoption of the above technical solutions:

[0022] The clinical nutrition virtual ward management system provided by the present application obtains the first nutrition consultation data before the nutrition plan to be evaluated is used and the second nutrition consultation data after the nutrition plan to be evaluated is used, and uses AI-based natural language processing and analysis technology to semantically encode the nutrition plan to be evaluated, the first nutrition consultation data and the second nutrition consultation data, so as to intelligently judge whether the nutrition plan to be evaluated is effective based on the differential change between the first nutrition consultation data and the second nutrition consultation data and the semantic interaction fusion characteristics between the nutrition plan to be evaluated. In this way, the nutrition consultation data of the patient can be automatically processed and analyzed, reducing manual intervention, thereby improving the intelligent level of clinical nutrition management and helping to provide personalized nutrition advice to patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0024] Figure 1 This is a system block diagram of a clinical nutrition virtual ward management system according to an embodiment of the present application.

[0025] Figure 2 Schematic diagram of the architecture of a clinical nutrition virtual ward management system according to an embodiment of the present application.

[0026] Figure 3 It is a block diagram of the feature dynamic interaction integration module in the clinical nutrition virtual ward management system according to an embodiment of the present application.

[0027] Figure 4 It is a block diagram of a gating response calculation unit in a clinical nutrition virtual ward management system according to an embodiment of the present application.

[0028] Figure 5 This is a block diagram of a multi-scale semantic fusion unit in a clinical nutrition virtual ward management system according to an embodiment of the present application.

[0029] Figure 6 This is a block diagram of a training module in a clinical nutrition virtual ward management system according to an embodiment of the present application. DETAILED DESCRIPTION

[0030] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0031] Clinical nutrition is an integral part of the medical field and is increasingly important in preventing, treating and recovering from disease. Clinical nutrition management involves detailed nutritional assessment, diagnosis and treatment of patients in a healthcare setting to ensure that patients receive the appropriate nutritional supply, which helps improve health, treat disease and prevent malnutrition.

[0032] In traditional clinical nutrition practice, doctors and nutritionists need to evaluate whether patients have improved through a specific nutrition plan. However, this evaluation usually relies on manual analysis of patients' medical files and nutrition consultation records. This process is not only time-consuming and laborious, but also may be affected by personal subjective judgment, and there is a risk of omission or wrong evaluation.

[0033] Therefore, in response to the above technical problems, the technical concept of the present application is to obtain the first nutrition consultation data before the nutrition plan to be evaluated is used and the second nutrition consultation data after the nutrition plan to be evaluated is used, and use AI-based natural language processing and analysis technology to semantically encode the nutrition plan to be evaluated, the first nutrition consultation data and the second nutrition consultation data, so as to intelligently judge whether the nutrition plan to be evaluated is effective based on the differential changes between the first nutrition consultation data and the second nutrition consultation data and the semantic interaction fusion characteristics between the nutrition plan to be evaluated. In this way, the patient's nutrition consultation data can be automatically processed and analyzed, reducing manual intervention, thereby improving the intelligence level of clinical nutrition management and helping to provide patients with personalized nutrition advice.

[0034] Figure 1 This is a system block diagram of a clinical nutrition virtual ward management system according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the architecture of a clinical nutrition virtual ward management system according to an embodiment of the present application. Figure 1 and Figure 2As shown, in the clinical nutrition virtual ward management system 100, it includes: a first nutrition consultation data acquisition module 110, which is used to obtain the first nutrition consultation data before the nutrition plan to be evaluated is used; a second nutrition consultation data acquisition module 120, which is used to obtain the second nutrition consultation data after the nutrition plan to be evaluated is used; a nutrition plan to be evaluated semantic coding module 130, which is used to semantically encode the nutrition plan to be evaluated to obtain a semantic coding feature vector of the nutrition plan to be evaluated; a first and second nutrition consultation data semantic coding module 140, which is used to semantically encode the first nutrition consultation data and the second nutrition consultation data respectively to obtain a first nutrition consultation data semantic coding feature vector and a second nutrition consultation data semantic coding feature vector; a differential feature calculation module 150 , used to calculate the difference feature vector between the semantic coding feature vector of the first nutrition consultation data and the semantic coding feature vector of the second nutrition consultation data as the semantic coding feature vector of nutrition consultation changes; a feature dynamic interaction integration module 160, used to input the semantic coding feature vector of the nutrition plan to be evaluated and the semantic coding feature vector of the nutrition consultation changes into a vector dynamic interaction union module based on a switch mechanism to obtain a nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector as a nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation feature; a classification result generation module 170, used to obtain a classification result based on the nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation feature, and the classification result is used to indicate whether the nutrition plan to be evaluated is effective.

[0035] In an embodiment of the present application, the first nutrition consultation data acquisition module 110 and the second nutrition consultation data acquisition module 120 are used to obtain the first nutrition consultation data before the nutrition scheme to be evaluated is used and the second nutrition consultation data after the nutrition scheme to be evaluated is used. It should be understood that the nutrition consultation data generally includes the patient's physical examination results, nutrition assessment results, etc. The first nutrition consultation data provides the patient's nutritional status and health status before starting the nutrition scheme as baseline data; the second nutrition consultation data provides the patient's nutritional status after using the nutrition scheme to be evaluated. By comparing the second nutrition consultation data (after using the nutrition scheme) with the first nutrition consultation data, the impact of the nutrition scheme on the patient's nutritional status and health can be understood, thereby evaluating whether the nutrition scheme is effective.

[0036] In an embodiment of the present application, the semantic encoding module 130 of the nutritional scheme to be evaluated is used to semantically encode the nutritional scheme to be evaluated to obtain a semantically encoded feature vector of the nutritional scheme to be evaluated. Accordingly, it is considered that the nutritional scheme to be evaluated contains important semantic information and conceptual content about the nutritional scheme to be evaluated. Therefore, in the technical solution of the present application, the nutritional scheme to be evaluated is semantically encoded to obtain a semantically encoded feature vector of the nutritional scheme to be evaluated. In particular, in a specific embodiment of the present application, the nutritional scheme to be evaluated is word segmented and then input into a converter-based semantic encoder of the nutritional scheme to be evaluated to obtain a semantically encoded feature vector of the nutritional scheme to be evaluated, so that an entire complete scheme can be divided into multiple words, each word contains the semantic information of the nutritional scheme, so as to better understand the semantic information between each word in the entire scheme.

[0037] In an embodiment of the present application, the first and second nutrition consultation data semantic encoding module 140 is used to semantically encode the first nutrition consultation data and the second nutrition consultation data, respectively, to obtain a semantically encoded feature vector of the first nutrition consultation data and a semantically encoded feature vector of the second nutrition consultation data. It should be understood that the first nutrition consultation data and the second nutrition consultation data also contain important semantic associations and interdependencies between the contexts of the nutrition consultation. Based on this, in order to more clearly understand and capture the semantic information of the nutrition consultation data, in the technical solution of the present application, the first nutrition consultation data and the second nutrition consultation data are semantically encoded, respectively, to capture and mine the key semantic information in the first and second nutrition data, respectively, to obtain a semantically encoded feature vector of the first nutrition consultation data and a semantically encoded feature vector of the second nutrition consultation data.

[0038] In an embodiment of the present application, the differential feature calculation module 150 is used to calculate the differential feature vector between the semantic coding feature vector of the first nutrition consultation data and the semantic coding feature vector of the second nutrition consultation data as the semantic coding feature vector of the nutrition consultation change. Accordingly, considering that the semantic coding feature vector of the first nutrition consultation data and the semantic coding feature vector of the second nutrition consultation data respectively represent the semantic information before and after the scheme to be evaluated is not used. Therefore, in order to clearly understand the difference in changes before and after the scheme to be evaluated is used, and to facilitate the subsequent better understanding of the effect and impact of the nutrition scheme to be evaluated, in the technical solution of the present application, the differential feature vector between the semantic coding feature vector of the first nutrition consultation data and the semantic coding feature vector of the second nutrition consultation data is calculated as the semantic coding feature vector of the nutrition consultation change. That is to say, by calculating the differential feature vector of the first and second nutrition consultation data, the changes before and after the implementation of the nutrition scheme to be evaluated can be effectively captured, so as to provide a more intuitive and objective indicator to evaluate the effect of the nutrition scheme to be evaluated, and comprehensively reflect the changes brought about by the nutrition scheme, thereby providing better support and basis for the subsequent evaluation of the effect of the nutrition scheme and the impact on individuals.

[0039] In an embodiment of the present application, the feature dynamic interaction integration module 160 is used to input the semantic coding feature vector of the nutrition scheme to be evaluated and the semantic coding feature vector of the nutrition consultation change into a vector dynamic interaction joint module based on a switch mechanism to obtain a nutrition scheme to be evaluated-nutrition consultation change semantic interaction fusion representation vector as a nutrition scheme to be evaluated-nutrition consultation change semantic interaction fusion representation feature. It should be understood that the semantic coding feature vector of the nutrition scheme to be evaluated and the semantic coding feature vector of the nutrition consultation change respectively express the semantic feature information of the nutrition scheme to be evaluated and the change information of the semantic difference between the two before and after the nutrition scheme to be evaluated is not used, and there is key semantic feature information of different degrees between the two about whether the scheme to be evaluated is effective. Therefore, in order to dynamically adjust and integrate the semantic coding feature vector of the nutrition scheme to be evaluated and the semantic coding feature vector of the nutrition consultation change, so as to achieve effective interaction and fusion between the two, in the technical solution of the present application, the semantic coding feature vector of the nutrition scheme to be evaluated and the semantic coding feature vector of the nutrition consultation change are input into a vector dynamic interaction joint module based on a switch mechanism to obtain a nutrition scheme to be evaluated-nutrition consultation change semantic interaction fusion representation vector.

[0040] In detail, the vector dynamic interaction joint module based on the switch mechanism first performs feature fusion on the semantic coding feature vector of the nutrition plan to be evaluated and the semantic coding feature vector of the nutrition consultation changes, and calculates a gating value through a gated response function to dynamically adjust the importance of the fusion weight between the characteristics of the nutrition plan to be evaluated and the characteristics of the nutrition consultation changes, and then uses the gated value as a weight to perform weighted fusion on the semantic coding feature vector of the nutrition plan to be evaluated and the semantic coding feature vector of the nutrition consultation changes, thereby realizing dynamic semantic interaction between the semantic coding characteristics of the nutrition plan to be evaluated and the semantic coding characteristics of the nutrition consultation changes, so as to enable the model to better capture the semantic association and mutual influence between the two, thereby more accurately evaluating the effectiveness of the nutrition plan to be evaluated.

[0041] Specifically, Figure 3 FIG. 1 is a block diagram of a feature dynamic interaction integration module in a clinical nutrition virtual ward management system according to an embodiment of the present application. Figure 3 As shown, the feature dynamic interactive integration module 160 includes: a feature joint processing unit 161, which is used to input the semantic coding feature vector of the nutrition plan to be evaluated and the semantic coding feature vector of the nutrition consultation change into a feature joint module for cascade processing to obtain a joint feature vector of the nutrition plan to be evaluated and the semantic coding of the nutrition consultation change; a gated response calculation unit 162, which is used to input the joint feature vector of the semantic coding of the nutrition plan to be evaluated and the nutrition consultation change into a gated response function to obtain a gated value of a semantic coding information fusion response of the nutrition plan to be evaluated and the nutrition consultation change; and a multi-scale semantic fusion unit 163, which is used to calculate the position-weighted sum of the semantic coding feature vector of the nutrition plan to be evaluated and the semantic coding feature vector of the nutrition consultation change using the gated value of the semantic coding information fusion response of the nutrition plan to be evaluated and the nutrition consultation change as a weight to obtain a semantic interactive fusion representation vector of the nutrition plan to be evaluated and the nutrition consultation change.

[0042] More specifically, Figure 4 FIG. 1 is a block diagram of a gated response calculation unit in a clinical nutrition virtual ward management system according to an embodiment of the present application. Figure 4 As shown, the gated response calculation unit 162 includes: a biased feature vector calculation subunit 1621, which is used to calculate the matrix product of the joint feature vector of the semantic coding of the nutrition plan to be evaluated-nutritional consultation changes and the parameter matrix, and then add the obtained feature vector and the bias vector by position to obtain the joint biased feature vector of the semantic coding of the nutrition plan to be evaluated-nutritional consultation changes; a nutrition plan to be evaluated response gating value calculation subunit 1622, which is used to input the joint biased feature vector of the semantic coding of the nutrition plan to be evaluated-nutritional consultation changes into a sigmoid activation function to obtain the information fusion response gating value of the semantic coding of the nutrition plan to be evaluated-nutritional consultation changes.

[0043] More specifically, Figure 5 FIG. 4 is a block diagram of a multi-scale semantic fusion unit in a clinical nutrition virtual ward management system according to an embodiment of the present application. Figure 5 As shown, the multi-scale semantic fusion unit 163 includes: a semantic coding gating adjustment subunit 1631 of the nutrition scheme to be evaluated, which is used to calculate the positional product between the semantic coding feature vector of the nutrition scheme to be evaluated and the gating value of the semantic coding information fusion response of the nutrition scheme to be evaluated-nutrition consultation changes to obtain the semantic coding gating adjustment vector of the nutrition scheme to be evaluated; a nutrition consultation change semantic coding gating adjustment subunit 1632, which is used to calculate the positional product between the semantic coding feature vector of the nutrition consultation changes and one minus the gating value of the semantic coding information fusion response of the nutrition scheme to be evaluated-nutrition consultation changes to obtain the semantic coding gating adjustment vector of the nutrition consultation changes; a semantic interaction fusion subunit 1633, which is used to add the semantic coding gating adjustment vector of the nutrition scheme to be evaluated and the semantic coding gating adjustment vector of the nutrition consultation changes by position to obtain the semantic interaction fusion representation vector of the nutrition scheme to be evaluated-nutrition consultation changes.

[0044] In the embodiment of the present application, specifically, the feature dynamic interaction integration module is used to: input the semantic coding feature vector of the nutrition plan to be evaluated and the semantic coding feature vector of the nutrition consultation change into the vector dynamic interaction joint module based on the switch mechanism, and process them with the following dynamic fusion formula to obtain the nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector; wherein, the dynamic fusion formula is:

[0045]

[0046] Wherein, v1 and v2 are the semantic encoding feature vector of the nutrition plan to be evaluated and the semantic encoding feature vector of the nutrition consultation change, respectively, [·,·] is the cascade processing, W is the parameter matrix, b is the bias vector, σ(·) is the sigmoid function, is the semantic encoding information fusion response gating value of the nutrition plan to be evaluated-nutrition consultation change, and vc is the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation change.

[0047] In an embodiment of the present application, the classification result generation module 170 is used to obtain a classification result based on the semantic interaction fusion representation feature of the nutrition plan to be evaluated-nutrition consultation changes, and the classification result is used to indicate whether the nutrition plan to be evaluated is effective. Specifically, in an embodiment of the present application, the classification result generation module is used to: input the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation changes into a classifier to obtain a classification result, and the classification result is used to indicate whether the nutrition plan to be evaluated is effective. That is, the semantic interaction fusion representation feature of the nutrition plan to be evaluated-nutrition consultation changes obtained by dynamic interaction fusion of the semantic coding feature vector of the nutrition plan to be evaluated and the semantic coding feature vector of the nutrition consultation changes is used for classification processing, so as to intelligently judge whether the nutrition plan to be evaluated is effective. In this way, the nutrition consultation data of the patient can be automatically processed and analyzed, reducing manual intervention, thereby improving the intelligent level of clinical nutrition management, and helping to provide personalized nutrition advice to patients.

[0048] In the technical solution of the present application, the semantic coding feature vector of the nutrition scheme to be evaluated and the semantic coding feature vector of the nutrition consultation change respectively represent the text semantic features of the nutrition scheme to be evaluated and the text semantic differential features between the first nutrition consultation data before the nutrition scheme to be evaluated was used and the second nutrition consultation data after the nutrition scheme to be evaluated was used. Considering that the second nutrition consultation data after the use of the nutrition scheme to be evaluated and the first nutrition consultation data before the use of the nutrition scheme to be evaluated are different from the nutrition scheme to be evaluated at the text expression level, the closeness between the semantic coding feature vector of the nutrition scheme to be evaluated and the semantic coding feature vector of the nutrition consultation change in the text semantic space may be insufficient, which will cause the semantic interaction fusion representation vector of the nutrition scheme to be evaluated-nutrition consultation change obtained by the vector dynamic interaction joint module based on the switch mechanism to have a more significant local distribution difference, thereby causing its harmony deviation relative to the predetermined category when performing classification regression through the classifier, affecting the classification iteration effect.

[0049] Based on this, in a preferred embodiment, the semantic interaction fusion representation vector of the nutrition plan to be evaluated and the nutrition consultation change is input into a classifier to obtain a classification result, including:

[0050] Performing L random sampling on the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation change to obtain L random perturbation eigenvalues ​​of the nutrition plan to be evaluated-nutrition consultation change;

[0051] Arrange the L random disturbance eigenvalues ​​of the nutrition plan to be evaluated-nutrition consultation changes in order from large to small to obtain a random disturbance ordered vector of the nutrition plan to be evaluated-nutrition consultation changes;

[0052] Calculating the inner product of the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation change to obtain the energy modulation characteristic value of the nutrition plan to be evaluated-nutrition consultation change;

[0053] Based on the energy modulation characteristic value of the nutrition plan to be evaluated-nutrition consultation change, the random disturbance ordered vector of the nutrition plan to be evaluated-nutrition consultation change is semantically scaled to obtain a semantically enhanced random disturbance ordered vector of the nutrition plan to be evaluated-nutrition consultation change;

[0054] Calculating the product of the random disturbance ordered vector of the semantically reinforced nutrition plan to be evaluated-nutrition consultation change and the transposed vector of the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation change to obtain the orthogonal modulation matrix of the best sample prompt of the nutrition plan to be evaluated-nutrition consultation change;

[0055] Calculating the matrix product between the best sample prompt orthogonal modulation matrix of the nutrition plan to be evaluated-nutrition consultation change and the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation change to obtain an optimized semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation change;

[0056] The optimized nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector is input into the classifier to obtain the classification result.

[0057] The above optimization process is expressed as follows:

[0058] S i =∈ i ,i=1,2,...,L

[0059]

[0060] u=average(V)

[0061]

[0062] V order =Order{S i}

[0063]

[0064] in, represents a normal distribution, represents the Gaussian density probability function, u and σ 2 Respectively represent the mean and variance of the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation change, ∈ irepresents a random variable that follows a normal distribution, L represents the number of random sampling, S i represents the random disturbance eigenvalue of the i-th nutrition plan to be evaluated-nutrition consultation change, V represents the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation change, v i represents the i-th eigenvalue of the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation change, average(·) represents the average value of the calculated vector, L1 represents the length of the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation change, Order{S i} indicates that the random disturbance eigenvalues ​​of the nutrition plan-nutrition consultation change to be evaluated are arranged in order from large to small, V order represents the ordered vector of random disturbance of the nutrition plan to be evaluated-nutrition consultation change, e represents the energy modulation eigenvalue of the nutrition plan to be evaluated-nutrition consultation change, represents the square of the vector's two-norm, ⊙ represents the point multiplication by position, represents matrix multiplication, (·) T Represents the transpose of a vector, M t represents the orthogonal modulation matrix of the best example of nutrition plan to be evaluated-nutrition consultation change prompt, V ′ Represents the optimized nutrition plan to be evaluated - the semantic interaction fusion representation vector of nutrition consultation changes.

[0065] Accordingly, by randomly sampling the semantic interaction fusion representation vector of the nutrition plan-nutrition consultation change to be evaluated L times to obtain a set of new feature values ​​based on the original features but containing slight changes, randomness is introduced to explore different aspects of the feature space, and by simulating the noise or variation in the actual data, it helps the model to better generalize to unseen data. Then, these randomly perturbed feature values ​​of the nutrition plan-nutrition consultation change to be evaluated are arranged in order from large to small to form an ordered vector of randomly perturbed changes in the nutrition plan-nutrition consultation to be evaluated, so as to identify which perturbation directions may be the most important for the current problem. Such an ordering not only helps the algorithm focus on the feature differences that can best distinguish different categories, but also helps to reveal hidden data structures or patterns. Subsequently, the self-inner product of the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation changes is calculated to obtain the energy modulation eigenvalue of the nutrition plan to be evaluated-nutrition consultation changes. The energy modulation eigenvalue of the nutrition plan to be evaluated-nutrition consultation changes reflects the overall strength or "energy" of the original feature vector, and provides a benchmark for measuring the importance of the feature. Using this energy modulation eigenvalue of the nutrition plan to be evaluated-nutrition consultation changes as a reference point, the previously formed random perturbation ordered vector of the nutrition plan to be evaluated-nutrition consultation changes is semantically scaled to construct a semantically enhanced random perturbation ordered vector of the nutrition plan to be evaluated-nutrition consultation changes. In this way, those disturbances that appear to be more important in the original feature space are amplified, thereby enhancing the model's sensitivity to changes in key features, while also retaining relatively small but potentially useful disturbance information.

[0066] Then, the product of the random perturbation ordered vector of the semantically enhanced nutrition plan to be evaluated-nutrition consultation change and the transpose of the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation change is calculated to obtain the orthogonal modulation matrix of the best example prompt of the nutrition plan to be evaluated-nutrition consultation change. The purpose is to find the interaction mode between the features under two different perspectives. The formed matrix can be regarded as a conversion tool for capturing and representing the complex linear relationship between features. Finally, the final feature structure modulation is completed by applying the above-constructed orthogonal modulation matrix of the best example prompt of the nutrition plan to be evaluated-nutrition consultation change to the original semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation change, so as to make the features more in line with the requirements of the objective function, reduce redundant information, and maximize the retention of useful information, so as to improve learning efficiency and prediction performance.

[0067] In summary, the feature manifold optimization process of the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation changes is clarified, which extracts the best prompt examples based on the fine-grained target retrieval of random perturbations, and then guides adversarial robust learning based on the best prompt examples to complete the structure of the high-dimensional manifold of the feature vector in the feature space, so as to improve the detail and structure of the feature expression. In this way, the harmony of the feature set of the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation changes in the predetermined category within the associated target classification domain is improved, thereby improving the classification training speed and the accuracy of the classification results when the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation changes is classified by the classifier.

[0068] It is worth mentioning that those skilled in the art should be aware that before applying a deep neural network model for inference, the deep neural network model must be trained so that the deep neural network can implement specific functional capabilities.

[0069] Specifically, in the technical solution of the present application, the clinical nutrition virtual ward management system also includes a training module for training the converter-based semantic encoder of the nutrition plan to be evaluated, the switch mechanism-based vector dynamic interaction joint module and the classifier.

[0070] Figure 6 FIG. 1 is a block diagram of a training module in a clinical nutrition virtual ward management system according to an embodiment of the present application. Figure 6As shown, the training module 200 includes: a training data acquisition unit 210, which is used to acquire training data, wherein the training data includes a training nutrition scheme to be evaluated, a first training nutrition consultation data before the training nutrition scheme to be evaluated is used, a second training nutrition consultation data after the training nutrition scheme to be evaluated is used, and a true and effective result, wherein the true and effective result is a true value of whether the training nutrition scheme to be evaluated is effective; a training nutrition scheme to be evaluated semantic encoding unit 220, which is used to input the training nutrition scheme to be evaluated semantic encoder based on the converter after word segmentation processing to obtain a semantic encoding feature vector of the training nutrition scheme to be evaluated; a training nutrition consultation data semantic encoding unit 230, which is used to semantically encode the first training nutrition consultation data and the second training nutrition consultation data respectively to obtain a first training nutrition consultation data semantic encoding feature vector and a second training nutrition consultation data semantic encoding feature vector; a training nutrition consultation change semantic encoding unit 230; The encoding unit 240 is used to calculate the difference feature vector between the first training nutrition consultation data semantic encoding feature vector and the second training nutrition consultation data semantic encoding feature vector as the training nutrition consultation change semantic encoding feature vector; the training data feature dynamic interaction integration unit 250 is used to input the training nutrition plan to be evaluated semantic encoding feature vector and the training nutrition consultation change semantic encoding feature vector into the vector dynamic interaction joint module based on the switch mechanism to obtain the training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector; the classification loss unit 260 is used to input the training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector into the classifier to obtain the classification loss function value; the model training unit 270 is used to train the converter-based nutrition plan to be evaluated semantic encoder, the vector dynamic interaction joint module based on the switch mechanism and the classifier based on the classification loss function value and through gradient descent back propagation.

[0071] It should be understood that in the technical solution of the present application, the semantic coding feature vector of the training nutrition scheme to be evaluated and the semantic coding feature vector of the training nutrition consultation change respectively represent the text semantic features of the training nutrition scheme to be evaluated and the text semantic differential features between the first training nutrition consultation data before the nutrition scheme to be evaluated is not used and the second training nutrition consultation data after the nutrition scheme to be evaluated is used. Considering that the second training nutrition consultation data after the use of the training nutrition scheme to be evaluated and the first training nutrition consultation data before the training nutrition scheme to be evaluated are different from the training nutrition scheme to be evaluated at the text expression level, the closeness between the semantic coding feature vector of the training nutrition scheme to be evaluated and the semantic coding feature vector of the training nutrition consultation change in the text semantic space may not be enough, which will cause the training nutrition scheme to be evaluated-nutrition consultation change semantic interaction fusion representation vector obtained by the vector dynamic interaction joint module based on the switch mechanism to have a more significant local distribution difference, thereby causing its harmony deviation relative to the predetermined category when performing classification regression through the classifier, affecting the classification iteration effect.

[0072] Based on this, in the preferred embodiment, when the converter-based nutritional scheme to be evaluated semantic encoder, the switch mechanism-based vector dynamic interaction joint module and the classifier are trained based on the classification loss function value, wherein, in each iteration of the training, the training nutritional scheme to be evaluated-nutritional consultation change semantic interaction fusion representation vector is iteratively optimized,

[0073] Among them, the iterative optimization process includes: first calculating the autocorrelation matrix of the transposed representation vector of the training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion and itself, and calculating the inner product of the i-th and j-th row vectors of the autocorrelation matrix as the matrix value of the (i, j)-th position of the weight matrix to obtain the weight matrix, then, after matrix multiplication of the weight matrix and the autocorrelation matrix, further matrix-vector multiplication with the training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector to obtain a correction vector, and finally, dot multiplication of the correction vector with the training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector to obtain the optimized training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector.

[0074] Specifically, in this preferred embodiment, the training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector is optimized to obtain an optimized training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector, and the process formula is expressed as follows:

[0075]

[0076] Wherein, V is the semantic interaction fusion representation vector of the nutrition plan to be evaluated and the nutrition consultation change, V T is the transposed vector of the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation change in the training, is the matrix multiplication, M is the autocorrelation matrix, V i and V j are the i-th and j-th row vectors of the autocorrelation matrix, V j T is the transposed vector of the row vector of the jth row in the autocorrelation matrix, L is the number of row vectors in the autocorrelation matrix, m i,j is the matrix value of the (i, j)th position of the weight matrix, M ′ is the weight matrix, R is a real number set, ⊙ is the position point multiplication, V ′ It is the optimized training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector.

[0077] Accordingly, in a preferred embodiment, based on the self-correlation dimension of the training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector as the object to be modulated, the spatial sub-dimensional complexity of the high-dimensional feature space of the feature distribution is used to perform correlation expansion, thereby introducing the decomposable correlation dimension offset of the training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector into the heterogeneous correlation embedding space, thereby enhancing its correlation self-consistency through the joint fine-tuning of the decomposable dimension set represented by the heterogeneous correlation embedding space, so as to improve the predetermined category harmony of the feature set of the training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector in the associated target classification domain, thereby improving the classification training speed and classification result accuracy of the training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector when it is classified by the classifier. In this way, the patient's nutrition consultation data can be automatically processed and analyzed, reducing manual intervention, thereby improving the intelligent level of clinical nutrition management, and helping to provide personalized nutrition advice to patients.

[0078] In summary, the clinical nutrition virtual ward management system 100 based on the embodiment of the present application is explained, which obtains the first nutrition consultation data before the nutrition plan to be evaluated is used and the second nutrition consultation data after the nutrition plan to be evaluated is used, and uses AI-based natural language processing and analysis technology to semantically encode the nutrition plan to be evaluated, the first nutrition consultation data and the second nutrition consultation data, so as to intelligently judge whether the nutrition plan to be evaluated is effective according to the differential changes between the first nutrition consultation data and the second nutrition consultation data and the semantic interaction fusion characteristics between the nutrition plan to be evaluated. In this way, the patient's nutrition consultation data can be automatically processed and analyzed, reducing manual intervention, thereby improving the intelligence level of clinical nutrition management and helping to provide patients with personalized nutrition advice.

[0079] As described above, the clinical nutrition virtual ward management system 100 according to the embodiment of the present application can be implemented in various terminal devices, such as a server for clinical nutrition virtual ward management. In one example, the clinical nutrition virtual ward management system 100 according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the clinical nutrition virtual ward management system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the clinical nutrition virtual ward management system 100 can also be one of the many hardware modules of the terminal device.

[0080] Alternatively, in another example, the clinical nutrition virtual ward management system 100 and the terminal device may also be separate devices, and the clinical nutrition virtual ward management system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0081] In other embodiments of the present application, the critical care medicine-clinical nutrition multidisciplinary comprehensive diagnosis and treatment team (multi-disciplinary team, MDT) working mode is also innovatively explored, the standardized nutrition treatment path for critically ill patients (nutrition risk screening-nutrition assessment-nutrition diagnosis-nutrition treatment-outpatient follow-up) is promoted, and the "clinical nutrition management virtual ward" working mode is established. The virtual ward has an independent nutrition diagnosis and treatment information system, which can complete the information docking of HIS and LIS systems. The nutrition screening, nutrition assessment, nutrition diagnosis, nutrition consultation and other data of the nutrition diagnosis and treatment information system can be read into the big medical record by the hospital electronic medical record system. Based on the multidisciplinary diagnosis and treatment model, relying on virtual ward management, and guided by malnutrition, nutrition treatment work is carried out, and a standardized clinical nutrition diagnosis and treatment path based on "nutrition screening-assessment diagnosis-nutrition intervention" is established, which can effectively improve the diagnosis rate of malnutrition and the accuracy of graded diagnosis, avoid the unreasonable treatment phenomenon of patients without nutrition risk, shorten the waiting time for patient treatment, and realize individualized and precise treatment from the formulation of the plan to the selection of the treatment path. In the era of medical big data, by establishing virtual management wards, we can achieve in-depth integration of nutrition diagnosis and treatment with clinical medicine while saving medical costs. At the same time, we can record, save, and compare the nutrition diagnosis and treatment process of the intervention cases in a timely and accurate manner and include it in medical record management. This provides strong data support for the clinical nutrition department to carry out clinical nutrition-related research, and is conducive to improving the clinical nutrition diagnosis and treatment and scientific research level of the nutrition department and enhancing its comprehensive competitiveness.

[0082] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details of the above application are only for the purpose of illustration and ease of understanding, rather than limitation, and the above details do not limit the present application to the use of the above specific details to be implemented.

Claims

1. Clinical nutrition virtual ward management system, characterized by: include: A first nutrition consultation data acquisition module, used to acquire first nutrition consultation data before the nutrition plan to be evaluated is used; A second nutrition consultation data acquisition module, used to acquire second nutrition consultation data after using the nutrition plan to be evaluated; A semantic coding module for the nutritional scheme to be evaluated, used for semantically coding the nutritional scheme to be evaluated to obtain a semantic coding feature vector of the nutritional scheme to be evaluated; A first and second nutrition consultation data semantic encoding module, for performing semantic encoding on the first nutrition consultation data and the second nutrition consultation data respectively to obtain a first nutrition consultation data semantic encoding feature vector and a second nutrition consultation data semantic encoding feature vector; A differential feature calculation module, used for calculating a differential feature vector between the first nutrition consultation data semantic coding feature vector and the second nutrition consultation data semantic coding feature vector as a nutrition consultation change semantic coding feature vector; A feature dynamic interaction integration module is used to input the semantic coding feature vector of the nutrition plan to be evaluated and the semantic coding feature vector of the nutrition consultation change into a vector dynamic interaction combination module based on a switch mechanism to obtain a nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector as a nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation feature; The classification result generation module is used to obtain a classification result based on the semantic interaction fusion representation feature of the nutrition plan to be evaluated-nutrition consultation change, and the classification result is used to indicate whether the nutrition plan to be evaluated is effective.

2. The clinical nutrition virtual ward management system according to claim 1, characterized in that: The semantic coding module of the nutritional scheme to be evaluated is used to: perform word segmentation processing on the nutritional scheme to be evaluated and then input the result into a converter-based semantic encoder of the nutritional scheme to be evaluated to obtain a semantic coding feature vector of the nutritional scheme to be evaluated.

3. The clinical nutrition virtual ward management system according to claim 2, characterized in that: The feature dynamic interaction integration module includes: A feature joint processing unit, used for inputting the semantic coding feature vector of the nutrition scheme to be evaluated and the semantic coding feature vector of the nutrition consultation change into a feature joint module for cascade processing to obtain a semantic coding joint feature vector of the nutrition scheme to be evaluated and the nutrition consultation change; A gated response calculation unit, used for inputting the nutrition plan to be evaluated-nutrition consultation change semantic coding joint feature vector into a gated response function to obtain a nutrition plan to be evaluated-nutrition consultation change semantic coding information fusion response gated value; A multi-scale semantic fusion unit is used to calculate the position-weighted sum of the semantic coding feature vector of the nutrition scheme to be evaluated and the semantic coding feature vector of the nutrition consultation change by using the semantic coding information fusion response gate value of the nutrition scheme to be evaluated-nutrition consultation change as a weight to obtain the semantic interaction fusion representation vector of the nutrition scheme to be evaluated-nutrition consultation change.

4. The clinical nutrition virtual ward management system according to claim 3, characterized in that: The gate response calculation unit comprises: A biased feature vector calculation subunit is used to calculate the matrix product of the nutrition plan to be evaluated-nutrition consultation change semantic coding joint feature vector and the parameter matrix, and then add the obtained feature vector and the bias vector by position to obtain the nutrition plan to be evaluated-nutrition consultation change semantic coding joint biased feature vector; The subunit for calculating the response gating value of the nutrition scheme to be evaluated is used to input the semantic coding and biased feature vector of the nutrition scheme to be evaluated-nutritional consultation changes into a sigmoid activation function to obtain the fusion response gating value of the semantic coding information of the nutrition scheme to be evaluated-nutritional consultation changes.

5. The clinical nutrition virtual ward management system according to claim 4, characterized in that: The multi-scale semantic fusion unit comprises: The semantic coding gating adjustment subunit of the nutrition scheme to be evaluated is used to calculate the positional product between the semantic coding feature vector of the nutrition scheme to be evaluated and the gating value of the fusion response of the semantic coding information of the nutrition scheme to be evaluated-nutrition consultation change to obtain the semantic coding gating adjustment vector of the nutrition scheme to be evaluated; A nutrition consultation change semantic coding gating adjustment subunit, used for calculating the positional product between the nutrition consultation change semantic coding feature vector and one minus the nutrition plan to be evaluated-nutrition consultation change semantic coding information fusion response gating value to obtain a nutrition consultation change semantic coding gating adjustment vector; The semantic interaction fusion subunit is used to add the semantic coding gate adjustment vector of the nutrition plan to be evaluated and the semantic coding gate adjustment vector of the nutrition consultation change according to the position to obtain the nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector.

6. The clinical nutrition virtual ward management system according to claim 5, characterized in that: The classification result generating module is used to: input the semantic interaction fusion representation vector of the nutrition plan to be evaluated-nutrition consultation change into a classifier to obtain a classification result, and the classification result is used to indicate whether the nutrition plan to be evaluated is effective.

7. The clinical nutrition virtual ward management system according to claim 6, characterized in that: It also includes a training module for training the converter-based semantic encoder of the nutritional plan to be evaluated, the switch mechanism-based vector dynamic interaction joint module and the classifier.

8. The clinical nutrition virtual ward management system according to claim 7, characterized in that: The training module comprises: a training data acquisition unit, configured to acquire training data, wherein the training data includes a training nutrition program to be evaluated, first training nutrition consultation data before the training nutrition program to be evaluated is used, second training nutrition consultation data after the training nutrition program to be evaluated is used, and a true and effective result, wherein the true and effective result is a true value of whether the training nutrition program to be evaluated is effective; A semantic encoding unit for training the nutritional scheme to be evaluated is used to input the word segmentation process of the nutritional scheme to be evaluated into the converter-based semantic encoder of the nutritional scheme to be evaluated to obtain a semantic encoding feature vector of the nutritional scheme to be evaluated; A training nutrition consultation data semantic encoding unit, used for performing semantic encoding on the first training nutrition consultation data and the second training nutrition consultation data respectively to obtain a first training nutrition consultation data semantic encoding feature vector and a second training nutrition consultation data semantic encoding feature vector; A training nutrition consultation change semantic coding unit, used for calculating a difference feature vector between the first training nutrition consultation data semantic coding feature vector and the second training nutrition consultation data semantic coding feature vector as a training nutrition consultation change semantic coding feature vector; A training data feature dynamic interaction integration unit, used for inputting the semantic coding feature vector of the training nutrition scheme to be evaluated and the semantic coding feature vector of the training nutrition consultation change into the vector dynamic interaction combination module based on the switch mechanism to obtain a training nutrition scheme to be evaluated-nutrition consultation change semantic interaction fusion representation vector; A classification loss unit, used for inputting the training nutrition plan to be evaluated-nutrition consultation change semantic interaction fusion representation vector into the classifier to obtain a classification loss function value; A model training unit is used to train the converter-based nutritional scheme semantic encoder to be evaluated, the switch mechanism-based vector dynamic interaction joint module and the classifier based on the classification loss function value and through back propagation of gradient descent.

Images