Human body obesity prediction system and method based on QCT technology

Abstract

The invention relates to a QCT technology-based human body obesity prediction system and method, and relates to the field of health management, and the method comprises the steps: regularly collecting and obtaining a BMI parameter sequence, a user feature information sequence and a QCT image sequence, and analyzing and obtaining a fat distribution sequence; predicting according to the BMI parameter sequence and the user feature information sequence to obtain first obesity trend information, and predicting according to the fat distribution sequence to obtain second obesity trend information; weight is configured according to the representative information sequence of the user, weighting calculation is conducted on the first obesity trend information and the second obesity trend information, and obesity trend information is obtained and serves as an obesity prediction result. The technical problems that a traditional human body obesity prediction method often depends on static indexes such as BMI, personalized prediction cannot be carried out according to user characteristics, and obesity prediction accuracy is insufficient can be solved; more accurate and personalized obesity prediction can be realized, and the accuracy and reliability of human body obesity prediction are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of health management, and in particular to a human obesity prediction system and method based on QCT technology. Background Art

[0002] Obesity, as an increasingly serious public health problem worldwide, has been widely recognized as a risk factor for a variety of metabolic diseases and chronic diseases, such as cardiovascular disease, diabetes, non-alcoholic fatty liver disease, etc.

[0003] At present, traditional obesity diagnosis and prediction methods mostly rely on static indicators such as BMI (body mass index). Although these indicators are simple and easy to obtain, they have some significant limitations in practical applications. For example, some individuals with normal BMI may have higher visceral fat content (such as latent obesity), while individuals with higher BMI may not actually be obese due to their large muscle mass.

[0004] In summary, traditional human obesity prediction methods often rely on static indicators such as BMI, are unable to make personalized predictions based on user characteristics, and have the technical problem of insufficient accuracy in obesity prediction. Summary of the invention

[0005] The present invention aims to solve the technical problem that traditional human obesity prediction methods often rely on static indicators such as BMI, cannot perform personalized predictions based on user characteristics, and have insufficient accuracy in obesity prediction. A human obesity prediction system and method based on QCT technology is provided to solve the problem.

[0006] The technical solution of the present invention to solve the above technical problems is as follows:

[0007] In a first aspect, the present invention provides a human obesity prediction system based on QCT technology, including: an initial fat distribution recognition module, used to collect initial BMI parameters and initial user characteristic information of a target user, analyze user representative information, perform QCT detection on the target user to obtain an initial QCT image, configure QCT recognition accuracy according to the user representative information, and obtain initial fat distribution by recognizing the initial QCT image; a fat distribution sequence acquisition module, used to continue to regularly collect BMI parameters, user characteristic information and QCT images of the target user, obtain BMI parameter sequence, user characteristic information sequence and QCT image sequence, analyze user representative information sequence, configure QCT recognition accuracy sequence, and analyze to obtain fat distribution sequence; an obesity trend prediction module, used to predict and obtain first obesity trend information of the target user according to the BMI parameter sequence and user characteristic information sequence, and predict and obtain second obesity trend information of the target user according to the fat distribution sequence; an obesity prediction result acquisition module, used to configure weights according to the user representative information sequence, perform weighted calculation on the first obesity trend information and the second obesity trend information, and obtain obesity trend information as an obesity prediction result.

[0008] In a second aspect, the present invention provides a human obesity prediction method based on QCT technology, including: collecting initial BMI parameters and initial user characteristic information of a target user, analyzing user representative information, performing QCT detection on the target user to obtain an initial QCT image, configuring QCT recognition accuracy based on the user representative information, and identifying the initial QCT image to obtain an initial fat distribution; continuing to regularly collect the BMI parameters, user characteristic information and QCT images of the target user to obtain a BMI parameter sequence, a user characteristic information sequence and a QCT image sequence, analyzing the user representative information sequence, configuring a QCT recognition accuracy sequence, and analyzing to obtain a fat distribution sequence; predicting and obtaining a first obesity trend information of the target user based on the BMI parameter sequence and the user characteristic information sequence, and predicting and obtaining a second obesity trend information of the target user based on the fat distribution sequence; configuring weights based on the user representative information sequence, performing weighted calculation on the first obesity trend information and the second obesity trend information to obtain obesity trend information as an obesity prediction result.

[0009] The beneficial effects of the present invention are as follows: the present invention collects the initial BMI parameters and initial user characteristic information of the target user, analyzes the user representative information, performs QCT detection on the target user to obtain an initial QCT image, configures the QCT recognition accuracy according to the user representative information, and obtains the initial fat distribution by recognizing the initial QCT image.

[0010] Then, the target user's BMI parameters, user characteristic information and QCT images are collected regularly to obtain BMI parameter sequences, user characteristic information sequences and QCT image sequences, the user representative information sequences are analyzed, the QCT recognition accuracy sequences are configured, and the fat distribution sequences are analyzed to obtain the first obesity trend information of the target user according to the BMI parameter sequence and the user characteristic information sequence, and the second obesity trend information of the target user is predicted according to the fat distribution sequence. Finally, the weights are configured according to the user representative information sequence, and the first obesity trend information and the second obesity trend information are weightedly calculated to obtain the obesity trend information as the obesity prediction result.

[0011] This application introduces QCT technology and combines user characteristics and BMI parameters to predict human obesity, which can achieve more accurate and personalized obesity prediction, significantly improve the accuracy and reliability of human obesity prediction, and thus provide a scientific basis for human health management. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A schematic diagram of the structure of a human obesity prediction system based on QCT technology provided by the present invention;

[0013] Figure 2 This is a schematic diagram of the process of the human obesity prediction method based on QCT technology provided by the present invention.

[0014] In the accompanying drawings, the components represented by the reference numerals are described as follows:

[0015] Initial fat distribution identification module 01, fat distribution sequence acquisition module 02, obesity trend prediction module 03, obesity prediction result acquisition module 04. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0017] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0018] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in the present invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present invention.

[0019] Embodiment 1, as Figure 1 As shown, the embodiment of the present invention provides a human obesity prediction system based on QCT technology, including:

[0020] The initial fat distribution recognition module 01 is used to collect the initial BMI parameters and initial user characteristic information of the target user, analyze the user representative information, perform QCT detection on the target user to obtain an initial QCT image, configure the QCT recognition accuracy according to the user representative information, and obtain the initial fat distribution by recognizing the initial QCT image.

[0021] Furthermore, the initial fat distribution identification module 01 is also used for:

[0022] Collecting initial BMI parameters and initial user characteristic information of the target user; obtaining a user database, using the initial BMI parameters and initial user characteristic information to search in the user database to obtain the same user ratio; using the same user ratio as the user representative information of the target user.

[0023] Specifically, in traditional obesity prediction methods, BMI (body mass index) is often used as the main indicator for assessing obesity. However, BMI is calculated only through weight and height, and cannot reflect an individual's fat distribution, muscle mass, and its relationship with related diseases. Therefore, it has certain limitations, especially for individuals with abdominal obesity, hidden obesity, etc., the BMI indicator cannot accurately predict their degree of obesity.

[0024] First, the initial BMI parameters and initial user characteristic information of the target user are collected. The initial BMI parameters refer to the weight and height data of the target user. The initial user characteristic information includes factors that affect obesity, such as age, gender, diet structure, exercise habits, and lifestyle. Then, a user database is obtained, which contains BMI parameters and related personalized characteristic information of different users and is constructed by collecting a large amount of user data; then, the initial BMI parameters and initial user characteristic information are used to search in the user database, that is, to find user groups with similar characteristics in the database, match and search according to the similarity, find the most similar user group, and calculate the proportion of users with the same characteristics as the target user in the retrieved user group, that is, the same user ratio. This ratio reflects the representativeness of the target user in the database. The larger the same user ratio, the closer the target user is to the similar user characteristics known in the database. Finally, the obtained same user ratio is used as the user representative information of the target user.

[0025] Furthermore, the initial fat distribution identification module 01 is also used for:

[0026] Performing QCT detection on the target user to obtain an initial QCT image; clustering the same user data in the user database to obtain multiple user data clustering results, and extracting the maximum same user ratio; and calculating the QCT recognition accuracy based on the same user ratio and the maximum same user ratio.

[0027] Specifically, QCT technology is based on chest CT scan data. By using QCT quality control phantoms and analysis software, it has the advantages of no repeated scanning, no increased radiation, and no prolonged scanning time. It can accurately measure abdominal fat area, pancreatic fat content, liver fat content, abdominal muscle area and muscle fat content. For example, QCT can accurately distinguish the distribution of abdominal fat tissue and divide it into subcutaneous fat and visceral fat, thereby accurately identifying individuals with abdominal obesity, especially those with hidden obesity with normal BMI; QCT can also quantify the fat percentage content of pancreatic tissue, which plays an important role in the evaluation of non-alcoholic pancreatitis. In order to achieve more accurate obesity assessment and management, QCT technology can accurately quantify multiple indicators such as visceral fat, subcutaneous fat, liver fat, etc. through CT images, providing a scientific basis for personalized obesity management.

[0028] First, the target user is scanned by CT using QCT technology to obtain detailed image data of the body fat distribution and obtain an initial QCT image. On the other hand, the same user data in the user database is clustered. The clustering method can cluster users with the same BMI and the same user characteristic information based on the user's age, height, weight, diet structure, lifestyle and other characteristic information, or cluster similar user groups together, such as using K-means, hierarchical clustering and other algorithms for clustering analysis to obtain multiple user data clustering results, wherein each user data clustering result is marked with the number of users, that is, the number of the same or similar users; then the maximum number of users in the user data clustering result is selected, and the ratio of the maximum number of users to the total number of users in the user database is set as the maximum same user ratio.

[0029] Then, according to the same user ratio and the maximum same user ratio, the QCT recognition accuracy is calculated. Exemplarily, the ratio of the same user ratio to the maximum same user ratio is calculated, and 1 minus the ratio is used as the QCT recognition accuracy. For example, assuming that the same user ratio is 10% and the maximum same user ratio is 25%, the ratio between the two is 0.4, and the QCT recognition accuracy is 1 minus 0.4, which is 0.6; wherein, when the same user ratio is larger, it means that the target user is more similar to other users, belongs to the majority of users, is more representative, and has a larger amount of corresponding sample data, then a lower recognition accuracy is set, that is, fewer branches are used for recognition, which can save computing resources while maintaining a high accuracy of fat distribution recognition; and the smaller the same user ratio is, the more different the target user is from other users, and belongs to a special category of users, such as athletes. At this time, more branches need to be used for recognition to improve the accuracy of fat distribution recognition. By dynamically adjusting the accuracy of QCT image analysis, computing resources can be reasonably utilized to improve the efficiency and accuracy of QCT images.

[0030] According to the QCT recognition accuracy, the initial QCT image is integratedly recognized to obtain the initial fat distribution.

[0031] Furthermore, the present invention further comprises the following steps:

[0032] A set of sample QCT images is collected, and the fat distribution of the user in each sample QCT image is annotated to obtain a set of sample fat distributions, wherein each sample fat distribution includes the fat areas of multiple body parts of the user; the sample QCT image set and the sample fat distribution set are divided to obtain K sets of QCT supervised training data, where K is an integer greater than 1; the K sets of QCT supervised training data are used to train a QCT image recognition path including K QCT image recognition branches based on a convolutional neural network; the QCT recognition accuracy is multiplied by K and rounded to obtain the number of initial QCT image recognition branches T, where T is an integer greater than or equal to 1 and less than or equal to K; the initial QCT image is input into T random QCT image recognition branches in the QCT image recognition path to obtain T identified fat distributions; the mean of the fat areas of multiple body parts in the T identified fat distributions is calculated to obtain the initial fat distribution of the target user.

[0033] Specifically, based on the QCT image database, a set of sample QCT images is collected, and then the user's fat distribution in each sample QCT image is annotated, that is, the fat area of ​​each body part in the image is marked, such as the fat area of ​​subcutaneous fat, visceral fat, liver fat and other areas. Each annotated fat area should include the value of the fat area to ensure that the data has quantitative characteristics and can provide a reliable basis for subsequent analysis to obtain a sample fat distribution set. Then, the sample QCT image set and the sample fat distribution set are divided into K equal parts, where K is an integer greater than 1, and the specific value of K can be set according to actual needs, for example, K is 10; K pieces of QCT supervised training data are obtained, where each piece of supervised training data includes a certain proportion of sample QCT images and corresponding fat distribution annotations.

[0034] K QCT image recognition branches are constructed based on the convolutional neural network, where each QCT image recognition branch includes an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer. The input data of the input layer is the sample QCT image, and the output data of the output layer is the sample fat distribution. The convolutional layer is used to extract local features in the QCT image, such as fat distribution in the abdomen, liver, pancreas and other areas. The pooling layer is used to reduce the dimension of the feature map (such as maximum pooling or average pooling), reduce the amount of calculation and improve the robustness of the model. The fully connected layer fuses the features extracted from the convolutional layer and the pooling layer. Further, the sample QCT image is used as input, the sample fat distribution is used as supervision, and the K QCT supervised training data are used to supervise the K QCT image recognition branches. During the training process, each QCT image is forward propagated through a convolutional neural network (CNN) branch to generate a preliminary prediction of the fat distribution; then, according to the predicted fat distribution and the actual fat distribution annotation of each QCT image, a loss function (such as mean square error, cross entropy, etc.) is calculated, and the loss function measures the gap between the predicted value and the actual label; further, the gradient is calculated according to the loss function, and the weights in the network are adjusted by back propagation to reduce the error. After each back propagation adjustment, the parameters of the CNN branch will be gradually optimized, and the network can more accurately identify the fat distribution characteristics; iterative training is performed, and the weight parameters of the convolutional neural network are updated using an optimization algorithm (such as Adam optimizer, SGD, etc.) until the model reaches the expected accuracy on the training set, and the trained K QCT image recognition branches are obtained. Finally, a QCT image recognition path is formed according to the K QCT image recognition branches. Through the above training process, the K QCT image recognition branches can learn the fat distribution characteristics of different user groups respectively, and finally generate accurate fat distribution prediction results for each target user.

[0035] The QCT recognition accuracy is further multiplied by K and rounded to obtain the number of initial QCT image recognition branches T, where T is an integer greater than or equal to 1 and less than or equal to K. For example, assuming K is 10 and the QCT recognition accuracy is 0.6, T is 6. In this way, the number of branches is adjusted according to the representative information of the user data, thereby balancing the calculation accuracy and efficiency. Then, T QCT image recognition branches are randomly selected from the K QCT image recognition branches of the QCT image recognition path, and the initial QCT image is input into the T QCT image recognition branches to obtain T identified fat distributions. Finally, the fat areas of multiple body parts in the T identified fat distributions are averaged, and the initial fat distribution of the target user is constructed based on the average of the fat areas of multiple body parts.

[0036] By dynamically adjusting the number of recognition branches T, the QCT image recognition accuracy and computing resources can be fully utilized to improve the flexibility and prediction efficiency of the model. By processing image data in parallel through multiple branches and then synthesizing the final fat distribution through mean calculation, noise and errors can be effectively reduced, thereby obtaining a more stable and accurate fat distribution prediction result.

[0037] The fat distribution sequence acquisition module 02 is used to continue to regularly collect the BMI parameters, user characteristic information and QCT images of the target user, obtain the BMI parameter sequence, user characteristic information sequence and QCT image sequence, analyze the user representative information sequence, configure the QCT recognition accuracy sequence, and analyze and obtain the fat distribution sequence.

[0038] Specifically, a data collection time interval is configured (e.g., 5 days, i.e., data collection is performed every 5 days); according to the collection time interval, the BMI parameters, user characteristic information, and QCT images of the target user continue to be regularly collected, and the BMI parameters, user characteristic information, and QCT images of the target user are arranged in the order of the collection time nodes to obtain a BMI parameter sequence (recording the BMI data of the target user that changes over time), a user characteristic information sequence (a time series containing various characteristic information of the user, such as age, height, dietary structure, lifestyle, etc.), and a QCT image sequence (recording the image data of each QCT scan of the user, for monitoring changes in fat distribution).

[0039] On the other hand, at each data acquisition node, the BMI parameters and user characteristic information are retrieved in the user database to obtain the same user ratio, and then the user representative information sequence is obtained. Then, the ratio of the same user ratio in each user representative information to the maximum same user ratio is calculated, and the ratio is subtracted from 1 to obtain the QCT recognition accuracy, and the QCT recognition accuracy sequence corresponding to the user representative information sequence is obtained. Then, according to the QCT recognition accuracy sequence, the QCT image recognition branches of the adapted number in the QCT image recognition path are called to recognize the QCT image sequence, and the fat distribution sequence corresponding to the QCT image sequence is output.

[0040] The obesity trend prediction module 03 is used to predict the first obesity trend information of the target user according to the BMI parameter sequence and the user characteristic information sequence, and to predict the second obesity trend information of the target user according to the fat distribution sequence.

[0041] Furthermore, the obesity trend prediction module 03 is also used for:

[0042] According to obesity test sample data of multiple users, a sample BMI parameter sequence set, a sample user characteristic information sequence set, a sample fat distribution sequence set and a sample obesity trend information set are collected, wherein each sample obesity trend information includes an obesity probability; the sample BMI parameter sequence set and the sample user characteristic information sequence set are used as binary input data, the sample fat distribution sequence set is used as input data, and the sample obesity trend information set is used to respectively train a first obesity prediction branch and a second obesity prediction branch, and combine to obtain an obesity prediction path; the BMI parameter sequence and the user characteristic information sequence are input into the first obesity prediction branch to predict and obtain the first obesity trend information of the target user, and the fat distribution sequence is input into the second obesity prediction branch to predict and obtain the second obesity trend information of the target user.

[0043] Specifically, first, obesity test sample data of multiple users are obtained, and based on the obesity test sample data, a sample BMI parameter sequence set, a sample user feature information sequence set, a sample fat distribution sequence set and a sample obesity trend information set are collected, wherein the sample BMI parameter sequence, the sample user feature information sequence, the sample fat distribution sequence and the sample obesity trend information have a corresponding relationship, and each sample obesity trend information includes an obesity probability, which reflects the risk of a user developing obesity within a certain period of time. It can be obtained by clinically tracking multiple users to continuously collect sample data and counting the proportion of obese users. The proportion of obese users is used as the obesity probability, and then as the sample obesity trend information. It can also be obtained based on medical evaluation conducted by technical personnel in this field, or it can be calculated through obesity-related indicators and fat distribution data.

[0044] A first obesity prediction branch and a second obesity prediction branch are constructed based on a BP neural network. The first obesity prediction branch and the second obesity prediction branch are BP neural network models that can be iteratively optimized in machine learning. The first obesity prediction branch includes an input layer, multiple hidden layers, and an output layer. The input data of the input layer are a sample BMI parameter sequence and a sample user feature information sequence, and the output data of the output layer is sample obesity trend information. The second obesity prediction branch also includes an input layer, multiple hidden layers, and an output layer. The input data of the input layer is a sample fat distribution sequence, and the output data of the output layer is sample obesity trend information.

[0045] The sample BMI parameter sequence set and the sample user feature information sequence set are further used as binary input data, and the sample obesity trend information set is used as supervision data to perform supervised training on the first obesity prediction branch. First, the BMI parameter sequence and user feature information sequence of each sample are input as binary input data into the input layer of the neural network model. The input data passes through the input layer and multiple hidden layers of the neural network and is finally transmitted to the output layer. The neurons of each layer are weighted and nonlinearly transformed until the output layer generates a predicted value. Then, the obesity trend information generated by the output layer is compared with the actual obesity trend information, and the loss function (such as mean square error, Mean SquaredError), the loss function reflects the gap between the model prediction result and the actual label (obesity trend information); then, the gradient of each neuron is calculated through the back-propagation algorithm, and the gradient value is back-propagated to each previous layer. The purpose of back-propagation is to optimize the weight of the network to minimize the loss function; further, the weights and biases of each layer in the network are updated through optimization algorithms (such as gradient descent) so that the model gradually fits the training data and reduces the prediction error; iterative training is performed, and the parameters of the model are optimized through forward propagation and back-propagation each time until the preset stopping standard is reached (for example, the loss value converges or the maximum number of iterations is reached), and the first obesity prediction branch that has been trained is obtained. After the training is completed, the model can predict obesity trend information, that is, obesity probability, based on the BMI parameter sequence and the user feature information sequence.

[0046] On the other hand, the same method of training the first obesity prediction branch is used, the sample fat distribution sequence set is used as input data, and the sample obesity trend information set is used as supervision data, and the second obesity prediction branch is supervised and trained until the loss value converges or the maximum number of iterations is reached, thereby obtaining a trained second obesity prediction branch. Then, the obesity prediction path is obtained based on the combination of the first obesity prediction branch and the second obesity prediction branch.

[0047] Finally, the BMI parameter sequence and user feature information sequence are input into the first obesity prediction branch to predict the first obesity trend information (first obesity probability) of the target user; the fat distribution sequence is input into the second obesity prediction branch to predict the second obesity trend information (second obesity probability) of the target user.

[0048] The obesity prediction result obtaining module 04 is used to configure weights according to the user representative information sequence, perform weighted calculation on the first obesity trend information and the second obesity trend information, and obtain obesity trend information as an obesity prediction result.

[0049] Furthermore, the obesity prediction result obtaining module 04 is also used for:

[0050] Clustering the same user data in the user database to obtain multiple user data clustering results, and calculating an average ratio of the same users; calculating a ratio of the same user ratio in multiple user representative information in the user representative information sequence to the average ratio of the same users, and calculating the mean to obtain an average user representative degree; taking the average user representative degree and multiplying it by a preset QCT weight to obtain a QCT weight; and calculating the basic data weight according to the QCT weight.

[0051] Specifically, first, cluster the same user data in the user database, such as using K-means, hierarchical clustering and other algorithms to obtain multiple user data clustering results, and calculate multiple identical user ratios of the multiple user data clustering results, that is, the ratio of the number of users to the total number in each user data clustering result; then calculate the mean of the multiple identical user ratios to obtain the average identical user ratio. Further obtain the identical user ratios in the multiple user representative information in the user representative information sequence described in the aforementioned content to obtain multiple identical user ratios, and then calculate the ratios of the multiple identical user ratios of the target user at multiple time nodes to the average identical user ratio, and calculate the mean of the multiple ratios to obtain the average user representativeness.

[0052] QCT identifies fat distribution based on medical images. Due to image errors and training errors, there are certain errors in predicting obesity. Basic data analysis of obesity such as the user's BMI also has certain errors. Therefore, the first and second obesity trend information predicted based on the fat distribution sequence and the user feature information sequence are weighted and fused to improve the accuracy of obesity prediction.

[0053] Specifically, the average user representativeness is multiplied by a preset QCT weight (such as 0.5) to obtain the QCT weight, that is, the QCT weight is dynamically adjusted according to the user representativeness. When the target user is highly representative over a long period of time, it means that the user belongs to the majority of common users, and the accuracy of basic data such as BMI in determining whether the user is obese is low (because common users have normal BMI but more or less fat distribution, while basic data such as BMI of users with rare basic data reflect obesity more accurately, such as lower or higher BMI), and the weight of the QCT technology will be relatively large, indicating that QCT data has a greater impact on obesity prediction and is more accurate.

[0054] Next, the basic data weight is calculated based on the QCT weight, for example, 1 minus the QCT weight, and the difference between the two is used as the basic data weight, and the sum of the QCT weight and the basic data weight is 1. The basic data weight reflects the degree of influence of basic data such as BMI on the prediction results. When the representativeness of the target users is higher (that is, the proportion of the same users is larger), the basic data (such as BMI) is less accurate, the weight of the QCT technology is greater, and the weight of the basic data is smaller.

[0055] For example, if the average user representativeness is 1.1 and the preset QCT weight is 0.5, the calculated QCT weight is 0.55 and the basic data weight is 0.45.

[0056] Furthermore, the obesity prediction result obtaining module 04 is also used for:

[0057] The first obesity trend information and the second obesity trend information are weightedly calculated using the basic data weight and the QCT weight to obtain obesity trend information; and the obesity trend information is used as an obesity prediction result.

[0058] Specifically, the basic data weight and QCT weight are finally used to perform weighted calculation on the first obesity trend information and the second obesity trend information to obtain obesity trend information (comprehensive obesity probability), and the obesity trend information is used as the obesity prediction result. Through this weighting scheme, the weights of QCT technology and basic data in obesity prediction can be dynamically adjusted according to user representative information to improve the accuracy and personalization of obesity prediction. This method takes into account the diversity of user characteristics, avoids the limitations of traditional methods that rely only on static data (such as BMI), and also helps improve the accuracy of QCT-based obesity identification.

[0059] The human obesity prediction system based on QCT technology provided by the embodiment of the present invention has at least the following technical effects:

[0060] By collecting the initial BMI parameters and initial user characteristic information of the target user, analyzing the user representative information, performing QCT detection on the target user to obtain the initial QCT image, configuring the QCT recognition accuracy according to the user representative information, and identifying the initial QCT image to obtain the initial fat distribution; then continuing to regularly collect the BMI parameters, user characteristic information and QCT images of the target user, obtaining the BMI parameter sequence, user characteristic information sequence and QCT image sequence, analyzing the user representative information sequence, configuring the QCT recognition accuracy sequence, and analyzing to obtain the fat distribution sequence; then predicting and obtaining the first obesity trend information of the target user according to the BMI parameter sequence and the user characteristic information sequence, and predicting and obtaining the second obesity trend information of the target user according to the fat distribution sequence; finally configuring the weight according to the user representative information sequence, performing weighted calculation on the first obesity trend information and the second obesity trend information, and obtaining the obesity trend information as the obesity prediction result; that is, by introducing QCT technology and combining user characteristics and BMI parameters to predict human obesity, more accurate and personalized obesity prediction can be achieved, and the accuracy and reliability of human obesity prediction can be significantly improved, thereby providing a scientific basis for human health management.

[0061] Embodiment 2, as Figure 2 As shown, based on the same inventive concept of the human obesity prediction system based on QCT technology provided in Example 1, the embodiment of the present invention also provides a human obesity prediction method based on QCT technology, including:

[0062] Collect the initial BMI parameters and initial user characteristic information of the target user, analyze the user representative information, perform QCT detection on the target user to obtain an initial QCT image, configure the QCT recognition accuracy according to the user representative information, and obtain the initial fat distribution by identifying the initial QCT image; continue to regularly collect the BMI parameters, user characteristic information and QCT images of the target user to obtain a BMI parameter sequence, a user characteristic information sequence and a QCT image sequence, analyze the user representative information sequence, configure the QCT recognition accuracy sequence, and analyze to obtain the fat distribution sequence; predict and obtain the first obesity trend information of the target user according to the BMI parameter sequence and the user characteristic information sequence, and predict and obtain the second obesity trend information of the target user according to the fat distribution sequence; configure weights according to the user representative information sequence, perform weighted calculation on the first obesity trend information and the second obesity trend information, and obtain obesity trend information as an obesity prediction result.

[0063] Furthermore, the human obesity prediction method based on QCT technology also includes: collecting initial BMI parameters and initial user characteristic information of the target user; obtaining a user database, using the initial BMI parameters and initial user characteristic information to search in the user database to obtain the same user ratio; and using the same user ratio as the user representative information of the target user.

[0064] Furthermore, the human obesity prediction method based on QCT technology also includes: performing QCT detection on the target user to obtain an initial QCT image; clustering the same user data in the user database to obtain multiple user data clustering results, and extracting the maximum same user ratio; calculating the QCT recognition accuracy based on the same user ratio and the maximum same user ratio; and performing integrated recognition on the initial QCT image based on the QCT recognition accuracy to obtain the initial fat distribution.

[0065] Furthermore, the human obesity prediction method based on QCT technology also includes: collecting a sample QCT image set, and marking the fat distribution of the user in each sample QCT image to obtain a sample fat distribution set, wherein each sample fat distribution includes the fat area of ​​multiple body parts of the user; dividing the sample QCT image set and the sample fat distribution set to obtain K QCT supervised training data, where K is an integer greater than 1; using the K QCT supervised training data, based on a convolutional neural network, training a QCT image recognition path including K QCT image recognition branches; multiplying the QCT recognition accuracy by K and rounding the result to obtain the initial number of QCT image recognition branches T, where T is an integer greater than or equal to 1 and less than or equal to K; inputting the initial QCT image into T random QCT image recognition branches in the QCT image recognition path to identify and obtain T identified fat distributions; calculating the average of the fat areas of multiple body parts in the T identified fat distributions to obtain the initial fat distribution of the target user.

[0066] Furthermore, the human obesity prediction method based on QCT technology also includes: according to obesity test sample data of multiple users, collecting a sample BMI parameter sequence set, a sample user feature information sequence set, a sample fat distribution sequence set and a sample obesity trend information set, wherein each sample obesity trend information includes an obesity probability; using the sample BMI parameter sequence set and the sample user feature information sequence set as binary input data, using the sample fat distribution sequence set as input data, using the sample obesity trend information set, respectively training a first obesity prediction branch and a second obesity prediction branch, and combining to obtain an obesity prediction path; inputting the BMI parameter sequence and the user feature information sequence into the first obesity prediction branch, predicting to obtain the first obesity trend information of the target user, and inputting the fat distribution sequence into the second obesity prediction branch, predicting to obtain the second obesity trend information of the target user.

[0067] Furthermore, the human obesity prediction method based on QCT technology also includes: clustering the same user data in the user database to obtain multiple user data clustering results, and calculating the average same user ratio; calculating the ratio of the same user ratio in multiple user representative information in the user representative information sequence to the average same user ratio, and calculating the mean to obtain the average user representativeness; using the average user representativeness, multiplying it by a preset QCT weight to obtain the QCT weight; according to the QCT weight, calculating the basic data weight.

[0068] Furthermore, the human obesity prediction method based on QCT technology also includes: using the basic data weight and the QCT weight to perform weighted calculation on the first obesity trend information and the second obesity trend information to obtain obesity trend information; and using the obesity trend information as an obesity prediction result.

[0069] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0070] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0071] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0072] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0073] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0074] Although preferred embodiments of the present invention have been described, additional changes and modifications may occur to these embodiments once those skilled in the art understand the basic inventive concepts.

[0075] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention belong to the scope of the present invention and its equivalent technologies, the present invention is also intended to include these changes and variations.

Claims

1. A human obesity prediction system based on QCT technology, characterized in that: The system comprises: An initial fat distribution recognition module is used to collect the initial BMI parameters and initial user characteristic information of the target user, analyze the user representative information, perform QCT detection on the target user to obtain an initial QCT image, configure the QCT recognition accuracy according to the user representative information, and recognize the initial QCT image to obtain the initial fat distribution; The fat distribution sequence acquisition module is used to continue to regularly collect the BMI parameters, user characteristic information and QCT images of the target user, obtain the BMI parameter sequence, user characteristic information sequence and QCT image sequence, analyze the user representative information sequence, configure the QCT recognition accuracy sequence, and analyze and obtain the fat distribution sequence; An obesity trend prediction module, used to predict and obtain the first obesity trend information of the target user according to the BMI parameter sequence and the user characteristic information sequence, and to predict and obtain the second obesity trend information of the target user according to the fat distribution sequence; The obesity prediction result obtaining module is used to configure weights according to the user representative information sequence, perform weighted calculation on the first obesity trend information and the second obesity trend information, and obtain obesity trend information as an obesity prediction result.

2. The human obesity prediction system based on QCT technology according to claim 1, characterized in that: Collect the target user's initial BMI parameters and initial user characteristic information, and analyze the user's representative information, including: Collecting the target user's initial BMI parameters and initial user characteristic information; Acquire a user database, and use the initial BMI parameter and initial user characteristic information to search in the user database to obtain the same user ratio; The same user ratio is used as user representative information of the target user.

3. The human obesity prediction system based on QCT technology according to claim 2, characterized in that: Performing QCT detection on the target user to obtain an initial QCT image, configuring QCT recognition accuracy according to the user representative information, and recognizing the initial QCT image to obtain initial fat distribution, including: Performing QCT detection on the target user to obtain an initial QCT image; Clustering the same user data in the user database to obtain multiple user data clustering results, and extracting the maximum same user ratio; Calculating and obtaining the QCT recognition accuracy according to the same user ratio and the maximum same user ratio; According to the QCT recognition accuracy, the initial QCT image is integratedly recognized to obtain the initial fat distribution.

4. The human obesity prediction system based on QCT technology according to claim 3, characterized in that: According to the QCT recognition accuracy, the initial QCT image is integratedly recognized to obtain the initial fat distribution, including: Collecting a set of sample QCT images, and marking the fat distribution of the user in each sample QCT image to obtain a set of sample fat distributions, wherein each sample fat distribution includes fat areas of multiple body parts of the user; Dividing the sample QCT image set and the sample fat distribution set to obtain K pieces of QCT supervised training data, where K is an integer greater than 1; Using the K pieces of QCT supervised training data, based on a convolutional neural network, a QCT image recognition path including K QCT image recognition branches is trained; The QCT recognition accuracy is multiplied by K and the result is rounded to obtain the number T of initial QCT image recognition branches, where T is an integer greater than or equal to 1 and less than or equal to K; Inputting the initial QCT image into T random QCT image recognition branches in the QCT image recognition path to obtain T recognized fat distributions; The average of the fat areas of multiple body parts within the T identified fat distributions is calculated to obtain the initial fat distribution of the target user.

5. The human obesity prediction system based on QCT technology according to claim 1, characterized in that: Predicting and obtaining first obesity trend information of the target user according to the BMI parameter sequence and the user characteristic information sequence, and predicting and obtaining second obesity trend information of the target user according to the fat distribution sequence, including: According to the obesity test sample data of multiple users, a sample BMI parameter sequence set, a sample user feature information sequence set, a sample fat distribution sequence set and a sample obesity trend information set are collected, wherein each sample obesity trend information includes an obesity probability; The sample BMI parameter sequence set and the sample user feature information sequence set are used as binary input data, the sample fat distribution sequence set is used as input data, and the sample obesity trend information set is used to train the first obesity prediction branch and the second obesity prediction branch respectively, and the obesity prediction path is obtained by combining them; The BMI parameter sequence and the user characteristic information sequence are input into the first obesity prediction branch to predict and obtain the first obesity trend information of the target user, and the fat distribution sequence is input into the second obesity prediction branch to predict and obtain the second obesity trend information of the target user.

6. The human obesity prediction system based on QCT technology according to claim 2, characterized in that: Configuring weights according to the user representative information sequence includes: Clustering the same user data in the user database to obtain multiple user data clustering results, and calculating an average ratio of the same users; Calculating the ratio of the same user ratio in the plurality of user representative information in the user representative information sequence to the average same user ratio, and calculating the average to obtain the average user representativeness; The average user representativeness is multiplied by a preset QCT weight to obtain a QCT weight; According to the QCT weight, the basic data weight is calculated.

7. The human obesity prediction system based on QCT technology according to claim 6, characterized in that: Performing weighted calculation on the first obesity trend information and the second obesity trend information to obtain obesity trend information as an obesity prediction result includes: Using the basic data weight and the QCT weight, weighted calculation is performed on the first obesity trend information and the second obesity trend information to obtain obesity trend information; The obesity trend information is used as an obesity prediction result.

8. A method for predicting human obesity based on QCT technology, characterized in that: The method is performed by the human obesity prediction system based on QCT technology according to any one of claims 1 to 7, and comprises: Collecting initial BMI parameters and initial user characteristic information of the target user, analyzing user representative information, performing QCT detection on the target user to obtain an initial QCT image, configuring QCT recognition accuracy according to the user representative information, and recognizing the initial QCT image to obtain initial fat distribution; Continue to regularly collect the BMI parameters, user characteristic information and QCT images of the target user, obtain a BMI parameter sequence, a user characteristic information sequence and a QCT image sequence, analyze the user representative information sequence, configure the QCT recognition accuracy sequence, and analyze and obtain the fat distribution sequence; Predicting and obtaining first obesity trend information of the target user according to the BMI parameter sequence and the user characteristic information sequence, and predicting and obtaining second obesity trend information of the target user according to the fat distribution sequence; According to the user representative information sequence, weights are configured, and weighted calculation is performed on the first obesity trend information and the second obesity trend information to obtain obesity trend information as an obesity prediction result.

Images