A portable animal body composition analyzer

Abstract

The present application relates to the technical field of animal body composition detection, and provides a portable animal body composition analyzer, which comprises a data acquisition module responsible for acquiring multi-frequency bioelectrical impedance data and near-infrared spectrum data of domestic pets; a data processing module for carrying out noise filtering, dynamic baseline correction and normalization processing on the collected original data; an algorithm analysis module for correcting data errors by using breed inference and dynamic correction methods, extracting body fat distribution, tissue metabolism and body fluid distribution characteristics by using a multi-modal data fusion method, and calculating a growth and development health score through an integrated learning supervised model; a database management module for storing individual information, appearance characteristics, measurement data and analysis results, supporting data query and backtracking; and a user interface for inputting animal information and hair state, and displaying analysis results, breed similarity and body composition change trend and health score.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of animal body composition detection, and more particularly to a portable animal body composition analyzer. BACKGROUND

[0002] Body composition analysis is a key technology for evaluating the fat, lean body mass (mainly composed of protein and water) and bone mineral content of an organism. By measuring these parameters, the health status, growth and development level and metabolic function of an individual can be understood in depth. For domestic pets in the growth and development period, the dynamic changes of body composition can directly reflect their growth and development status and nutritional health level. Domestic pets experience significant changes in body composition during the growth and development period, including fat accumulation, protein synthesis and body fluid balance regulation. Therefore, accurate measurement of their body composition is crucial for understanding the growth and development process and formulating a scientific feeding plan.

[0003] Document 1 (Scholz AM, et al, Non-invasive methods for the determination of body and carcass composition in livestock. Animal, 2015) points out that body composition analysis technology plays an important role in livestock breeding optimization, health assessment and scientific research. The study details non-invasive technologies including ultrasonic wave (US), dual-energy X-ray absorptiometry (DXA), computed tomography (CT) and magnetic resonance imaging (MRI), which can measure the distribution and volume of tissue fat, lean meat and bone with high precision, providing important support for optimizing animal health management and improving production performance. In particular, the DXA technology is widely recognized for its portability and efficiency, but its device complexity and high cost limit its use in daily monitoring.

[0004] Document 2 (Karelis AD, et al, Validation of a portable bioelectrical impedance analyzer for the assessment of body composition. Applied Physiology, Nutrition, and Metabolism, 2013) studied the accuracy and applicability of a portable bioelectrical impedance analyzer (Inbody 230). The results of the study showed that the device, through multi-frequency bioelectrical impedance measurement technology, can efficiently assess body fat mass, body fat percentage, and lean body mass, and the measurement results have a high correlation with DXA (r = 0.94-0.99). However, the device has systematic bias in the assessment of trunk and limb lean body mass. The portable device is particularly suitable for daily monitoring scenarios outside the laboratory due to its ease of operation and high precision.

[0005] Although existing body composition analysis techniques such as DXA and portable bioelectrical impedance analyzers can meet certain research needs, they still have significant limitations when it comes to the special needs of dynamic body composition monitoring of domestic pets during the growth and development period. For example, existing devices lack the ability to integrate multi-modal data, and cannot fully integrate the distribution information of fat, protein, and water. These technical deficiencies limit the comprehensive understanding and accurate monitoring of the dynamic changes in body composition of domestic pets during the growth and development period. Moreover, due to significant differences in body size, hair characteristics, and body composition distribution among different breeds of domestic pets, the lack of breed-specific correction mechanisms limits the accuracy of the measurement results. In particular, existing devices do not adequately consider the impact of the pet's hair condition, and cannot effectively compensate for the shaved and unshaved states, resulting in measurement bias. At the same time, there is a lack of body composition inference capability for unknown breeds of pets, making it difficult to provide accurate assessments for cross-breed animals. SUMMARY

[0006] To overcome the above-mentioned defects of the prior art, the present application provides a portable animal body composition analyzer, which collects multi-frequency bioelectrical impedance and near-infrared spectroscopy data, uses a breed inference unit and a dynamic correction unit for time series prediction and compensation, extracts body fat distribution, tissue metabolism, and body fluid distribution characteristics using a multi-modal data fusion method, and calculates a growth and development health score based on an integrated learning supervised model, solving the problems of insufficient multi-modal data fusion analysis capability and lack of breed-specific adaptation, hair condition impact compensation, and dynamic correction during the growth and development period of pets, achieving accurate monitoring and health assessment of the body composition of domestic pets during the growth and development period.

[0007] To achieve the above-mentioned purposes, the present application provides the following technical solutions:

[0008] A portable animal body composition analyzer comprises a data acquisition module, a data processing module, a database management module, an algorithm analysis module and a user interface.

[0009] As a further aspect of the present application, the data acquisition module is connected to the data processing module, the data processing module is connected to the algorithm analysis module, the algorithm analysis module is bidirectionally connected to the database management module, the algorithm analysis module is connected to the user interface, and the database management module is connected to the user interface.

[0010] The data acquisition module is configured to acquire raw data of the body composition of a domestic pet from bioelectrical impedance equipment and near-infrared spectroscopy equipment, and the bioelectrical impedance equipment and the near-infrared spectroscopy equipment are configured for a portable device. The bioelectrical impedance technology is used to measure the impedance of body tissue to weak current, and the near-infrared spectroscopy technology is used to analyze the reflection or absorption characteristics of different tissues to light.

[0011] The algorithm analysis module is configured to analyze and calculate the preprocessed bioelectrical impedance data and near-infrared spectroscopy data, including a breed inference unit, a dynamic correction unit, a body composition analysis unit and a health score unit. The breed inference unit is based on appearance feature data and combines a known breed database or cluster analysis to analyze pets of known breeds and unknown breeds, and generates breed similarity information and a breed inference adjustment factor.

[0012] The dynamic correction unit is configured to use a time series model based on historical data and combine the breed inference adjustment factor to dynamically correct the body composition data of a domestic pet, and is also configured to calculate a hair state compensation coefficient based on hair state information.

[0013] The dynamic correction unit is configured to use a time series model based on historical data and combine the breed inference adjustment factor to dynamically correct the body composition data of a domestic pet during growth and development, and output the corrected body composition data. The dynamic correction unit is also configured to receive user input hair state information, calculate a hair state compensation coefficient based on the hair state information and the breed inference adjustment factor, and apply the hair state compensation coefficient to the correction process of the body composition measurement data. The hair state information includes a hair state identifier and a hair feature parameter.

[0014] The body composition analysis unit is configured to receive the corrected data and analyze the body fat distribution, tissue metabolism and body fluid distribution indicators of a domestic pet through a multi-modal data fusion method. The health score unit is configured to evaluate the health status of a domestic pet during growth and development based on an integrated learning supervised model, and calculate a growth and development health score for an individual. The score calculation combines the output results of the breed inference unit for weight adjustment.

[0015] As a further scheme of the present application, the measuring points of the data acquisition module are divided into core measuring points and auxiliary measuring points, the core measuring points include the midline of the neck, the median of the chest, the median of the abdomen, the median of the back and the characteristic measuring points determined according to the animal species, and the positioning is performed based on the anatomical landmark points during measurement; the auxiliary measuring points include the points on both sides of the back and the distal points of the limbs, and the positions thereof are determined according to the relative positions of the core measuring points; the data acquisition module can complete preliminary analysis after obtaining the core measuring point data, and the auxiliary measuring point data is collected again when the accuracy needs to be improved.

[0016] As a further scheme of the present application, the breed inference unit is configured to perform the following operations:

[0017] Receiving pet appearance feature data input by a user through a user interface, the appearance features including body size, body proportion, hair length, fur color, ear shape and tail feature;

[0018] Converting the appearance feature data into a feature vector, comparing the feature vector with a known breed feature vector library in a database, and calculating a similarity;

[0019] According to the similarity, a processing path is determined, when the similarity is higher than a preset threshold, a random forest classification algorithm is used for breed classification; when the similarity is lower than the preset threshold, a K-Means clustering algorithm is used for body type category division; the training data of the random forest classification algorithm is derived from the appearance feature vectors in the known breed database;

[0020] According to the results of the breed classification or the body type category division, a breed inference adjustment factor is calculated, which is used for subsequent calculation of the processing unit.

[0021] As a further scheme of the present application, the breed adjustment factor is a weight coefficient calculated based on the breed similarity, and the generation step of the breed inference adjustment factor includes:

[0022] Step 1: extracting a feature vector according to the pet appearance features input by the user;

[0023] Step 2: calculating the cosine similarity of the feature vector with each breed feature vector in the known breed database;

[0024] Step 3: taking the breed corresponding to the maximum similarity as a candidate breed, and calculating an adjustment factor, wherein:

[0025] ;

[0026] In the formula, is the maximum similarity, is the adjustment factor.

[0027] As a further scheme of the present application, the dynamic correction unit is configured to perform the following operations:

[0028] A historical database is established to store the body composition measurement data of the domestic pet at different growth stages, the measurement data including bioelectrical impedance data and near-infrared spectrum data reflecting the body fat content, protein content, and water content;

[0029] Time series analysis is performed on the historical data to extract the characteristic change pattern of each growth stage, and an error compensation model is constructed;

[0030] The collected data of each measurement point is input into the error compensation model to generate expected body fat content, protein content, and water content parameter values;

[0031] The expected body fat content, protein content, and water content parameter values are compared with the actual measurement values to calculate the error values of each index;

[0032] Based on the error values and the breed inference adjustment factor output by the breed inference unit, a compensation function is constructed for correction, the compensation function considering the relationship between the error vector, the breed standard body weight, and the actual body weight;

[0033] The compensation function is applied to the real-time data of the measurement points to generate corrected index values, and the correction results are stored in the historical database.

[0034] The time series model uses an improved LSTM network, the input layer of which includes historical body composition data (body fat rate, protein content, and water content) and breed inference adjustment factors, the hidden layer is set as a bidirectional gated recurrent unit (BiGRU), and the output layer generates error compensation values through a full connection network; during model training, the historical data is divided using a sliding window method, the window length is 3 months, the step length is 7 days, and the loss function is a weighted combination of MAE and breed weight.

[0035] As a further scheme of the present application, the dynamic correction unit is further configured to perform the following operations:

[0036] The hair state identifier input by the user through the user interface is received, the hair state identifier being used to indicate whether the pet is in a shaved state or a hairy state;

[0037] When the hair state identifier is the hairy state, the hair compensation coefficient library data is called from the database management module;

[0038] According to the breed inference adjustment factor output by the breed inference unit and the hair length grade and hair density characteristic value in the appearance feature data, a hair state compensation coefficient and a dynamic compensation value are calculated; the corrected body composition data is modified as:

[0039] ;

[0040] In the formula, is the original measurement value, is the hair state compensation coefficient, is the dynamic compensation value, is the breed inference adjustment factor;

[0041] The corrected body composition data is output to the body composition analysis unit for subsequent analysis.

[0042] Method for constructing the hair compensation coefficient library:

[0043] Collect the body composition data of the same pet in the shaved and hairy states to form a paired sample set;

[0044] Calculate the hair interference compensation equation through a linear regression model:

[0045] ;

[0046] In the formula, is the hair length grade (1-5 grades), is the hair density characteristic value;

[0047] Pre-store the compensation coefficients of different breeds into the database, and the index key is the breed ID + hair state.

[0048] As a further scheme of the present application, the body composition analysis unit is configured to perform the following operations:

[0049] Receive the corrected data and the original data output by the dynamic correction unit;

[0050] Extract the electrical impedance characteristic parameters of each measurement point from the bioelectrical impedance data to construct a body fat distribution feature vector;

[0051] Extract the spectrum features related to protein from the near-infrared spectrum data to construct a tissue metabolism feature vector;

[0052] Extract the features related to water from the near-infrared spectrum data to construct a body fluid distribution feature vector;

[0053] Use a weighted average or feature splicing fusion method to fuse the three obtained feature vectors, and the fusion process is performed by combining each feature vector through a weight coefficient;

[0054] Calculate the body fat rate index, body fluid distribution index, and tissue metabolism index based on the fused features, and input them into the subsequent unit.

[0055] As a further scheme of the present application, the health score unit is configured to perform the following operations:

[0056] The health evaluation index system for the growth and development period of the domestic pet is constructed, including a body fat rate index, a body fluid distribution index and a tissue metabolism index.

[0057] Based on historical data, a correlation model between the evaluation indexes is constructed, and the output results of the animal category, breed classification and breed inference unit are taken as influencing factors and incorporated into the model.

[0058] The XGBoost integrated learning method is adopted to train the health score model.

[0059] Real-time monitoring data are input into the health score model to calculate the growth and development health score.

[0060] The hyperparameters of the XGBoost model are set as follows: a learning rate of 0.05, a tree depth of 8, a minimum leaf sample number of 10, a regularization term λ of 1.0, and input features including a body fat rate index, a body fluid distribution index, a tissue metabolism index and a breed inference adjustment factor.

[0061] As a further scheme of the present application, the data processing module is configured to perform noise filtering, dynamic baseline correction and normalization processing on the collected original bioelectrical impedance data and near-infrared spectrum data. Noise filtering is mainly aimed at high-frequency interference in bioelectrical impedance signals and random noise in spectrum data; dynamic baseline correction is used to eliminate baseline drift in near-infrared spectrum data; and normalization processing makes data from different sources comparable, facilitating subsequent data fusion and analysis.

[0062] As a further scheme of the present application, the database management module includes: an individual information table for storing basic information and breed information of the domestic pet at different growth and development stages; an appearance feature table for storing characteristic vectors of standard breeds and appearance feature data of unknown breed pets input by the user; a clustering model table, a measurement data table and an analysis result table; and a hair compensation coefficient library for storing hair compensation coefficients under different hair lengths and densities.

[0063] As a further scheme of the present application, the user interface comprises: an animal information input area, including a known breed selection mode and an unknown breed characteristic input mode, the known breed mode being used for selecting an animal category, a specific breed and inputting appearance characteristics, and the unknown breed mode being used for inputting appearance characteristics such as body size, hair characteristics, ear shape and tail characteristics; a hair state selection area, used for selecting a current hair state of the pet, including a shaved state and a hairy state, when the hairy state is selected, a hair length grade and an estimated hair density can be further inputted; a real-time data display area, used for displaying numerical values of a body fat rate index, a body fluid distribution index and a tissue metabolism index, and breed similarity analysis results of the unknown breed pet; and a result analysis area, used for displaying body composition change trends, standard reference ranges and growth and development health scores of four growth and development stages.

[0064] Compared with the prior art, the portable animal body composition analyzer has the beneficial effects that:

[0065] The present application can simultaneously obtain multi-modal data of fat, protein and water of a domestic pet by combining a data acquisition module with multi-frequency bioelectrical impedance technology and near-infrared spectroscopy technology, and can use a breed inference unit and a dynamic correction unit to perform breed-specific adjustment, time series prediction and error compensation on the measurement data. In contrast, the bioelectrical impedance analysis device in the prior art can realize portable measurement, but only relies on single impedance data for body composition analysis, lacks a compensation mechanism for animal breed differences and hair states, and lacks real-time correction capability for dynamic changes, especially in a daily use environment, the data accuracy is difficult to guarantee.

[0066] The algorithm analysis module of the present application adopts a feature fusion algorithm and a Bayesian network model, extracts body composition distribution characteristics by integrating multi-modal data, and calculates a growth and development health score, which can directly complete complex data analysis locally without relying on large computing devices or external data support, which further enhances the advantages of the system in portability and independence. The prior art such as the DXA device has high measurement accuracy, but is large in size and high in cost, and is only suitable for use in fixed laboratory environments, and is difficult to meet the daily family monitoring needs. BRIEF DESCRIPTION OF DRAWINGS

[0067] Figure 1 It is a 3D analysis and reconstruction example diagram for the small sheep image data in document 1.

[0068] Figure 2 It is an ultrasound comparison diagram of back fat of Iberian pigs and German Landrace pigs in document 1.

[0069] Figure 3 It is a structure display diagram of a flexible dry electrode.

[0070] Figure 4A system structure schematic diagram of a portable animal body composition analyzer according to the present application.

[0071] Figure 5 A user interface block diagram of a portable animal body composition analyzer according to the present application.

[0072] Figure 6 An algorithm flow chart in a portable animal body composition analyzer according to the present application. DETAILED DESCRIPTION

[0073] The technical solutions in the embodiments will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the present application.

[0074] A portable animal body composition analyzer comprises a data acquisition module, a data processing module, a database management module, an algorithm analysis module and a user interface.

[0075] The data acquisition module in the embodiments of the present application is connected with the data processing module, the data processing module is connected with the algorithm analysis module, the algorithm analysis module is bidirectionally connected with the database management module, the algorithm analysis module is connected with the user interface, and the database management module is connected with the user interface.

[0076] The data acquisition module is used for acquiring original data of a domestic pet body composition from bioelectrical impedance equipment and near-infrared spectrum equipment, the bioelectrical impedance equipment and the near-infrared spectrum equipment are configured for a portable device; the bioelectrical impedance equipment is used for measuring impedance of a body tissue to a weak current, adopts a four-electrode measurement system, and has a working frequency range of 5 kHz to 1 MHz, and can measure impedance value, phase angle, resistance value and reactance value; the near-infrared spectrum equipment is used for analyzing reflection or absorption characteristics of different tissues to light, has a working wavelength range of 700 nm to 2500 nm, a resolution of 10 nm, and adopts a diffuse reflection mode for measurement.

[0077] The measurement points of the data acquisition module in the embodiment of the present application are divided into core measurement points and auxiliary measurement points, the core measurement points include a neck midline (located on the median line of the neck, at the level of cervical vertebrae C4-C5), a chest midline (located at the midpoint of the sternum), an abdominal midline (located 2 cm above the navel), a back midline (located at the junction of the chest and back, at the level of about T13 vertebrae), and characteristic measurement points determined according to the animal species, and positioning is performed based on anatomical landmark points during measurement; the auxiliary measurement points include two back side points and distal limb points, and the positions of the auxiliary measurement points are determined according to the relative positions of the core measurement points; the data acquisition module can complete preliminary analysis after obtaining the core measurement point data, and the auxiliary measurement point data can be collected when it is necessary to improve the accuracy.

[0078] The data processing module in the embodiment of the present application is configured to perform noise filtering, dynamic baseline correction and normalization processing on the collected raw bioelectrical impedance data and near-infrared spectrum data. Noise filtering is mainly aimed at high-frequency interference in bioelectrical impedance signals and random noise in spectrum data; dynamic baseline correction is used to eliminate baseline drift in near-infrared spectrum data; and normalization processing makes data from different sources comparable, facilitating subsequent data fusion and analysis.

[0079] The algorithm analysis module is used for analyzing and calculating the preprocessed bioelectrical impedance data and near-infrared spectrum data, including a breed inference unit, a dynamic correction unit, a body composition analysis unit and a health score unit.

[0080] The breed inference unit is based on appearance feature data and combines a known breed database or cluster analysis to analyze pets of known breeds and unknown breeds, and generate breed similarity information and breed inference adjustment factors.

[0081] The breed inference unit in the embodiment of the present application is configured to perform the following operations:

[0082] Receiving pet appearance feature data input by a user through a user interface, the appearance features including body size, body proportion, hair length, hair color, ear shape and tail features;

[0083] Converting the appearance feature data into a feature vector, and comparing the feature vector with a known breed feature vector library in a database to calculate a similarity;

[0084] According to the similarity to determine a processing path, when the similarity is higher than a preset threshold, a random forest classification algorithm is used for breed classification; when the similarity is lower than the preset threshold, a K-Means clustering algorithm is used for body type category division; the training data of the random forest classification algorithm is derived from the appearance feature vectors in the known breed database;

[0085] According to the results of breed classification or body type classification, a breed inference adjustment factor is calculated for subsequent processing unit calculation.

[0086] The breed adjustment factor in the embodiment of the present application is a weight coefficient calculated based on breed similarity, and the generation step of the breed inference adjustment factor includes:

[0087] Step 1: Extract the feature vector according to the pet appearance features input by the user;

[0088] Step 2: Calculate the cosine similarity of the feature vector and the feature vector of each breed in the known breed database;

[0089] Step 3: Take the breed corresponding to the maximum similarity as the candidate breed, and calculate the adjustment factor α, wherein:

[0090] ;

[0091] In the formula, is the maximum similarity.

[0092] The dynamic correction unit is configured to adopt a time series model based on historical data and combine the breed inference adjustment factor to dynamically correct the body composition data of the domestic pet during the growth and development period, and output the corrected body composition data.

[0093] The dynamic correction unit in the embodiment of the present application is configured to perform the following operations:

[0094] A historical database is established to store the body composition measurement data of the domestic pet at different growth and development stages, and the measurement data includes bioelectrical impedance data and near-infrared spectrum data reflecting the body fat content, protein content and water content;

[0095] Time series analysis is performed on the historical data to extract the characteristic change pattern of each growth and development stage, and an error compensation model is constructed;

[0096] The data of each measurement point collected is input into the error compensation model to generate expected body fat content, protein content and water content parameter values;

[0097] The expected body fat content, protein content and water content parameter values are compared with the actual measurement values to calculate the error values of each index;

[0098] Based on the error values and the breed inference adjustment factor output by the breed inference unit, a compensation function is constructed for correction, and the compensation function considers the relationship between the error vector, the breed standard body weight and the actual body weight;

[0099] The compensation function is applied to real-time data of the measuring point to generate corrected index values, and the correction result is stored in a historical database.

[0100] The time series model adopts an improved LSTM network, the input layer of which includes historical body composition data (body fat rate, protein content, water content) and breed inference adjustment factors, the hidden layer is provided with a bidirectional gated recurrent unit (BiGRU), and the output layer generates error compensation values through a full connection network; during model training, the historical data is segmented by using a sliding window method, the window length is 3 months, the step length is 7 days, and the loss function is a weighted combination of MAE and breed weight.

[0101] The dynamic correction unit is further configured to receive user input hair state information, calculate a hair state compensation coefficient according to the hair state information and the breed inference adjustment factor, and apply the hair state compensation coefficient to the correction process of the body composition measurement data, wherein the hair state information includes a hair state identifier and a hair feature parameter.

[0102] The dynamic correction unit in the embodiment of the present application is further configured to perform the following operations:

[0103] receiving a hair state identifier input by a user through a user interface, the hair state identifier being used to indicate whether the pet is in a shaved state or a hairy state;

[0104] When the hair state identifier is in the hairy state, the hair compensation coefficient library data is called from the database management module;

[0105] According to the breed inference adjustment factor output by the breed inference unit and the hair length grade and the hair density characteristic value in the appearance feature data, a hair state compensation coefficient and a dynamic compensation value are calculated; the corrected body composition data is corrected as:

[0106] ;

[0107] In the formula, is the original measurement value, is the hair state compensation coefficient, is the dynamic compensation value, is the breed inference adjustment factor;

[0108] The corrected body composition data is output to a body composition analysis unit for subsequent analysis.

[0109] The construction method of the hair compensation coefficient library:

[0110] Collecting body composition data of the same pet in a shaved state and a hairy state to form a paired sample set;

[0111] The hair interference compensation equation is calculated by a linear regression model:

[0112] ;

[0113] wherein, is a hair length grade (1-5 grades), is a hair density characteristic value;

[0114] The compensation coefficients of different breeds are pre-stored in a database, and the index key is breed ID + hair state.

[0115] The historical data is subjected to time series analysis to extract the characteristic change pattern of each growth and development stage:

[0116] The body composition analysis unit is configured to receive the corrected data and analyze the body fat distribution, tissue metabolism and body fluid distribution indexes of the domestic pet through a multi-modal data fusion method.

[0117] As a further scheme of the present application, the body composition analysis unit is configured to perform the following operations:

[0118] Receive the corrected data and original data output by the dynamic correction unit;

[0119] Extract the electrical impedance characteristic parameters of each measurement point from the bioelectrical impedance data to construct a body fat distribution feature vector;

[0120] Extract the protein-related spectral features from the near-infrared spectrum data to construct a tissue metabolism feature vector;

[0121] Extract the water-related features from the near-infrared spectrum data to construct a body fluid distribution feature vector;

[0122] The multi-modal data fusion method of weighted average or feature splicing is used to fuse the three kinds of feature vectors obtained, and the fusion process is performed by combining each feature vector through a weight coefficient;

[0123] Based on the fused features, the body fat rate index, body fluid distribution index and tissue metabolism index are calculated and input into the subsequent unit.

[0124] The health score unit is configured to evaluate the health status of the domestic pet during the growth and development period based on an integrated learning supervised model, and calculate the growth and development health score of the individual, wherein the score calculation combines the output results of the breed inference unit for weight adjustment.

[0125] As a further scheme of the present application, the health score unit is configured to perform the following operations:

[0126] An index system for health assessment of domestic pets during the growth and development period is constructed, including the body fat rate index, the body fluid distribution index and the tissue metabolism index;

[0127] Based on historical data, a correlation model between evaluation indexes is constructed, and the output results of the animal category, breed classification and breed inference unit are taken as influencing factors into the model;

[0128] An XGBoost integrated learning method is used to train a health score model;

[0129] Real-time monitoring data is input into the health score model to calculate the growth and development health score.

[0130] The hyperparameters of the XGBoost model are set as follows: learning rate 0.05, tree depth 8, minimum leaf sample number 10, regularization term λ = 1.0, and input features including body fat rate index, body fluid distribution index, tissue metabolism index and breed inference adjustment factor.

[0131] The database management module in the embodiment of the application comprises: an individual information table for storing basic information and breed information of a domestic pet at different growth and development stages; an appearance feature table for storing feature vectors of standard breeds and appearance feature data of unknown breed pets input by a user; a clustering model table, a measurement data table and an analysis result table; and a hair compensation coefficient library for storing hair compensation coefficients under different hair lengths and densities.

[0132] The user interface in the embodiment of the application comprises: an animal information input area comprising a known breed selection mode and an unknown breed feature input mode, the known breed mode being used to select an animal category, a specific breed and input appearance features, and the unknown breed mode being used to input appearance features such as body size, hair features, ear shape and tail features; a hair state selection area for selecting a current hair state of a pet, comprising a shaved state and a hairy state, when the hairy state is selected, a hair length grade and an estimated hair density can be further input; a real-time data display area for displaying numerical values of a body fat rate index, a body fluid distribution index and a tissue metabolism index and breed similarity analysis results of an unknown breed pet; and a result analysis area for displaying body composition change trends, standard reference ranges and growth and development health scores of four growth and development stages.

[0133] Example 1. In a pet clinic environment, a veterinarian uses a portable animal body composition analyzer to analyze the body composition of a 3-year-old known breed golden retriever. The analysis process is as follows:

[0134] The veterinarian first starts the portable animal body composition analyzer and enters the user interface. In the animal information input area, the veterinarian selects the "known breed selection mode", further selects the animal category as "dog" and the specific breed as "golden retriever", and inputs the basic information of the pet, including the name "Lucky", the age "3 years old", the weight "32 kg", and the gender "male".

[0135] Next, the veterinarian selects the current coat condition of the pet as "with coat" in the coat condition selection area, and further inputs the coat length level as "level 3" (medium-length coat) and the estimated coat density as "0.8" (relatively dense). These information are recorded in the individual information table and the appearance feature table of the database management module.

[0136] After the breed inference unit receives the input pet appearance feature data, since the breed is known, the system directly retrieves the standard feature vector of the golden retriever from the database, and calculates the breed inference adjustment factor a = 1.0 (complete match with the known breed).

[0137] Subsequently, the veterinarian performs data collection. First, the veterinarian uses the bioelectrical impedance device and the near-infrared spectroscopy device to measure the core measurement points. The veterinarian places the measurement probes in the midline of the neck, the midline of the chest, the midline of the abdomen, and the midline of the back of the pet according to the anatomical landmark map displayed by the device. In order to improve the measurement accuracy, the veterinarian also measures the auxiliary measurement points, including the points on both sides of the back and the distal points of the fore and hind limbs.

[0138] After the data collection module collects the raw data, the data is transmitted to the data processing module. The data processing module filters out high-frequency noise from the bioelectrical impedance data, removes random noise from the near-infrared spectroscopy data using wavelet transform method, and performs dynamic baseline correction through polynomial fitting method, and finally performs maximum-minimum normalization processing on the data to make the data from different sources comparable.

[0139] The processed data is transmitted to the dynamic correction unit. The dynamic correction unit retrieves the previous measurement records of the dog from the historical database as a reference, and builds a time series model based on the improved LSTM network. This model takes into account the historical body composition data and the breed inference adjustment factor to generate an error compensation value.

[0140] Since the pet is in a coat condition, the dynamic correction unit also retrieves the hair compensation coefficient library data from the database management module. According to the input coat length level "level 3" and coat density "0.8", the system calculates the hair condition compensation coefficient and the dynamic compensation value. Finally, the original measurement value is corrected to the corrected measurement value through the formula.

[0141] The corrected data is transmitted to the body composition analysis unit. The body composition analysis unit extracts the electrical impedance characteristic parameters of each measurement point from the bioelectrical impedance data, constructs a body fat distribution feature vector; extracts the protein-related spectral features from the near-infrared spectrum data, constructs a tissue metabolism feature vector; and extracts the water-related features from the near-infrared spectrum data, constructs a body fluid distribution feature vector. Subsequently, the system uses a weighted average fusion method to fuse the three feature vectors, and finally calculates the body fat rate, body fluid distribution index and tissue metabolism index.

[0142] Next, the health score unit receives the output results of the body composition analysis unit. Based on the health score model trained by the XGBoost ensemble learning method, the system calculates the growth and development health score of the golden retriever.

[0143] Finally, the real-time data display area of the user interface displays the numerical values of the body fat rate, body fluid distribution and tissue metabolism index. The result analysis area shows the body composition change trend graph of the dog in the adult stage, the standard reference range, and the growth and development health score. The veterinarian can view these data and provide health advice to the pet owner based on the results.

[0144] After the completion of the entire analysis process, the system stores the data of this measurement in the measurement data table and the analysis result table of the database management module, so as to be used as historical data for reference in the future.

[0145] Example 2. This embodiment introduces pheromone auxiliary technology, which is specifically aimed at the problem of unstable pet activity and low measurement cooperation in the home environment. The pheromone auxiliary technology is realized by an integrated or detachable pheromone diffusion module, which can release low-concentration pet calming pheromones (such as cat pheromone F3 component, dog DAP pheromone, etc.) before measurement to reduce the stress response of animals and improve stability during measurement.

[0146] The working mechanism of the pheromone diffusion module is as follows:

[0147] Before the device starts, the user can choose whether to turn on the pheromone diffusion function and its parameter settings;

[0148] After the device starts, the system controls the amount of pheromone release according to the preset process, so that it diffuses at a predetermined time before measurement and continues to release until the end of measurement;

[0149] The pheromone diffusion module uses a low-power fan diffusion method to ensure uniform release while not interfering with the measurement accuracy of the device.

[0150] This technology is particularly suitable for home environments or handling animals that are easily frightened, effectively reducing stress reactions and improving measurement success rates and data stability. The system supports users in selecting different types of pheromone formulations based on animal species to match individual needs of different animals.

[0151] In a home environment, a user uses a portable animal body composition analyzer equipped with this technology to analyze the body composition of a young, unknown breed of feline. The specific operation process is as follows:

[0152] The user takes out the portable animal body composition analyzer and observes that the device has an integrated pheromone diffusion module. Considering the special nature of home environment measurements, the user enables this function, loads the pheromone formulation suitable for felines, and selects "Enable Pheromone Diffusion Function" in the system settings, setting appropriate diffusion parameters.

[0153] Subsequently, the user starts the device and enters the user interface. In the animal information input area, select "Unknown breed feature input mode", specify the animal category as "cat", input the basic information of the animal, including the identification name, age "8 months", actual body weight and gender information. At this time, the pheromone diffusion module starts to work, and the system controls the pheromone to be released according to the preset program.

[0154] The user follows the interface guidance and inputs the appearance feature data of the feline, including body size, body proportion, hair length level, coat color, ear shape, and tail characteristics. In the hair state selection area, the user specifies the current hair state as "hair state" and inputs the corresponding hair parameters.

[0155] The breed inference unit receives the input appearance feature data and converts it into a feature vector, which is compared with the known feline breed feature vectors in the database to calculate the similarity. The system detects that the similarity with a certain breed is below the preset threshold, so it uses a clustering algorithm to divide the body type category and calculates the corresponding breed inference adjustment factor.

[0156] After the pheromone diffusion module works for a preset time, the system prompts that the effective concentration has been reached. The user starts data collection and uses bioelectrical impedance equipment and near-infrared spectroscopy equipment to collect data from the measurement points of the feline. Under the action of pheromones, the stress reaction of the animal is reduced, and it remains in a relatively stable state, allowing the user to complete the core measurement point data collection and also collect data from some auxiliary measurement points.

[0157] After data collection is complete, the collected data is transmitted to the data processing module. The system records relevant information about the use of pheromones during the measurement process. The data processing module performs noise filtering, dynamic baseline correction, and normalization processing on the bioelectrical impedance data and near-infrared spectroscopy data. The system evaluates the stability and quality of the data and confirms that it meets the processing standards.

[0158] The processed data is transmitted to the dynamic correction unit. Since it is the first measurement, the system lacks the historical data of the animal, so it retrieves the standard development data of the same type and age range of animals from the database as a reference. The system generates an error compensation value suitable for young felines based on a time series model.

[0159] Considering that the animal is in a hairy state, the dynamic correction unit retrieves the hair compensation coefficient data from the database, combines the input hair parameters and breed inference adjustment factor to calculate the hair state compensation coefficient and dynamic compensation value, and corrects the original measurement value.

[0160] The corrected data is transmitted to the body composition analysis unit. This unit extracts feature parameters from various data to construct corresponding feature vectors, and calculates body fat rate, body fluid distribution index, and tissue metabolism index through multi-modal data fusion method.

[0161] The health score unit performs health assessment based on the body composition analysis results. The system considers that the animal is in the growth and development period, and adjusts the weight according to the output results of the breed inference unit to calculate the growth and development health score, and provides health status assessment and corresponding suggestions.

[0162] The user interface displays the analysis results, including body composition index values, breed similarity analysis results, and comparison with standard reference range and health score. The system also provides evaluation of the use of pheromone auxiliary function, indicating the effect of this function on improving measurement stability.

[0163] After the measurement is completed, the system stores the data into the database, creates an individual record of the animal, records the measurement data and analysis results, and provides basic data for subsequent monitoring.

[0164] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any skilled person in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0165] Finally, the above is only a preferred embodiment of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A portable animal body composition analyzer comprising a data acquisition module, a data processing module, a database management module and a user interface, characterized in that, Further comprising an algorithm analysis module, wherein the data acquisition module is connected with the data processing module, the data processing module is connected with the algorithm analysis module, the algorithm analysis module is bidirectionally connected with the database management module, the algorithm analysis module is connected with the user interface, and the database management module is connected with the user interface; The data acquisition module is configured to collect raw data of the body composition of the domestic pet from a bioelectrical impedance device and a near-infrared spectroscopy device, and the bioelectrical impedance device and the near-infrared spectroscopy device are configured to be used for a portable device. The algorithm analysis module is configured to analyze and calculate the preprocessed bioelectrical impedance data and near-infrared spectroscopy data, and includes a breed inference unit, a dynamic correction unit, a body composition analysis unit and a health score unit, wherein: the breed inference unit generates breed similarity information and a breed inference adjustment factor based on the appearance feature data; The dynamic correction unit is configured to use a time series model based on historical data and combine the breed inference adjustment factor to dynamically correct the body composition data of the domestic pet, and is further configured to calculate a hair state compensation coefficient according to the hair state information; The body composition analysis unit is configured to receive the corrected data and analyze the body fat distribution, tissue metabolism and body fluid distribution indicators of the domestic pet through a multi-modal data fusion method; and the health score unit is configured to evaluate the health condition of the domestic pet in the growth and development period based on an integrated learning supervised model and calculate the growth and development health score of the individual. The dynamic correction unit is further configured to perform the following operations: receive a hair state identifier input by a user through the user interface, the hair state identifier being used to indicate whether the pet is in a shaved state or a hairy state; when the hair state identifier is in the hairy state, retrieve hair compensation coefficient library data from the database management module; According to the variety inference adjustment factor output by the variety inference unit and the hair length grade and the hair density characteristic value in the appearance characteristic data, a hair state compensation coefficient and a dynamic compensation value are calculated; the corrected body composition data Amended to read: ； wherein is the original measured value, is the hair condition compensation factor, is the dynamic compensation value, is the breed inference adjustment factor; output the corrected body composition data to the body composition analysis unit for subsequent analysis.

2. The portable animal composition analyzer of claim 1, wherein, The measurement points of the data acquisition module are divided into core measurement points and auxiliary measurement points, the core measurement points include the middle line of the neck, the middle of the chest, the middle of the abdomen, the middle of the back and feature measurement points determined according to the animal species, and the positioning is performed based on the anatomical landmark points during measurement; the auxiliary measurement points include points on both sides of the back and distal points of the limbs, and the positions thereof are determined according to the relative positions of the core measurement points; the data acquisition module can complete preliminary analysis after obtaining the core measurement point data, and the auxiliary measurement point data is collected again when the accuracy needs to be improved.

3. The portable animal composition analyzer of claim 1, wherein, The breed inference unit is configured to perform the following operations: receive pet appearance feature data input by a user through the user interface, the appearance features including body size, body proportion, hair length, hair color, ear shape and tail features; convert the appearance feature data into a feature vector and compare it with the known breed feature vector library in the database to calculate the similarity; determine the processing path according to the similarity, when the similarity is higher than a preset threshold, use a random forest classification algorithm for breed classification; when the similarity is lower than the preset threshold, use a K-Means clustering algorithm for body type classification; and the training data of the random forest classification algorithm is derived from the appearance feature vectors in the known breed database. According to the results of breed classification or body type classification, a breed inference adjustment factor is calculated for subsequent processing units.

4. The portable animal composition analyzer of claim 1, wherein, The dynamic correction unit is configured to perform the following operations: A historical database is established to store body composition measurement data of domestic pets at different growth stages, including bioelectrical impedance data and near-infrared spectroscopy data reflecting body fat content, protein content, and water content. Time series analysis is performed on the historical data to extract characteristic change patterns at each growth stage and build an error compensation model. The collected data of each measurement point is input into the error compensation model to generate expected body fat content, protein content, and water content parameter values. The expected body fat content, protein content, and water content parameter values are compared with the actual measurement values to calculate error values for each index. Based on the error values and the breed inference adjustment factor output by the breed inference unit, a compensation function is constructed for correction, considering the relationship between the error vector, the standard body weight of the breed, and the actual body weight. The compensation function is applied to the real-time data of the measurement points to generate corrected index values, and the correction results are stored in the historical database.

5. The portable animal composition analyzer of claim 1, wherein, The body composition analysis unit is configured to perform the following operations: Receive the corrected data and original data output by the dynamic correction unit. Extract bioelectrical impedance characteristic parameters of each measurement point from the bioelectrical impedance data to construct a body fat distribution feature vector. Extract protein-related spectral features from the near-infrared spectroscopy data to construct a tissue metabolism feature vector. Extract water-related features from the near-infrared spectroscopy data to construct a body fluid distribution feature vector. Perform multi-modal data fusion on the three obtained feature vectors using a weighted average or feature splicing fusion method, and the fusion process is performed by combining the feature vectors through weight coefficients. Calculate body fat rate, body fluid distribution, and tissue metabolism indicators based on the fused features and input them into subsequent units.

6. The portable animal composition analyzer of claim 1, wherein, The health score unit is configured to perform the following operations: Construct a health evaluation index system for domestic pets during the growth and development period, including body fat rate, body fluid distribution, and tissue metabolism indicators. Based on the historical data, build a correlation model between the evaluation indicators, and include the output results of the animal category, breed classification, and breed inference unit as influencing factors in the model. Use the XGBoost ensemble learning method to train a health score model. Input real-time monitoring data into the health score model to calculate the growth and development health score.

7. The portable animal composition analyzer of claim 1, wherein, The data processing module is configured to perform noise filtering, dynamic baseline correction, and normalization processing on the collected original bioelectrical impedance data and near-infrared spectroscopy data.

8. The portable animal composition analyzer of claim 1, wherein, The database management module includes: an individual information table for storing basic information and breed information of domestic pets at different growth stages; an appearance feature table for storing characteristic vectors of standard breeds and appearance feature data of unknown breed pets input by users; a clustering model table, a measurement data table, and an analysis result table; a hair compensation coefficient library for storing hair compensation coefficients under different hair lengths and densities.

9. The portable animal composition analyzer of claim 1, wherein, The user interface comprises: an animal information input area, including a known breed selection mode and an unknown breed characteristic input mode, the known breed mode being used for selecting an animal category, a specific breed and inputting appearance characteristics, and the unknown breed mode being used for inputting size, hair characteristics, ear shape and tail appearance characteristics; a hair state selection area, used for selecting a current hair state of the pet, including a shaved state and a hairy state, when the hairy state is selected, further inputting a hair length grade and an estimated hair density; a real-time data display area, used for displaying numerical values of a body fat rate index, a body fluid distribution index and a tissue metabolism index, and breed similarity analysis results of the unknown breed pet; and a result analysis area, used for displaying body composition change trends, standard reference ranges and growth and development health scores of four growth and development stages.

Images