Mare follicular development noninvasive real-time monitoring method and system based on multi-parameter fusion
By combining surface-adhesive monitoring components with deep learning fusion of ultrasound, temperature, and blood flow parameters, the problems of non-invasiveness and accuracy in monitoring mare follicle development have been solved, enabling non-invasive, continuous, and dynamic monitoring of follicle development and prediction of ovulation time.
Patent Information
- Application Number
- CN202511443435.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-10
- Publication Date
- 2026-01-02
AI Technical Summary
Existing mare follicle development monitoring technologies are not non-invasive or continuous enough, rely on professional personnel, have large instantaneous monitoring errors, and are inaccurate in judging single indicators, making it difficult to achieve precise breeding.
Using a surface-adhesive monitoring component, combined with multi-dimensional parameters such as ultrasound, temperature, and blood flow, and through deep learning architecture for feature fusion, it outputs an assessment of follicle development stage and a prediction of ovulation time.
It enables non-invasive, continuous, and dynamic monitoring of follicle development, improves the accuracy of ovulation time prediction, simplifies operation, and lowers the threshold for aquaculture applications.
Smart Images
Figure CN121242622A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of animal breeding technology and medical imaging technology, more particularly to a mare follicle development non-invasive real-time monitoring method and system based on multi-parameter fusion. BACKGROUND
[0002] In modern horse breeding, accurately grasping mare follicle development and ovulation time is the core to improve the pregnancy rate, which is directly related to breeding efficiency and breed inheritance. The current industry "gold standard" is rectal ultrasound scanning technology, which can obtain follicle basic images through the probe to support breeding work, but with the increasing demand for precision and low stress, its disadvantages become more and more obvious: First, the risk of invasive operation is high. The probe needs to be repeatedly inserted through the rectum, which not only causes mare stress, but also easily causes rectal mucosa damage, destroys the intestinal microenvironment, increases the risk of infection, and threatens the health of mares. Second, it is difficult to popularize due to the dependence on professional personnel. The operation requires experienced veterinarians to accurately control the position of the probe, which is limited by personnel capacity. Remote or large-scale horse farms are difficult to monitor in time, and key window periods cannot be covered for 24 hours without interruption. Third, the error of instantaneous monitoring is large. Only the state at a specific time point can be recorded in "snapshot mode", and the development process of follicles cannot be dynamically captured. Due to the randomness of ovulation and individual differences, the ovulation moment is easily missed, and the prediction error is several hours to one day, which reduces the pregnancy rate. Fourth, the single index judgment is not accurate. Only the follicle diameter is relied on, and the key parameters such as endometrial state and ovarian blood flow are ignored. These parameters are closely related to follicle development and implantation environment, and a single model cannot fully reflect the physiological state, affecting the reliability of monitoring.
[0003] In summary, the existing technology has defects in non-invasiveness and continuity, which restricts the breeding efficiency. The industry urgently needs a monitoring method that is non-invasive, continuous, and automatically collects images and fuses multiple physiological parameters to break through the technical limitations and provide support for precise breeding. SUMMARY
[0004] Therefore, the present application provides a mare follicle development non-invasive real-time monitoring method and system based on multi-parameter fusion to solve the problems in the background art.
[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions: A mare follicle development non-invasive real-time monitoring method based on multi-parameter fusion, comprising the following steps: The monitoring area of the mare's lower abdomen corresponding to the ovary and uterus is pretreated, and the monitoring component is arranged in a surface-adhesive manner; The monitoring component emits ultrasonic waves to the mare's ovarian tissue, receives the reflected echo, and generates two-dimensional / three-dimensional ultrasound image data of the ovarian follicle and endometrium, and collects temperature data and light intensity time series data; The received ultrasonic image data, temperature data and light intensity time series data are preprocessed respectively to obtain follicle parameters, endometrial parameters, pelvic temperature parameters and blood flow related parameters of the ovary peripheral tissue; A multi-parameter fusion model is constructed, the obtained follicle parameters, endometrial parameters, pelvic temperature parameters and blood flow related parameters are input into the model, a deep learning architecture is used for feature fusion, and an evaluation result of the mare follicle development stage and a predicted ovulation time are output.
[0006] Optionally, when the predicted ovulation time shows that the mare enters the ovulation window period, an early warning signal is pushed through a mobile terminal APP, the early warning information includes follicle development stage, blood flow state description data, estimated ovulation time and recommended mating window period, and non-invasive real-time monitoring of mare follicle development is completed.
[0007] Optionally, the monitoring assembly includes an ultrasonic monitoring assembly, an infrared temperature monitoring assembly and a diffuse spectrum monitoring assembly, the ultrasonic monitoring assembly emits ultrasonic waves to the mare ovary tissue, receives reflected echoes and generates two-dimensional / three-dimensional ultrasonic image data of the ovary follicle and endometrium, and the infrared temperature monitoring assembly collects pelvic region surface temperature data.
[0008] Optionally, the blood flow related parameters include actual autocorrelation data combination, theoretical autocorrelation data combination based on light intensity time series data, and first mean square root error relative average, second mean square root error relative average and blood flow state description data calculated based on the two types of autocorrelation data combination.
[0009] Optionally, the theoretical autocorrelation data combination is simulated by a Monte Carlo model combined with a semi-infinite model, wherein the semi-infinite model is used to limit the boundary information in the simulation process, and the parameter configuration of the Monte Carlo model includes ovary tissue absorption coefficient, reduced scattering coefficient, light source detector spacing, photon number and anisotropy factor.
[0010] Optionally, the preprocessing of the ultrasonic image data includes beamforming, adaptive Gaussian filter denoising of ultrasonic echo signals, extraction of follicle contour by using Canny edge detection algorithm, and calculation of follicle parameters including follicle diameter, follicle wall thickness and spheroidity; the endometrial image data is subjected to gray scale analysis to obtain endometrial parameters including endometrial thickness and gland density.
[0011] Optionally, the deep learning architecture of the multi-parameter fusion model adopts a convolutional neural network-long short-term memory network fusion architecture; in the feature fusion stage of the multi-parameter fusion model, 25%-30% weight is allocated to the follicle spheroidity, and 20%-25% weight is allocated to the blood flow state description data.
[0012] A mare follicle development non-invasive real-time monitoring system based on multi-parameter fusion comprises: The monitoring component arrangement module is used for pretreating the monitoring area of the mare's lower abdominal surface corresponding to the ovaries and uterus, and arranging the monitoring component in a surface-adhesive manner. The parameter acquisition module is used for emitting ultrasonic waves to the mare's ovarian tissue through the monitoring component, receiving reflected echoes and generating two-dimensional / three-dimensional ultrasonic image data of the ovarian follicle and endometrium, and acquiring temperature data and light intensity time series data. The parameter preprocessing module is used for preprocessing the received ultrasonic image data, temperature data and light intensity time series data respectively, to obtain follicle parameters, endometrial parameters, pelvic temperature parameters and ovarian peripheral tissue blood flow-related parameters. The multi-parameter fusion model construction and ovulation monitoring module is used for constructing a multi-parameter fusion model, inputting the obtained follicle parameters, endometrial parameters, pelvic temperature parameters and blood flow-related parameters into the model, performing feature fusion using a deep learning architecture, and outputting mare follicle development stage evaluation results and ovulation time prediction values.
[0013] Compared with the prior art, the above technical solution provides a mare follicle development non-invasive real-time monitoring method and system based on multi-parameter fusion, which has the following beneficial effects: 1. The surface-adhesive non-invasive monitoring avoids the damage, infection and stress risk of traditional invasive operation, and can be monitored for a long time and multiple times. 2. The multi-dimensional parameters of ultrasound, temperature and blood flow are fused, and the parameter correlation is mined by deep learning to improve the follicle development stage judgment and ovulation time prediction accuracy. 3. The follicle development dynamic change can be continuously and dynamically monitored, the best breeding node can be captured in time, and monitoring lag can be avoided. 4. The operation is simple and easy to use, the device is portable and suitable for mare of different body types, and the data can be stored and traced, reducing the breeding application threshold. BRIEF DESCRIPTION OF DRAWINGS
[0014] In order to more clearly illustrate the technical solutions in the embodiments or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only embodiments of the present application, and those skilled in the art can obtain other drawings according to the provided drawings without creative labor.
[0015] Figure 1 The method flowchart provided by the present application is shown in the figure. Figure 2 The system structure schematic diagram provided by the present application is shown in the figure. DETAILED DESCRIPTION
[0016] The technical solutions in the embodiments of the present application will be apparently and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.
[0017] The embodiment of the present application discloses a non-invasive real-time monitoring method for mare follicle development based on multi-parameter fusion, as shown in the formula: Figure 1 The embodiment of the present application discloses a non-invasive real-time monitoring method for mare follicle development based on multi-parameter fusion, as shown in the formula: The monitoring area corresponding to the ovary and uterus of the mare's lower abdomen is pretreated, and the monitoring components are arranged in a surface-adhesive manner; The ultrasonic wave is emitted to the mare's ovarian tissue through the monitoring component, the reflected echo is received, and the two-dimensional / three-dimensional ultrasonic image data of the ovarian follicle and endometrium is generated, and the temperature data and light intensity time series data are collected; The received ultrasonic image data, temperature data and light intensity time series data are pretreated respectively to obtain follicle parameters, endometrial parameters, pelvic temperature parameters and ovarian peripheral tissue blood flow related parameters; A multi-parameter fusion model is constructed, the obtained follicle parameters, endometrial parameters, pelvic temperature parameters and blood flow related parameters are input into the model, feature fusion is performed by using a deep learning architecture, and the mare follicle development stage evaluation result and ovulation time prediction value are output.
[0018] Specifically: S1: The monitoring area corresponding to the ovary and uterus of the mare's lower abdomen is pretreated, and the ultrasonic monitoring component, infrared temperature monitoring component and diffuse spectrum monitoring component are arranged in a surface-adhesive manner to ensure that each component is closely attached to the mare's surface without invasive contact, and the monitoring point arrangement is completed; wherein the diffuse spectrum monitoring component includes a light source emitting end and a light signal receiving end for collecting light intensity time series data of ovarian peripheral tissue blood flow; S2: Start the monitoring process, emit ultrasonic waves with a frequency of 7-12 MHz to the mare's ovarian tissue through the ultrasonic monitoring component, receive the reflected echo and generate two-dimensional / three-dimensional ultrasonic image data of the ovarian follicle and endometrium; collect pelvic region surface temperature data through the infrared temperature monitoring component; emit near-infrared light to the ovarian peripheral tissue through the light source emitting end of the diffuse spectrum monitoring component, and synchronously collect light intensity time series data through the light signal receiving end, and real-time transmit the ultrasonic image data, temperature data and light intensity time series data to the background analysis system; S3: The background analysis system pretreats the received data: S31: Beamforming and adaptive Gaussian filtering are performed on the ultrasound echo signal for noise reduction. The Canny edge detection algorithm is used to extract the follicle contour, and the follicle diameter, follicle wall thickness, and sphericity are calculated (sphericity is calculated using the formula...). (Calculation: V is the follicle volume, S is the follicle surface area); grayscale analysis is performed on endometrial imaging data to obtain endometrial thickness and glandular density parameters; S32: Filter the body surface temperature data to obtain the real-time temperature change curve of the pelvic region; S33: Perform autocorrelation analysis on the light intensity time series data to obtain the actual autocorrelation data combination of blood flow in the peripheral tissues of the ovary. , ( (For time delay); Based on the Monte Carlo model combined with a semi-infinite model (boundary information is limited by the semi-infinite model, and model parameters include ovarian tissue absorption coefficient, reduced scattering coefficient, and light source detector spacing), the blood flow autocorrelation data generation process is simulated to obtain the theoretical autocorrelation data combination. ; S34: Calculate blood flow-related parameters: using the formula Determine the relative mean of the first root mean square error (where is the number of autocorrelation data points and 'mean' is the theoretical mean of the autocorrelation data points); based on the actual autocorrelation data combination, a predetermined blood flow index data combination (including blood flow rate and blood flow index) is determined using the formula. Determine the relative mean of the second root mean square error , For blood flow parameters, (mean value of blood flow parameters) Through formula , , where is the weighting factor, is used to calculate the state description data C of blood flow in the peripheral tissues of the ovary; S4: Construct a multi-parameter fusion model. Input the follicle parameters obtained in S31, the temperature parameters obtained in S32, and the blood flow status description data C obtained in S34 into the model. Use a CNN-LSTM deep learning architecture to perform feature fusion and output the assessment results of the mare's follicle development stage and the predicted value of ovulation time. The model is trained to convergence using more than 1,000 sets of historical data on mare follicle development containing ultrasound, temperature, and blood flow features. S5: When the ovulation time prediction value shows that the mares have entered the ovulation window, the background analysis system pushes an early warning signal through the mobile terminal APP. The early warning information includes the follicle development stage, sphericity value, blood flow status description data, estimated ovulation time and recommended mating window, thus completing the non-invasive real-time monitoring of mare follicle development.
[0019] Further, in the method for non-invasive real-time monitoring of mare follicular development based on multi-parameter fusion, the pretreatment of the monitoring area on the surface of the mare's lower abdomen corresponding to the ovaries and uterus is a key operation to ensure that the subsequent ultrasonic monitoring component, infrared temperature monitoring component, and diffuse spectral monitoring component can be stably attached to the surface, accurately collect data, and reduce the mare's stress response and data interference. The specific process includes the following complete process: First, the surface of the monitoring area is cleaned. Use soft medical degreasing cotton to dip 37-40℃ (close to the body temperature of the mare, to avoid temperature difference stimulation) sterile saline, gently wipe the surface projection area of the mare's lower abdomen corresponding to the ovaries and uterus (usually 5-15 cm behind the navel of the mare, 8-12 cm on both sides of the middle line of the abdomen, the size needs to be adjusted according to the body size of the mare to accurately cover the surface corresponding position of the ovaries and uterus), remove dirt, hair debris, dandruff, sweat and other impurities in the area; if the hair in this area is long (more than 1 cm), a medical electric shaver should be used to gently shave the hair in the direction of the hair to avoid scratching the skin during shaving, to ensure that the monitoring component is in direct contact with the skin and to reduce the reflection or shielding interference of hair on the ultrasonic signal, infrared temperature signal and near-infrared light signal.
[0020] Secondly, the skin condition of the surface is checked and pretreated. After cleaning, the skin of the monitoring area is checked by visual observation and gentle palpation to check if there are any abnormalities such as skin damage, swelling, scabbing, skin rash or parasitic infection; if there is slight skin dryness, a thin layer of medical coupling agent (the same material as the coupling agent used for subsequent ultrasonic monitoring, to avoid ingredient differences affecting the signal) should be applied for moisturizing to enhance the adhesion of the skin to the monitoring component; if any abnormalities such as skin damage are found, the monitoring point should be adjusted to the adjacent healthy skin area (to ensure that it still covers the monitoring range of the ovaries and uterus) and the abnormal area should be marked for the subsequent care of the breeding personnel.
[0021] Then, the surface is flattened and the coupling agent is applied. For slight skin wrinkles in the monitoring area, the operator should gently fix the mare's abdominal skin with one hand and flatten the wrinkles with the other hand with light force to avoid data collection gaps caused by loose fitting of the monitoring component; then, evenly apply a medical ultrasonic coupling agent (also suitable for ultrasonic monitoring and diffuse spectral monitoring, with good acoustic conductivity and light transmission) with a thickness of 0.5-1mm on the flattened monitoring area, the coupling agent should cover the entire monitoring area (1-2cm larger than the fitting area of the subsequent monitoring component) to ensure that it fills the small depressions on the skin surface and eliminates the air gap between the component and the skin - the air gap will seriously attenuate the ultrasonic signal and near-infrared light signal, and also cause deviation in the infrared temperature monitoring, while the coupling agent can stably penetrate the skin surface layer through its good conductivity, reaching the relevant tissues of the ovaries and uterus.
[0022] Finally, the pre-processed fitting test and adjustment are carried out. After the above operations are completed, the probe of the ultrasonic monitoring assembly is gently attached to the pre-processed monitoring area, and 1-2 groups of ultrasonic signals are preliminarily collected through the background system to observe the signal strength and clarity; if the signal exists fluctuation or interference, it is necessary to check whether the coupling agent is evenly applied, whether the skin is still not cleaned of impurities, or whether the assembly fitting force is moderate (it is appropriate that the assembly does not slip and the mare does not struggle obviously), and targeted adjustment is carried out; at the same time, the infrared temperature monitoring assembly and the diffuse spectral monitoring assembly are subjected to the same fitting test to ensure that each assembly can stably collect data, and the pre-processing operation is completed.
[0023] In S4, the specific construction process of the multi-parameter fusion model needs to be based on data preprocessing, error control as the core, deep learning architecture as the support, and orderly promoted in stages, as follows: First, the selection and standardization processing of model input parameters are completed, the follicle diameter, follicle wall thickness, sphericity, endometrial thickness, gland density obtained by ultrasonic monitoring, the pelvic temperature change curve obtained by infrared monitoring, and the ovarian peripheral tissue blood flow state description data obtained by diffuse spectrum monitoring are taken as the core input parameters, among them, the blood flow state description data needs to be strictly combined according to the file based on the Monte Carlo model combined with the semi-infinite model generation theory, and then the first root mean square error relative mean value is calculated by the actual and theoretical autocorrelation data, the second root mean square error relative mean value is calculated based on the blood flow index data, and finally the flow acquisition of the blood flow state description data is determined by the weighted sum, all parameters need to be converted into dimensionless data to eliminate the interference of different parameter order of magnitude differences on model training. Second, the CNN-LSTM fusion deep learning architecture is built as the main body of the model, among them, the CNN part is used to extract the local features of each parameter, the convolution layer and the pooling layer are used to capture the follicle shape change characteristics from the ultrasonic image derived parameters, extract the temperature fluctuation trend characteristics from the temperature curve, and identify the blood flow stability characteristics from the blood flow state data, and the LSTM part is used to mine the time sequence correlation of each parameter with time, and at the same time, the attention mechanism is introduced into the architecture, referring to the emphasis logic of the file on error control of blood flow parameters, the blood flow state description data and the follicle sphericity are allocated higher feature weights, and the dominant role of key parameters on model output is ensured. Finally, the model training and optimization are completed, the training data set needs to contain more than 1000 groups of monitoring data of mare follicle development in the whole cycle, and the error evaluation logic mentioned in the file is used in the training process, the deviation of the predicted follicle development stage and ovulation time from the actual situation is taken as the basis for calculating the loss function, the model parameters are adjusted by gradient descent algorithm iteration, at the same time, the cross validation method is used to avoid model overfitting, until the prediction error of the model on the ovulation time is stable within the preset range, and the classification accuracy of the follicle development stage is more than 90%, at this time, the multi-parameter fusion model is constructed, which can be used for real-time monitoring and ovulation prediction of mare follicle development.
[0024] With Figure 1 Corresponding to the method shown in the file, the application also discloses a mare follicle development non-invasive real-time monitoring system based on multi-parameter fusion for Figure 1 The implementation of the method, and the specific structure is as shown in Figure 2 The file shows that it includes: The monitoring component layout module is used for pretreating the monitoring area of the mare's lower abdominal surface corresponding to the ovary and uterus, and laying out the monitoring component in a surface fitting manner. The parameter acquisition module is configured to emit ultrasonic waves to the mare's ovarian tissue through the monitoring assembly, receive reflected echoes, and generate two-dimensional / three-dimensional ultrasonic image data of ovarian follicles and endometrium, and collect temperature data and light intensity time series data. The parameter preprocessing module is configured to preprocess the received ultrasonic image data, temperature data, and light intensity time series data respectively, and obtain follicle parameters, endometrial parameters, pelvic temperature parameters, and ovarian peripheral tissue blood flow related parameters. The multi-parameter fusion model construction and ovulation monitoring module is configured to construct a multi-parameter fusion model, input the obtained follicle parameters, endometrial parameters, pelvic temperature parameters, and blood flow related parameters into the model, perform feature fusion using a deep learning architecture, and output mare follicle development stage evaluation results and ovulation time prediction values.
[0025] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed by the embodiments, since it corresponds to the method disclosed by the embodiments, the description is relatively simple, and the related parts can be referred to the method part.
[0026] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for non-invasive real-time monitoring of mare follicular development based on multi-parameter fusion, characterized in that, Includes the following steps: The monitoring areas corresponding to the ovaries and uterus on the lower abdomen of the mare were pretreated, and the monitoring components were installed by attaching them to the body surface. The monitoring component emits ultrasound waves into the ovarian tissue of a mare, receives the reflected echoes, and generates two-dimensional / three-dimensional ultrasound image data of the ovarian follicles and endometrium, while also acquiring temperature data and light intensity time-series data. The received ultrasound image data, temperature data, and light intensity time series data were preprocessed to obtain follicle parameters, endometrial parameters, pelvic temperature parameters, and blood flow-related parameters of the ovarian peripheral tissues. A multi-parameter fusion model was constructed, and the acquired follicle parameters, endometrial parameters, pelvic temperature parameters, and blood flow-related parameters were input into the model. A deep learning architecture was used for feature fusion, and the results of the mare follicle development stage assessment and ovulation time prediction were output.
2. The method for non-invasive real-time monitoring of mare follicle development based on multi-parameter fusion according to claim 1, characterized in that, When the ovulation time prediction value shows that the mares have entered the ovulation window, an early warning signal is pushed through the mobile terminal APP. The early warning information includes follicle development stage, blood flow status description data, estimated ovulation time and recommended mating window, thus completing non-invasive real-time monitoring of mare follicle development.
3. The non-invasive real-time monitoring method for mare follicle development based on multi-parameter fusion according to claim 1, characterized in that, The monitoring components include an ultrasound monitoring component, an infrared temperature monitoring component, and a diffusion spectroscopy monitoring component. The ultrasound monitoring component emits ultrasound waves to the ovarian tissue of the mare, receives the reflected echoes, and generates two-dimensional / three-dimensional ultrasound image data of the ovarian follicles and endometrium. The infrared temperature monitoring component collects surface temperature data of the pelvic region.
4. The non-invasive real-time monitoring method for mare follicle development based on multi-parameter fusion according to claim 1, characterized in that, The blood flow-related parameters include the actual autocorrelation data combination and the theoretical autocorrelation data combination obtained based on light intensity time series data, as well as the first root mean square error relative mean, the second root mean square error relative mean and blood flow state description data calculated based on the two types of autocorrelation data combinations.
5. The non-invasive real-time monitoring method for mare follicle development based on multi-parameter fusion according to claim 4, characterized in that, The theoretical autocorrelation data combination is generated by simulation using a Monte Carlo model combined with a semi-infinite model. The semi-infinite model is used to limit the boundary information in the simulation process. The parameter configuration of the Monte Carlo model includes the ovarian tissue absorption coefficient, the reduced scattering coefficient, the distance between the light source and the detector, the number of photons, and the anisotropy factor.
6. The method for non-invasive real-time monitoring of mare follicle development based on multi-parameter fusion according to claim 1, characterized in that, Preprocessing of ultrasound image data includes: beamforming and adaptive Gaussian filtering to denoise the ultrasound echo signals; extraction of follicle contours using the Canny edge detection algorithm; and calculation of follicle parameters including follicle diameter, follicle wall thickness, and sphericity. Grayscale analysis is performed on endometrial image data to obtain endometrial parameters including endometrial thickness and glandular density.
7. The non-invasive real-time monitoring method for mare follicle development based on multi-parameter fusion according to claim 1, characterized in that, The deep learning architecture of the multi-parameter fusion model adopts a convolutional neural network-long short-term memory network fusion architecture. In the feature fusion stage, the multi-parameter fusion model assigns 25%-30% weight to follicle sphericity and 20%-25% weight to blood flow state description data.
8. A non-invasive real-time monitoring system for mare follicle development based on multi-parameter fusion, characterized in that, include: Monitoring component deployment module: used to preprocess the monitoring areas corresponding to the ovaries and uterus on the lower abdomen of the mare, and to deploy the monitoring components by attaching them to the body surface; Parameter acquisition module: used to transmit ultrasound waves to the ovarian tissue of the mare through the monitoring components, receive the reflected echoes and generate two-dimensional / three-dimensional ultrasound image data of ovarian follicles and endometrium, and acquire temperature data and light intensity time series data; Parameter preprocessing module: used to preprocess the received ultrasound image data, temperature data and light intensity time series data respectively to obtain follicle parameters, endometrial parameters, pelvic temperature parameters and blood flow related parameters of ovarian peripheral tissue; Multi-parameter fusion model construction and ovulation monitoring module: used to construct a multi-parameter fusion model. The model is input with the acquired follicle parameters, endometrial parameters, pelvic temperature parameters and blood flow related parameters. The deep learning architecture is used to perform feature fusion and output the assessment results of the mare's follicle development stage and the predicted value of ovulation time.