Digital content processing method and device based on computer

Abstract

The invention discloses a digital content processing method and device based on a computer, and relates to the field of digital content processing, and the method comprises the following steps: a processing module obtains historical visual recording body weight monitoring information and pregnancy state levels of a pregnant woman from a register, standardizes data, divides a data set, constructs a neural network, and trains the neural network; the training processing module inputs the actual feature vector of the pregnant woman into a trained neural network model to obtain corresponding early warning and display, the processing module fuses images to extract features, analyzes blood vessels, constructs a direction field network, constructs, processes and calculates graphs to obtain a placenta harmony effectiveness balance value, and the processing module compares the placenta harmony effectiveness balance value with a preset comparison interval to obtain a placenta harmony effectiveness comparison result. Corresponding measure processing and early warning are obtained, an efficient fusion processing mechanism of multi-module data can be achieved, and the timeliness and pertinence of medical intervention measures are improved.

Description

Technical Field

[0001] The present invention relates to the field of digital content processing, and particularly to a method and device for processing digital content based on a computer. Background Art

[0002] With the continuous development of the social economy, the attention to human health problems is increasing, especially the monitoring of the health status of pregnant women and fetuses. Although the existing technologies can achieve the single monitoring of some physiological characteristics such as weight and fetal heart rate, there are technical gaps in the fusion analysis of multi-modal data and the intelligent early warning of pregnant women's risks, and it is difficult to meet the clinical needs for the comprehensive and dynamic monitoring of the health status of pregnant women and fetuses. Therefore, a method and device for processing digital content based on a computer have emerged as the times require.

[0003] When the existing method and device for processing digital content based on a computer are running, they lack an efficient fusion processing mechanism for multi-module data, cannot comprehensively utilize multi-dimensional features to evaluate the status of pregnant women, the analysis of the placenta status stays at the basic image observation, and a scientific quantitative evaluation system is not constructed. The early warning of pregnant women's risks depends on single-dimensional features and does not comprehensively analyze multi-features by combining an intelligent model, resulting in low early warning accuracy and insufficient timeliness and pertinence of medical intervention measures.

[0004] To solve the above defects, a technical solution is provided now. Summary of the Invention

[0005] To solve the technical problems raised in the above background art, the present invention is proposed. An embodiment of the present invention provides a method and device for processing digital content based on a computer.

[0006] The object of the present invention can be achieved by the following technical solutions: A method for processing digital content based on a computer includes the following steps:

[0007] Step 1: Train a neural network model. The processing module obtains the historical visual recording weight monitoring information and pregnancy status level of a pregnant woman from a register, standardizes the data, divides the data set, constructs a neural network and trains it;

[0008] Step 2: Corresponding early warning and display. The trained processing module inputs the actual feature vector of the pregnant woman into the trained neural network model to obtain the corresponding early warning and display;

[0009] Step 3: Placenta status analysis. The processing module fuses the images to extract features, analyzes the blood vessels and constructs a direction field network, performs graphic construction processing calculations, and obtains the placental harmony efficiency equilibrium value;

[0010] Step 4: Measure response. The processing module compares the placental harmony efficiency equilibrium value with a preset comparison interval to obtain the corresponding measure processing and early warning.

[0011] Further, the steps for analyzing the trained neural network model are as follows:

[0012] The processing module obtains the historical visual recording weight monitoring information Z of the pregnant woman from the register his and the pregnancy status level G of the pregnant woman his , where Z his represents an m×n matrix, m represents the number of historical samples, n represents the number of features, and the number of features n is composed of four parts of features: vision, weight, monitoring, and recording. Specifically, it includes the weight change rate f1, the body posture tilt angle f2, the skin glossiness f3, the shoulder joint abduction angle f4, the facial muscle micro - tremor frequency f5, the pupil light reflex duration f6, the fetal heart rate f7, the fetal heart rhythm f8, the respiratory rate f9, and the intonation fluctuation f10. Among them, the fetal heart rate f7 refers to the number of fetal heartbeats per minute, the fetal heart rhythm f8 refers to the degree of uniformity of the heart - beat interval, and the intonation fluctuation f10 refers to the degree and frequency of the change in the voice pitch of the pregnant woman during voice communication. A feature vector is established and is standardized using Z - Score normalization respectively to remove the dimension. G his represents a 4×1 vector, representing the pregnancy status level corresponding to each historical sample, specifically the healthy development level, the low - risk controllable level, the medium - risk warning level, and the high - risk intervention level. The processed data is divided into a training set and a validation set with a ratio of 7:3. A neural network structure is constructed with 10 neurons in the input layer, 2 hidden layers are set in the hidden layer. The first hidden layer has 32 neurons, and the second hidden layer has 16 neurons. The activation function of the hidden layer selects the ReLU function. The output layer has 4 neurons, and these 4 neurons will respectively output the probabilities that the sample belongs to the healthy development level, the low - risk controllable level, the medium - risk warning level, and the high - risk intervention level, forming a neural network structure, and using the Gaussian distribution to randomly generate the initial values of the weights and biases. The mean squared error MSE loss function is used to measure the difference between the model prediction value and the true value, and the formula is J is the number of samples, t pred，j is the prediction value of the model for the j - th sample, t true，j is the true value of the j - th sample. The stochastic gradient descent SGD is selected as the optimizer. The data of the training set is input into the model batch by batch for training. After multiple batches of training, a complete training cycle is completed, which is called an epoch. A batch is the batch of data input each time. During the training of each batch, the model calculates the prediction value according to the input data, uses the loss function to calculate the error between the prediction value and the true value, and uses the optimizer to calculate the gradient according to the error to update the weights and biases of the model. When the loss function value of the validation set does not decrease for 10 consecutive epochs, the training is stopped, and the trained model is obtained.

[0013] Further, the corresponding warning and display analysis steps are as follows:

[0014] Input the actual feature vector of the pregnant woman into the trained model to obtain the probabilities of the healthy development level, low-risk controllable level, medium-risk warning level, and high-risk intervention level of the pregnant woman. If the sum of the probabilities of the medium-risk warning level and the high-risk intervention level is greater than 30%, an early warning message is sent to the deep module to send a deep monitoring instruction. If the sum of the probabilities of the medium-risk warning level and the high-risk intervention level is less than or equal to 30%, the information of the pregnant woman is marked in green in the display module.

[0015] Further, the analysis steps of the placental harmony efficiency equilibrium value are as follows:

[0016] Step 305: The processing module normalizes the placental connection heterogeneity dispersion value, the placental hub walking pressure difference value, and the placental isolated connection space distance value. A regular dodecahedron is constructed with the placental isolated connection space distance value as the side length as the basic framework, and then a pyramid is generated with the placental connection heterogeneity dispersion value as a parameter. Finally, the intersection part of the pyramid and the regular dodecahedron is cropped by an inclusion sphere with the placental hub walking pressure difference value as the radius, and the percentage of the sum of the remaining volume and the inclusion sphere volume in the volume of the regular dodecahedron is calculated to obtain the placental harmony efficiency equilibrium value.

[0017] Further, the numerical analysis steps of the placental isolated connection space distance value are as follows:

[0018] Step 304: If the Pearson correlation coefficient between each node and all other nodes is less than the set threshold of 0.1, the blood signal peak frequency distance value is greater than the threshold TQ4, the vascular blood flow resistance index η1 is less than the threshold ψ1, and the pulsatility index η2 is less than the threshold ψ2 in the analysis step 102 by the processing module, then the node is marked as an isolated node. For the finally formed direction field network, the direction field network is divided into several regions. If the weights of the connecting edges in the region are all lower than 0.4, the region is corresponding to a weak coupling slow flow region. If the weights of the connecting edges in the region are all higher than 0.65, the region is corresponding to a smooth flow strong connection region. Otherwise, it belongs to a stable flow uniform connection region. Calculate the shortest path distance between all nodes in the weak coupling slow flow region and the isolated nodes, and mark it as the placental isolated connection space distance value.

[0019] Further, the analysis steps of the placental connection heterogeneity dispersion value and the placental hub walking pressure difference value are as follows:

[0020] Obtain the standard deviation of blood flow velocity in each unobstructed strongly connected region, sum them up to obtain the placental connection heterogeneity dispersion value, calculate the degree of each node in the network, and if it is greater than the set threshold, mark the node as a hub node. Obtain the vascular running pulling rate of the blood vessels connected to the hub node, and mark it as the placental hub running pulling difference value. The vascular running pulling value is the proportion of the abnormal turns, twists or spiral shapes in the running of the blood vessels connected to the hub node. If the bending angle of the blood vessel exceeds 90 degrees and the bending radius is greater than the set threshold, it is determined that the blood vessel is abnormally turning. If the number of bending points of the blood vessel exceeds the set threshold and the number value of the bending direction is greater than the set threshold, it is determined that the blood vessel is abnormally twisted. If the distance from the outermost side of the blood vessel to the central axis is less than 65% of the normal minimum value or greater than 140% of the maximum value, it is determined that the blood vessel is abnormally spiral.

[0021] Furthermore, the connection edge weight analysis steps are as follows:

[0022] Step 303: The processing module establishes a strong connection edge between two nodes that receive similar signals of the first-level blood flow wave, sets the weight of the edge to 0.85, and at the same time, if the difference in the elastic modulus of the blood vessel walls where the two nodes are located is less than 10%, the corresponding weight is increased by 0.05. For two nodes that receive similar signals of the second-level blood flow wave, a connection with an edge weight of 0.65 is established between the main blood vessel nodes, and a connection with an edge weight of 0.55 is established between the microvascular nodes. If the blood flow energy density ratio between the two nodes is between 0.5 and 1.5, the weight is increased by 0.05. For two nodes that receive similar signals of the third-level blood flow wave, the edge weight of the main blood vessel nodes is 0.35, and the edge weight of the microvascular nodes is 0.25. If the difference in the blood flow vortex edge velocity between the two nodes is greater than the threshold TQ3, the corresponding weight is reduced by 0.05.

[0023] Furthermore, the blood flow wave similarity signal analysis steps are as follows:

[0024] Step 301: The imaging module transmits the obtained ultrasonic image and magnetic resonance imaging image to the processing module. The processing module uses scale-invariant feature transform to extract the edge and texture features of the blood vessels in the ultrasonic image, and accelerated robust features to extract the morphological and tissue contrast features of the placenta. The nearest neighbor algorithm calculates the similarity between the extracted features, finds the corresponding feature points in the magnetic resonance imaging image and the ultrasonic image, establishes the matching relationship between the feature points, and based on the matching relationship of the feature points, fuses the magnetic resonance imaging image and the ultrasonic image, performs denoising and enhancement preprocessing operations on the fused image, and uses edge detection and region growing algorithms to extract the features such as the boundary of the placenta, the distribution and morphology of the blood vessels.

[0025] Step 302: The processing module takes the placental vascular voxels extracted from the fused image as the nodes of the direction field network, and obtains the vascular blood flow resistance index η1 and the pulsatility index η2 of two nodes. If the blood flow resistance index η1 is greater than the set threshold TQ1 and the pulsatility index η2 is greater than the set threshold TQ2, it is determined that the blood vessel is a main blood vessel; otherwise, it is determined as a microvessel. The included angle value of the blood flow direction vectors of the two nodes is obtained. If the two-node blood vessels are main blood vessels, when the included angle value is 45 - 90 degrees, it corresponds to a first-level blood vessel angle-like deviation signal; when the included angle value is 30 - 45 degrees (excluding 45 degrees), it corresponds to a second-level blood vessel angle-like deviation signal; when the included angle value is less than 30 degrees, it corresponds to a third-level blood vessel angle-like deviation signal. If the two-node blood vessels are microvessels, when the included angle value is 30 - 90 degrees, it corresponds to a first-level blood vessel angle-like deviation signal; when the included angle value is 15 - 30 degrees (excluding 30 degrees), it corresponds to a second-level blood vessel angle-like deviation signal; when the included angle value is less than 15 degrees, it corresponds to a third-level blood vessel angle-like deviation signal. Using the continuously acquired fused image sequence, the change curve of the direction change of the two nodes over time is calculated, and the Pearson correlation coefficient of the curve is calculated. The time series of the blood flow direction is transformed into the frequency domain by Fourier transform, and the blood signal peak frequency distance value of the two nodes is calculated. If the two nodes are third-level blood vessel angle-like deviation signals and the Pearson correlation coefficient is greater than the threshold 0.8 and the blood signal peak frequency distance value is less than the threshold TQ2, then a first-level blood flow direction wave similarity signal is sent corresponding to the two nodes; if the two nodes are second-level blood vessel angle-like deviation signals and the Pearson correlation coefficient is greater than the threshold 0.8 and the blood signal peak frequency distance value is less than the threshold TQ2, then a second-level blood flow direction wave similarity signal is sent corresponding to the two nodes; otherwise, a third-level blood flow direction wave similarity signal is sent.

[0026] Further, the measure response analysis steps are as follows:

[0027] The processing module compares and analyzes the preset comparison intervals XJ1, XJ2, and XJ3 of the placental harmonic adaptation efficiency value. If the placental harmonic adaptation efficiency value is within the comparison interval XJ1, a signal indicating that the baby's growth state is good is sent, and there is no corresponding operation; if the placental harmonic adaptation efficiency value is within the comparison interval XJ2, a signal indicating that the baby's growth state is average is sent, corresponding to Measure 1; if the placental harmonic adaptation efficiency value is within the comparison interval XJ3, a signal indicating that the baby's growth state is poor is sent and a warning is issued, corresponding to Measure 2.

[0028] As a preferred embodiment of the present invention, a processing device for computer-based digital content includes an imaging module, a weight module, a monitoring module, a vision module, a processing module, a recording module, a depth module, a register, and a display module. The vision module includes a high-definition camera that captures videos at a rate of 60 frames per second from different angles, records the limb movement information and facial expression information of the pregnant woman, and transmits them to the processing module; the weight module includes an electronic weighing scale that monitors the weight change information of the pregnant woman and transmits it to the processing module; the recording module includes a recording device that records the voice of the pregnant woman and the surrounding environment sounds, analyzes audio features such as intonation fluctuations and breathing frequencies, and transmits them to the processing module; the monitoring module includes a fetal heart rate monitoring device that monitors the fetal heart rate information of the pregnant woman and transmits it to the processing module; the depth module is used to receive the warning information from the processing module and issue a depth monitoring instruction; the display module includes a display screen that visually presents the characteristic status information of the pregnant woman;

[0029] The imaging module includes a color Doppler ultrasound device and magnetic resonance imaging. The angle and depth of the ultrasound probe are adjusted to obtain images of different sections of the placenta. Magnetic resonance imaging is used to image the placenta and transmit it to the processing module;

[0030] The processing module trains and processes the historical vision recording weight monitoring feature vector and pregnancy status level vector of the pregnant woman to obtain a trained neural network model, and performs corresponding warnings and displays on the actual feature vector. It fuses the images to extract features, analyzes blood vessels and constructs a direction field network, performs graphic construction processing calculations, and obtains the placental harmony efficiency value;

[0031] The processing module has a built-in register. The register is used to store the neural network model and also stores the historical vision recording weight monitoring information Z his and the pregnancy status level G of the pregnant woman his , and is also used to store the placental image information obtained by the thunder imaging module. The register clears the placental image information in real time after it is read and used.

[0032] Compared with the prior art, the beneficial effects of the present invention are:

[0033] 1. In the present invention, the processing module obtains the historical vision recording weight monitoring information and pregnancy status level of the pregnant woman from the register, standardizes the data, divides the data set, constructs and trains a neural network. The trained processing module inputs the actual feature vector of the pregnant woman into the trained neural network model to obtain corresponding warnings and displays. The processing module fuses the images to extract features, analyzes blood vessels and constructs a direction field network, performs graphic construction processing calculations, and obtains the placental harmony efficiency value. It can efficiently fuse and process multi-module data, comprehensively utilize multi-dimensional features to evaluate the status of the pregnant woman, and not only stay in the basic image observation of the placental status, but also construct a scientific quantitative evaluation system.

[0034] 2. The present invention obtains corresponding measure processing and early warning by processing the placental harmony efficiency value and the preset comparison interval. The dimensional features on which the pregnant woman risk early warning depends are more diverse. It can comprehensively analyze multiple features in combination with an intelligent model, with high early warning accuracy, improving the timeliness and pertinence of medical intervention measures. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for description in the embodiments. The following drawings are not deliberately drawn to scale in actual size, and the focus is on showing the gist of the present invention.

[0036] Figure 1 It is a system block diagram of the present invention;

[0037] Figure 2 It is a method flow chart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0038] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings. Obviously, the described embodiments are only partial embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts also belong to the scope of protection of the present invention.

[0039] As Figure 1 shown, a processing device for digital content based on a computer includes an imaging module, a weight module, a monitoring module, a vision module, a processing module, a recording module, a depth module, a register, and a display module.

[0040] The vision module includes a high-definition camera that collects videos at a rate of 60 frames per second from different angles, records the limb movement information and facial expression information of the pregnant woman, and transmits them to the processing module; the weight module includes an electronic weighing scale that monitors the weight change information of the pregnant woman and transmits it to the processing module; the recording module includes a recording device that records the voice of the pregnant woman and the surrounding environment sounds for analyzing audio features such as intonation fluctuations and breathing frequencies, and transmits them to the processing module; the monitoring module includes a fetal heart monitoring device for monitoring the fetal heart information of the pregnant woman and transmitting it to the processing module; the depth module is used to issue a depth monitoring instruction in response to the early warning information from the processing module; the display module includes a display screen that visually presents the characteristic status information of the pregnant woman;

[0041] The imaging module includes a color Doppler ultrasound device and magnetic resonance imaging. The angle and depth of the ultrasound probe are adjusted to obtain images of different sections of the placenta, and magnetic resonance imaging is used to image the placenta and transmit them to the processing module;

[0042] The processing module trains and processes the historical visual recording weight monitoring feature vector and pregnancy status level vector of the pregnant woman to obtain a trained neural network model, and performs corresponding early warnings and displays on the actual feature vector, extracts features by fusing images, analyzes blood vessels and constructs a direction field network, performs graphic construction processing calculations, and obtains the placental harmony efficiency equilibrium value;

[0043] The processing module has a built-in register, which is used to store the neural network model, and also stores historical visual recording weight monitoring information Z internally his and the pregnancy status level G of the pregnant woman his , and is also used to store the placental image information obtained by the thunder image module. The register clears the placental image information in real time after it is read and used;

[0044] A method for processing digital content based on a computer includes the following steps:

[0045] Step 1: Train the neural network model. The training and processing module obtains the historical visual recording weight monitoring information and pregnancy status level of the pregnant woman from the register, standardizes the data, divides the data set, constructs a neural network and trains it;

[0046] Step 2: Corresponding early warning and display. The training and processing module inputs the actual feature vector of the pregnant woman into the trained neural network model to obtain corresponding early warnings and displays;

[0047] Step 3: Placental status analysis. The processing module extracts features by fusing images, analyzes blood vessels and constructs a direction field network, performs graphic construction processing calculations, and obtains the placental harmony efficiency equilibrium value;

[0048] Step 4: Measure response. The processing module compares the placental harmony efficiency equilibrium value with a preset comparison interval to obtain corresponding measure processing and early warnings;

[0049] Among them, training the neural network model includes the following steps:

[0050] The processing module obtains the historical visual recording weight monitoring information Z of the pregnant woman from the register his and the pregnancy status level G of the pregnant woman his , where Z hisDenote an \(m\times n\) matrix, where \(m\) represents the number of historical samples and \(n\) represents the number of features. The number of features \(n\) is composed of four parts: vision, weight, monitoring, and recording. Specifically, it includes the weight change rate \(f_1\), body tilt angle \(f_2\), skin glossiness \(f_3\), abduction angle of the shoulder joint \(f_4\), micro - tremor frequency of facial muscles \(f_5\), pupillary light reflex duration \(f_6\), fetal heart rate \(f_7\), fetal heart rhythm \(f_8\), respiratory rate \(f_9\), and intonation fluctuation \(f_{10}\). Among them, the fetal heart rate \(f_7\) refers to the number of fetal heartbeats per minute, the fetal heart rhythm \(f_8\) refers to the uniformity of the heart - beat intervals, and the intonation fluctuation \(f_{10}\) refers to the degree and frequency of the change in the pitch of the pregnant woman's voice during speech communication. Establish a feature vector And perform standardization using Z - Score standardization respectively to remove the dimension, \(G\) his Denote a \(4\times1\) vector, representing the pregnancy status level corresponding to each historical sample, specifically including the healthy development level, low - risk controllable level, medium - risk warning level, and high - risk intervention level. Divide the processed data into a training set and a validation set with a ratio of 7:3. Construct an input layer with 10 neurons, set 2 hidden layers. The first hidden layer has 32 neurons, and the second hidden layer has 16 neurons. Select the ReLU function as the activation function for the hidden layers. Construct an output layer with 4 neurons. These 4 neurons will respectively output the probabilities that the sample belongs to the healthy development level, low - risk controllable level, medium - risk warning level, and high - risk intervention level, forming a neural network structure, and use the Gaussian distribution to randomly generate the initial values of the weights and biases. Use the mean squared error (MSE) loss function to measure the difference between the predicted value and the true value of the model. The formula is \(J\) is the number of samples, \(t\) pred，j is the predicted value of the model for the \(j\) - th sample, \(t\) true，j is the true value of the \(j\) - th sample. Select stochastic gradient descent (SGD) as the optimizer. Input the data of the training set batch by batch into the model for training. After multiple batches of training, complete a full training cycle, which is called an epoch. A batch is the batch of data input each time. During the training of each batch, the model calculates the predicted value according to the input data, uses the loss function to calculate the error between the predicted value and the true value, and uses the optimizer to calculate the gradient according to the error to update the weights and biases of the model. When the loss function value of the validation set does not decrease for 10 consecutive epochs, stop training to obtain the trained model;

[0051] Among them, the corresponding warning and display include the following steps:

[0052] The actual feature vector of the pregnant woman Input into the trained model to obtain the probabilities of the healthy development level, low-risk controllable level, medium-risk warning level, and high-risk intervention level of the pregnant woman. If the sum of the probabilities of the medium-risk warning level and the high-risk intervention level is greater than 30%, a warning message is sent to issue a deep monitoring instruction to the deep module. If the sum of the probabilities of the medium-risk warning level and the high-risk intervention level is less than or equal to 30%, the information of the pregnant woman is marked in green in the display module;

[0053] Among them, the placental state analysis includes the following steps:

[0054] Step 301: The imaging module transmits the acquired ultrasound image and magnetic resonance imaging image to the processing module. The processing module uses scale-invariant feature transform to extract the edge and texture features of the blood vessels in the ultrasound image, accelerated robust features to extract the morphological and tissue contrast features of the placenta, and the nearest neighbor algorithm to calculate the similarity between the extracted features, find the corresponding feature points in the magnetic resonance imaging image and the ultrasound image, establish the matching relationship between the feature points, and fuse the magnetic resonance imaging image and the ultrasound image according to the matching relationship of the feature points. Perform denoising and enhancement preprocessing operations on the fused image, and use edge detection and region growing algorithms to extract features such as the boundary of the placenta, the distribution and morphology of blood vessels;

[0055] Step 302: The processing module takes the placental vascular voxels extracted from the fused image as the nodes of the direction field network, and obtains the vascular blood flow resistance index η1 and the pulsatility index η2 of two nodes. If the blood flow resistance index η1 is greater than the set threshold TQ1 and the pulsatility index η2 is greater than the set threshold TQ2, then it is determined that the blood vessel is a main blood vessel; otherwise, it is determined as a micro blood vessel. Obtain the included angle value of the blood flow direction vectors of the two nodes. If the two node blood vessels are main blood vessels, when the included angle value is 45 - 90 degrees, it corresponds to a first-level blood vessel angle-like deviation signal; when the included angle value is 30 - 45 degrees (excluding 45 degrees), it corresponds to a second-level blood vessel angle-like deviation signal; when the included angle value is less than 30 degrees, it corresponds to a third-level blood vessel angle-like deviation signal. If the two node blood vessels are micro blood vessels, when the included angle value is 30 - 90 degrees, it corresponds to a first-level blood vessel angle-like deviation signal; when the included angle value is 15 - 30 degrees (excluding 30 degrees), it corresponds to a second-level blood vessel angle-like deviation signal; when the included angle value is less than 15 degrees, it corresponds to a third-level blood vessel angle-like deviation signal. Using the continuously acquired fused image sequence, calculate the change curve of the direction change of the two nodes over time, calculate the Pearson correlation coefficient of the curve, use Fourier transform to convert the time series of the blood flow direction to the frequency domain, and calculate the blood signal peak frequency distance value of the two nodes. If the two nodes are third-level blood vessel angle-like deviation signals and the Pearson correlation coefficient is greater than the threshold 0.8 and the blood signal peak frequency distance value is less than the threshold TQ2, then a first-level blood flow direction wave similarity signal is sent corresponding to the two nodes. If the two nodes are second-level blood vessel angle-like deviation signals and the Pearson correlation coefficient is greater than the threshold 0.8 and the blood signal peak frequency distance value is less than the threshold TQ2, then a second-level blood flow direction wave similarity signal is sent corresponding to the two nodes. Otherwise, a third-level blood flow direction wave similarity signal is sent. It should be noted that the blood signal peak frequency distance value of the two nodes is the distance value of the peak frequency positions of the blood flow signals of the two nodes;

[0056] Step 303: The processing module establishes a strong connection edge between the two nodes that receive the first-level blood flow direction wave similarity signal, and sets the weight of the edge to 0.85. At the same time, if the difference in the elastic modulus of the blood vessel walls where the two nodes are located is less than 10%, the corresponding weight is increased by 0.05. For the two nodes that receive the second-level blood flow direction wave similarity signal, a connection with an edge weight of 0.65 is established between the main blood vessel nodes, and a connection with an edge weight of 0.55 is established between the micro blood vessel nodes. If the blood flow energy density ratio of the two nodes is between 0.5 and 1.5, the weight is increased by 0.05. For the two nodes that receive the third-level blood flow direction wave similarity signal, the edge weight of the main blood vessel node is 0.35, and the edge weight of the micro blood vessel node is 0.25. If the difference in the blood flow vortex edge speed between the two nodes is greater than the threshold TQ3, the corresponding weight is reduced by 0.05. It should be noted that the difference in the blood flow vortex edge speed between the two nodes is obtained by acquiring the blood flow velocity values at the vortex center and the edge positions through an ultrasonic Doppler instrument and calculating the absolute value of the difference;

[0057] Step 304: The processing module analyzes that in step 102, if the Pearson correlation coefficient between each node and all other nodes is less than the set threshold of 0.1, the blood signal peak frequency distance value is greater than the threshold TQ4, the vascular blood flow resistance index η1 is less than the threshold ψ1, and the pulsatility index η2 is less than the threshold ψ2, then the node is marked as an isolated node. For the finally formed orientation field network, the orientation field network is divided into several regions. If the weights of the connecting edges within the region are all lower than 0.4, then the region corresponds to a weakly coupled slow-flow region. If the weights of the connecting edges within the region are all higher than 0.65, then the region corresponds to a smooth-flow strongly coupled region. Otherwise, it belongs to a steady-flow evenly coupled region. Calculate the shortest path distance between all nodes in the weakly coupled slow-flow region and the isolated nodes, which is marked as the placental retention-isolated node connection space distance value. Obtain the standard deviation of the blood flow velocity in each smooth-flow strongly coupled region and sum them to get the placental smooth connection heterogeneity dispersion value. Calculate the degree of each node in the network. If it is greater than the set threshold, then the node is marked as a hub node. Obtain the vascular path pulling rate of the blood vessels connected to the hub node, which is marked as the placental hub path pulling heterogeneity value. It should be noted that the degree of each node is the number of edges connected to the node, and the vascular path pulling value is the proportion of the abnormal cases where the paths of the blood vessels connected to the hub node turn, twist, or are spiral. If the bending angle of the blood vessel exceeds 90 degrees and the bending radius is greater than the set threshold, then the blood vessel is determined to be abnormally turning. If the number of bending points of the blood vessel exceeds the set threshold and the number of bending direction values is greater than the set threshold, then the blood vessel is determined to be abnormally twisted. If the distance from the outermost side of the blood vessel to the central axis is less than 65% of the normal minimum value or greater than 140% of the maximum value, then the blood vessel is determined to be spirally abnormal;

[0058] Step 305: The processing module normalizes the values of the placental smooth connection heterogeneity dispersion value, the placental hub path pulling heterogeneity value, and the placental retention-isolated node connection space distance value, and shrinks the range to [0, 1]. Construct a regular dodecahedron with the placental retention-isolated node connection space distance value as the edge length as the basic framework, then generate a pyramid with the placental smooth connection heterogeneity dispersion value as a parameter. Finally, cut the intersection part of the pyramid and the regular dodecahedron with an inclusion sphere with the placental hub path pulling heterogeneity value as the radius, and calculate the percentage of the sum of the remaining volume and the inclusion sphere volume in the volume of the regular dodecahedron to obtain the placental harmony efficiency balance value;

[0059] Among them, the measure response includes the following steps:

[0060] The processing module compares and analyzes the preset comparison intervals XJ1, XJ2, and XJ3 of the placental harmony efficiency value. If the placental harmony efficiency value is within the comparison interval XJ1, it sends a signal indicating that the baby is in good growth condition and there is no corresponding operation. If the placental harmony efficiency value is within the comparison interval XJ2, it sends a signal indicating that the baby's growth condition is average, increases the frequency of fetal heart monitoring and ultrasound examinations, and dynamically evaluates the fetal development. If the placental harmony efficiency value is within the comparison interval XJ3, it sends a signal indicating that the baby's growth condition is poor, issues a warning, and initiates medical intervention, such as deeply investigating placental pathological problems, implementing clinical treatment, and considering terminating the pregnancy in advance in combination with the gestational week and fetal condition.

[0061] The above is a description of the present invention and should not be construed as a limitation thereof. Although several exemplary embodiments of the present invention have been described, those skilled in the art will readily understand that many modifications can be made to the exemplary embodiments without departing from the novel teachings and advantages of the present invention. Therefore, all such modifications are intended to be included within the scope of the present invention as defined by the claims. It should be understood that the above is a description of the present invention and should not be considered limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present invention is defined by the claims and their equivalents.

Claims

1. A method for processing computer-based digital content, characterized in that, It includes the following steps: Step 1: Train a neural network model. The processing module obtains the pregnant woman's historical visual recording, weight monitoring information, and pregnancy status level from the register, standardizes the data, divides the data set, constructs a neural network, and trains it; Step 2: Corresponding warning and display. The trained processing module inputs the actual feature vector of the pregnant woman into the trained neural network model to obtain the corresponding warning and display; Step 3: Placenta state analysis. The processing module fuses the features extracted from the image, analyzes the blood vessels, constructs a direction field network, performs graphic construction and processing calculations, and obtains the placental harmony efficiency equilibrium value; Step 4: Measure response. The processing module compares the placental harmony efficiency equilibrium value with the preset comparison interval to obtain the corresponding measure processing and warning.

2. A method for processing computer-based digital content according to claim 1, characterized in that The analysis steps of the trained neural network model are as follows: The processing module obtains the historical visual recording weight monitoring information Z of the pregnant woman from the register his and the pregnancy status level G of the pregnant woman his , where Z his represents an m×n matrix, m represents the number of historical samples, n represents the number of features, and the number of features n is composed of four parts of features: vision, weight, monitoring, and recording, specifically including the weight change rate f1, body posture tilt angle f2, skin glossiness f3, shoulder joint abduction angle f4, facial muscle microtremor frequency f5, pupillary light reflex duration f6, fetal heart rate f7, fetal heart rhythm f8, respiratory rate f9, and intonation fluctuation f10. Among them, the fetal heart rate f7 refers to the number of fetal heartbeats per minute, the fetal heart rhythm f8 refers to the uniformity of the heartbeat interval, and the intonation fluctuation f10 refers to the degree and frequency of the change in the pitch of the pregnant woman's voice during voice communication. A feature vector is established. And Z-Score standardization is respectively used for standardization to remove the dimension, G his represents a 4×1 vector, indicating the pregnancy status level corresponding to each historical sample, specifically the healthy development level, low-risk controllable level, medium-risk warning level, and high-risk intervention level. The processed data is divided into a training set and a validation set in a ratio of 7:

3. An input layer with 10 neurons is constructed, and 2 hidden layers are set. The first hidden layer has 32 neurons, and the second hidden layer has 16 neurons. The activation function of the hidden layer selects the ReLU function. An output layer with 4 neurons is constructed. These 4 neurons will respectively output the probabilities that the sample belongs to the healthy development level, low-risk controllable level, medium-risk warning level, and high-risk intervention level, forming a neural network structure. And the Gaussian distribution is used to randomly generate the initial values of the weights and biases. The mean squared error MSE loss function is used to measure the difference between the predicted value and the true value of the model. The formula is J is the number of samples, t pred，j is the predicted value of the model for the j-th sample, t true，j is the true value of the j-th sample. Stochastic gradient descent SGD is selected as the optimizer. The data of the training set is input into the model batch by batch for training. After multiple batches of training, a complete training cycle is completed, which is called an epoch. A batch is the batch of data input each time. During the training of each batch, the model calculates the predicted value according to the input data, uses the loss function to calculate the error between the predicted value and the true value, and uses the optimizer to calculate the gradient according to the error to update the weights and biases of the model. When the loss function value of the validation set does not decrease for 10 consecutive epochs, the training is stopped, and the trained model is obtained.

3. A method for processing digital content based on a computer according to claim 1, characterized in that, The analysis steps of the corresponding warning and display are as follows: Input the actual feature vector of the pregnant woman into the trained model to obtain the probabilities of the healthy development level, low-risk controllable level, medium-risk warning level, and high-risk intervention level of the pregnant woman. If the sum of the probabilities of the medium-risk warning level and the high-risk intervention level is greater than 30%, a warning message is sent to the depth module to send a depth monitoring instruction. If the sum of the probabilities of the medium-risk warning level and the high-risk intervention level is less than or equal to 30%, the information of the pregnant woman is marked in green in the display module.

4. A method for processing computer-based digital content according to claim 1, characterized in that, The analysis steps of the placental harmony efficiency equilibrium value are as follows: Step 305: The processing module normalizes the values of the placental connection heterogeneity dispersion value, the placental hub walking pressure difference value, and the placental isolated connection space distance value. A regular dodecahedron is constructed with the placental isolated connection space distance value as the edge length as the basic framework, and then a pyramid is generated with the placental connection heterogeneity dispersion value as the parameter. Finally, the intersection part of the pyramid and the regular dodecahedron is trimmed with an inclusion sphere with the placental hub walking pressure difference value as the radius, and the percentage of the sum of the remaining volume and the inclusion sphere volume in the regular dodecahedron volume is calculated to obtain the placental harmony efficiency equilibrium value.

5. A method for processing computer-based digital content according to claim 4, characterized in that, The numerical analysis steps of the placental isolated connection space distance value are as follows: Step 304: If the Pearson correlation coefficient between each node and all other nodes is less than the set threshold of 0.1, the blood signal peak frequency distance value is greater than the threshold TQ4, the blood vessel blood flow resistance index η1 is less than the threshold ψ1, and the pulsatility index η2 is less than the threshold ψ2 in the analysis step 102 of the processing module, then the node is marked as an isolated node. For the finally formed direction field network, the direction field network is divided into several regions. If the connection edge weights within the region are all lower than 0.4, then the region is corresponding to a weak coupling slow flow region. If the connection edge weights within the region are all higher than 0.65, then the region is corresponding to a smooth flow strong connection region, and the others are classified as a stable flow uniform connection region. The shortest path distance between all nodes in the weak coupling slow flow region and the isolated nodes is calculated and marked as the placental isolated connection space distance value.

6. A method for processing digital content based on a computer according to claim 4, characterized in that, The analysis steps of the placental connection heterogeneity dispersion value and the placental hub walking pressure difference value are as follows: Obtain the standard deviation of blood flow velocity in each unobstructed strongly connected region, and sum them to obtain the placental connectivity heterogeneity dispersion value. Calculate the degree of each node in the network. If it is greater than the set threshold, mark the node as a hub node. Obtain the vascular trajectory compression rate of the blood vessels connected to the hub node, and mark it as the placental hub trajectory compression difference value. The vascular trajectory compression value is the proportion of the trajectories of the blood vessels connected to the hub node that show abnormal turning, twisting, or spiral shapes. If the bending angle of the blood vessel exceeds 90 degrees and the bending radius is greater than the set threshold, the blood vessel is determined to be abnormally turned. If the number of bending points of the blood vessel exceeds the set threshold and the number value of the bending direction is greater than the set threshold, the blood vessel is determined to be abnormally twisted. If the distance from the outermost side to the central axis of the blood vessel is less than 65% of the normal minimum value or greater than 140% of the maximum value, the blood vessel is determined to be abnormally spiral-shaped.

7. A method for processing computer-based digital content according to claim 5, characterized in that, The steps for analyzing the weights of the connection edges are as follows: Step 303: The processing module establishes a strong connection edge between two nodes when they receive similar first-level blood flow wave signals, and sets the weight of the edge to 0.

85. At the same time, if the difference in the elastic modulus of the blood vessel walls where the two nodes are located is less than 10%, the corresponding weight is increased by 0.

05. When the two nodes receive similar second-level blood flow wave signals, a connection with a weight of 0.65 is established between the main blood vessel nodes, and a connection with a weight of 0.55 is established between the small blood vessel nodes. If the ratio of the blood flow energy density of the two nodes is between 0.5 and 1.5, the weight is increased by 0.

05. When the two nodes receive similar third-level blood flow wave signals, the weight of the main blood vessel node edge is 0.35, and the weight of the small blood vessel node edge is 0.

25. If the difference in the blood flow vortex edge velocity around the two nodes is greater than the threshold TQ3, the corresponding weight is reduced by 0.

05.

8. A method for processing digital content based on a computer according to claim 7, characterized in that, The steps for analyzing the similar blood flow wave signals are as follows: Step 301: The imaging module transmits the obtained ultrasound image and magnetic resonance imaging image to the processing module. The processing module uses scale-invariant feature transform to extract the edge and texture features of the blood vessels in the ultrasound image, and accelerated robust features to extract the morphological and tissue contrast features of the placenta. The nearest neighbor algorithm calculates the similarity between the extracted features, finds the corresponding feature points in the magnetic resonance imaging image and the ultrasound image, establishes a matching relationship between the feature points, and based on the matching relationship of the feature points, fuses the magnetic resonance imaging image and the ultrasound image. Perform denoising and enhancement preprocessing operations on the fused image, and use edge detection and region growing algorithms to extract features such as the boundary of the placenta, the distribution and morphology of the blood vessels. Step 302: The processing module takes the placental vascular voxels extracted from the fused image as the nodes of the direction field network, and obtains the vascular blood flow resistance index η1 and the pulsatility index η2 of two nodes. If the blood flow resistance index η1 is greater than the set threshold TQ1 and the pulsatility index η2 is greater than the set threshold TQ2, it is determined that the blood vessel is a main blood vessel; otherwise, it is determined to be a microvessel. The included angle value of the blood flow direction vectors of the two nodes is obtained. If the two-node blood vessels are main blood vessels, when the included angle value is 45 - 90 degrees, it corresponds to a first-level blood vessel angle-like deviation signal; when the included angle value is 30 - 45 degrees (excluding 45 degrees), it corresponds to a second-level blood vessel angle-like deviation signal; when the included angle value is less than 30 degrees, it corresponds to a third-level blood vessel angle-like deviation signal. If the two-node blood vessels are microvessels, when the included angle value is 30 - 90 degrees, it corresponds to a first-level blood vessel angle-like deviation signal; when the included angle value is 15 - 30 degrees (excluding 30 degrees), it corresponds to a second-level blood vessel angle-like deviation signal; when the included angle value is less than 15 degrees, it corresponds to a third-level blood vessel angle-like deviation signal. Using the continuously acquired fused image sequence, the change curve of the direction change of the two nodes over time is calculated, and the Pearson correlation coefficient of the curve is calculated. The time series of the blood flow direction is transformed into the frequency domain by Fourier transform, and the blood signal peak frequency distance value of the two nodes is calculated. If the two nodes are in the third-level blood vessel angle-like deviation signal and the Pearson correlation coefficient is greater than the threshold 0.8 and the blood signal peak frequency distance value is less than the threshold TQ2, then a first-level blood flow direction wave similarity signal is sent corresponding to the two nodes; if the two nodes are in the second-level blood vessel angle-like deviation signal and the Pearson correlation coefficient is greater than the threshold 0.8 and the blood signal peak frequency distance value is less than the threshold TQ2, then a second-level blood flow direction wave similarity signal is sent corresponding to the two nodes; otherwise, a third-level blood flow direction wave similarity signal is sent.

9. A method for processing digital content based on a computer according to claim 1, characterized in that, The measures for response analysis are as follows: The processing module compares and analyzes the placental harmony efficiency threshold with the preset comparison intervals XJ1, XJ2, and XJ3. If the placental harmony efficiency threshold is within the comparison interval XJ1, a signal indicating that the baby's growth status is good is sent, and there is no corresponding operation; if the placental harmony efficiency threshold is within the comparison interval XJ2, a signal indicating that the baby's growth status is average is sent, corresponding to measure one; if the placental harmony efficiency threshold is within the comparison interval XJ3, a signal indicating that the baby's growth status is poor is sent, and a warning is issued, corresponding to measure two.

10. A processing device for computer-based digital content, characterized in that Applied to a computer-based digital content processing method according to any one of claims 1-9, which includes an imaging module, a weight module, a monitoring module, a vision module, a processing module, a recording module, a depth module, a register, and a display module. The vision module includes a high-definition camera that acquires videos at a rate of 60 frames per second from different angles, records the limb movement information and facial expression information of the pregnant woman, and transmits them to the processing module; the weight module includes an electronic weighing scale that monitors the weight change information of the pregnant woman and transmits it to the processing module; the recording module includes a recording device that collects the voice of the pregnant woman and the surrounding environment sounds, analyzes audio features such as intonation fluctuations and breathing frequencies, and transmits them to the processing module; the monitoring module includes a fetal heart monitoring device that monitors the fetal heart information of the pregnant woman and transmits it to the processing module; the depth module is used to receive the warning information from the processing module and issue a depth monitoring instruction; the display module includes a display screen that visually presents the characteristic status information of the pregnant woman; The imaging module includes a color Doppler ultrasound device and magnetic resonance imaging. Adjust the angle and depth of the ultrasound probe to obtain images of different sections of the placenta, perform imaging on the placenta using magnetic resonance imaging, and transmit them to the processing module; The processing module trains and processes the historical vision, recording, weight, and monitoring feature vectors of the pregnant woman and the pregnancy status level vector to obtain a trained neural network model, and performs corresponding warnings and displays on the actual feature vectors, fuses the features extracted from the imaging, analyzes the blood vessels and constructs a direction field network, performs graphic construction processing calculations, and obtains the placental harmony efficiency balance value; The processing module has built-in registers, which are used to store the neural network model and also store historical visual recording body weight monitoring information Z his and the pregnancy status level G of the pregnant woman his , and is also used to store the placental image information obtained by the thunder imaging module. The register clears the placental image information in real time after it is read and used.

Images