CN109875556A - A kind of intelligence just before giving birth monitors system - Google Patents
A kind of intelligence just before giving birth monitors system Download PDFInfo
- Publication number
- CN109875556A CN109875556A CN201910162216.0A CN201910162216A CN109875556A CN 109875556 A CN109875556 A CN 109875556A CN 201910162216 A CN201910162216 A CN 201910162216A CN 109875556 A CN109875556 A CN 109875556A
- Authority
- CN
- China
- Prior art keywords
- layer
- giving birth
- before giving
- network
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Measuring And Recording Apparatus For Diagnosis (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention discloses a kind of intelligence just before giving birth to monitor system, which includes: uterine myoelectricity acquisition processing device and intelligent terminal, the uterine myoelectricity acquisition processing device and intelligent terminal communication connection；The uterine myoelectricity acquisition processing device extracts from collected electromyography signal for acquiring electromyography signal and obtains the first uterine contraction signal data；The intelligent terminal includes: at least one processor and at least one processor, processor is for realizing a kind of intelligence just before giving birth prediction technique step, the method step includes: that the input of the first uterine contraction signal data is just before giving birth identified prototype network to predict, thus output just before giving birth state；First uterine contraction signal data input Time in labor prediction model network is predicted, to export prediction Time in labor.By using the present invention, puerpera can be helped to grasp the production status of itself in real time, protect the safety of puerpera and fetus.The present invention just before giving birth monitors system as a kind of intelligence and can be widely applied in just before giving birth monitoring field.
Description
Technical field
The present invention relates to just before giving birth monitoring field more particularly to a kind of intelligence just before giving birth to monitor system.
Background technique
Just before giving birth predict to puerpera production and medical staff it is significant, accurate judgement just before giving birth feature carries out Operational preparation,
Delivery room is timely entered, is the important prerequisite for guaranteeing puerpera and smoothly producing.Currently, most of just before giving birth judge according further to puerpera
A possibility that judgement that itself makes the impression of uterine contraction, the first-born puerpera is inexperienced, generates erroneous judgement is larger, and two tires or more
The puerpera of tire experience, which is then possible to these factors insensitive on uterine contraction reaction, etc., can all influence just before giving birth to judge.If by instrument
Device auxiliary helps puerpera to grasp the production status of itself in real time, this is to protection by the method prediction just before giving birth signal of artificial intelligence
Puerpera and fetus are all of great significance.
Summary of the invention
In order to solve the above-mentioned technical problem, the object of the present invention is to provide a kind of intelligence just before giving birth to monitor system.
The embodiment of the invention provides a kind of intelligence just before giving birth to monitor system, including uterine myoelectricity acquisition processing device and intelligence
Terminal, the uterine myoelectricity acquisition processing device and intelligent terminal communication connection；
The uterine myoelectricity acquisition processing device is extracted from collected electromyography signal for acquiring electromyography signal
To the first uterine contraction signal data；
The intelligent terminal, comprising:
At least one processor；
At least one processor, for storing at least one program；
When at least one described program is executed by least one described processor, so that at least one described processor is realized
A kind of intelligence just before giving birth prediction technique step, the method step include:
According to the first uterine contraction signal data, just before giving birth status predication processing and/or Time in labor prediction processing are carried out；
Wherein, the just before giving birth status predication processing comprising:
The first testing data snapshot is obtained after doing the first pretreatment to the first uterine contraction signal data acquired, by first
The input of testing data snapshot just before giving birth identifies prototype network and carries out prediction processing, thus output just before giving birth state；
The Time in labor prediction processing comprising:
The second testing data snapshot is obtained after doing the second pretreatment to the first uterine contraction signal data acquired, by second
Testing data snapshot input Time in labor prediction model network is predicted, to export prediction Time in labor.
Further, the prototype network that just before giving birth identifies is obtained by following model training step:
Building just before giving birth identifies prototype network, obtains the second uterine contraction signal data, and the second uterine contraction signal data is inputted
Network training is carried out to prototype network is just before giving birth identified；
And/or the Time in labor prediction model network is obtained by following model training step:
Time in labor prediction model network is constructed, third uterine contraction signal data is obtained, by the third uterine contraction signal data
It is input to Time in labor prediction model network and carries out network training.
Further, the building just before giving birth identifies the step for prototype network, specifically includes:
First layer prototype network is constructed, the first layer prototype network includes: the first convolutional layer, the first normalization layer and the
The output end of one function actuation layer, first convolutional layer connects by the input terminal of the first normalization layer and first function excitation layer
It connects；
Second layer prototype network is constructed, the second layer prototype network includes: the second convolutional layer, the second normalization layer, the
Two function actuation layers, the first random deactivating layer and third convolutional layer, the output end of second convolutional layer successively pass through second and return
One change layer, second function excitation layer, the first random deactivating layer are connect with the input terminal of third convolutional layer, the first function excitation
The output end of layer is connect with the input terminal of the second convolutional layer；
Construct third layer prototype network, the third layer prototype network include by several memory unit layers be connected in series and
At memory unit layer group, the output end of the third convolutional layer connect with the input terminal of memory unit layer group；
The 4th layer model network is constructed, the 4th layer model network includes: third normalization layer, third function actuation
Layer, the first full articulamentum and the 4th function actuation layer, the output end of second convolutional layer successively pass through third normalization layer, the
Three function actuation layers, the first full articulamentum are connect with the input terminal of the 4th function actuation layer, the output of the memory unit layer group
It holds and is connect with the input terminal of third normalization layer.
Further, it is described by the second uterine contraction signal data be input to just before giving birth identify prototype network carry out network training this
One step, specifically includes:
It is obtained after successively carrying out the processing of the first maximum value down-sampling and first order combined treatment to the second uterine contraction signal data
To multiple first data snapshots；
First data snapshot and just before giving birth label are input to and just before giving birth identify prototype network the first training managing of progress；
It is described by the first data snapshot and just before giving birth label be input to just before giving birth identify prototype network carry out the first training managing this
One step includes: to be input to the first data snapshot and just before giving birth label just before giving birth to identify to obtain the first spy after prototype network is handled
Sign is as a result, judge whether fisrt feature result falls into the first just before giving birth characteristic interval, if so, the first just before giving birth characteristic interval is constant,
Conversely, then the first just before giving birth characteristic interval is expanded, in the just before giving birth characteristic interval of first after enabling fisrt feature result fall into expansion；
The first training managing step is repeated, all training managing has finished up to the first all data snapshots.
Further, it the step for the building Time in labor prediction model network, specifically includes:
First layer prototype network is constructed, the first layer prototype network includes: Volume Four lamination, the 4th normalization layer and the
The output end of five function actuation layers, the Volume Four lamination connects by the input terminal of the 4th normalization layer and the 5th function actuation layer
It connects；
Second layer prototype network is constructed, the second layer prototype network includes: the 5th convolutional layer, the 4th normalization layer, the
Six function actuation layers, the second random deactivating layer and the 6th convolutional layer, the output end of the 5th convolutional layer successively pass through the 4th and return
One change layer, the 6th function actuation layer, the second random deactivating layer are connect with the input terminal of the 6th convolutional layer, the 5th function actuation
The output end of layer is connect with the input terminal of the 5th convolutional layer；
Third layer prototype network is constructed, the third layer prototype network includes being connected by multiple identical second residual block layers
The residual block layer group of composition, wherein each second residual block layer include: the 5th normalization layer, the 7th function actuation layer, third with
Machine deactivating layer, the 7th convolutional layer, the 6th normalization layer, the 8th function actuation layer, the 4th random deactivating layer and the 8th convolutional layer, institute
The input terminal for stating the 5th normalization layer successively passes through the 7th function actuation layer, the random deactivating layer of third, the 7th convolutional layer, the 6th returns
One change layer, the 8th function actuation layer, the 4th random deactivating layer are connect with the output end of the 8th convolutional layer, the 6th convolutional layer
Output end is connect with the input terminal of residual block layer group；
The 4th layer model network is constructed, the 4th layer model network includes: the 7th normalization layer, the 9th function actuation
Layer, the second full articulamentum and the tenth function actuation layer, the input terminal of the normalization layer successively pass through the 7th normalization layer, the 9th
Function actuation layer, the second full articulamentum are connect with the output end of the tenth function actuation layer, the output end of the residual block layer group with
The input terminal connection of 7th normalization layer.
Further, described that the third uterine contraction signal data is input to Time in labor prediction model network progress network instruction
It the step for practicing, specifically includes:
It is obtained after successively carrying out the processing of the second maximum value down-sampling and second order combined treatment to third uterine contraction signal data
To multiple second data snapshots；
Second data snapshot and just before giving birth label are input to Time in labor prediction model network and carry out the second training managing；
It is described that second data snapshot and just before giving birth label are input at the second training of Time in labor prediction model network progress
The step for reason includes: that the second data snapshot and just before giving birth label are input to after Time in labor prediction model network is handled to obtain
To second feature as a result, judging whether second feature result falls into the second just before giving birth characteristic interval, if so, the second just before giving birth characteristic area
Between it is constant, conversely, then expanding the second just before giving birth characteristic interval, second feature result is enabled to fall into the second just before giving birth characteristic area after expanding
Between in；
The second training managing step is repeated, all training managing has finished up to the second all data snapshots.
Further, the method step is further comprising the steps of:
Obtain pregnant week data and amniotic fluid status data；
When according to pregnant week data and amniotic fluid status data to judge that current state is just before giving birth situation, then issue just before giving birth
Pre-warning signal.
Further, the uterine myoelectricity acquisition processing device is connected by blue tooth wireless communication mode and intelligent terminal communication
It connects.
Further, the intelligent terminal includes smart phone, IPAD, laptop computer and/or desktop computer.
Further, the uterine myoelectricity acquisition processing device realizes that myoelectricity is believed using four emg electrode signal pickup assemblies
Number acquisition.
The beneficial effects of the present invention are: by using the embodiment of the present invention, using the uterine contraction signal data acquired into
Row just before giving birth status predication processing and/or Time in labor prediction processing only need to assist by instrument, pass through the method for artificial intelligence
Prediction just before giving birth state and Time in labor, can help puerpera to grasp the production status of itself in real time, protect the peace of puerpera and fetus
Entirely, and just before giving birth identification model and Time in labor prediction model is utilized, therefore has the advantages that pinpoint accuracy and high-timeliness.
Detailed description of the invention
Fig. 1 is the first specific embodiment structural schematic diagram that a kind of intelligence of the present invention just before giving birth monitors system；
Fig. 2 is the step flow chart of the first specific embodiment of intelligence just before giving birth prediction technique of the invention.
Specific embodiment
The present invention is described in further detail in the following with reference to the drawings and specific embodiments.In for the examples below
Number of steps is arranged only for the purposes of illustrating explanation, does not do any restriction to the sequence between step, each in embodiment
The execution sequence of step can be adaptively adjusted according to the understanding of those skilled in the art.
As shown in Figure 1, the embodiment of the invention provides a kind of intelligence just before giving birth to monitor system, including uterine myoelectricity acquisition process
Device and intelligent terminal, the uterine myoelectricity acquisition processing device and intelligent terminal communication connection；
The uterine myoelectricity acquisition processing device is extracted from collected electromyography signal for acquiring electromyography signal
To the first uterine contraction signal data；
The intelligent terminal, comprising:
At least one processor；
At least one processor, for storing at least one program；
When at least one described program is executed by least one described processor, so that at least one described processor is realized
A kind of intelligence just before giving birth prediction technique step, as shown in Fig. 2, the method step includes:
S101, according to the first uterine contraction signal data, carry out just before giving birth status predication and handle；
S102, according to the first uterine contraction signal data, carry out Time in labor prediction processing.
Wherein, the step for just before giving birth status predication processing, includes:
S1011, the first testing data snapshot is obtained after doing the first pretreatment to the first uterine contraction signal data acquired,
The input of first testing data snapshot is just before giving birth identified into prototype network and carries out prediction processing, thus output just before giving birth state；
The Time in labor predicts that the step for processing includes:
S1021, the second testing data snapshot is obtained after doing the second pretreatment to the first uterine contraction signal data acquired,
Second testing data snapshot input Time in labor prediction model network is predicted, to export prediction Time in labor.
Specifically, when carrying out just before giving birth status predication, the first uterine contraction signal data long to 10 minutes acquired into
The long interception of row every five seconds clock does the first maximum value to each data slot to obtain the data slot after 120 interceptions
The processing of down-sampling value, obtains 49 numerical value being sequentially arranged, by each data slice from each data slot
Numerical value in section carries out permutation and combination, and the matrix of a 7*7 is successively arranged out in the way of a line by every 7 numerical value, and as the
One testing data snapshot, to obtain 120 the first testing data snapshots.This 120 the first testing data snapshots are input to
It has constructed in trained just before giving birth identification prototype network, the just before giving birth state of finally output prediction, output result includes: just before giving birth, not
Just before giving birth, data exception etc., and alarm signal is issued, to prompt prediction result to come out；When carrying out Time in labor prediction, to obtaining
The first long uterine contraction signal data of 10 minutes obtained carries out the long interception of every five seconds clock, to obtain the number after 120 interceptions
According to segment, the processing of the second maximum value down-sampling value is done to each data slot, obtains 2209 from each data slot
Numerical value in each data slot is carried out permutation and combination, successively with every 47 numerical value by a numerical value being sequentially arranged
The matrix of a 47*47, as the second testing data snapshot are arranged out for the mode of a line, thus obtain 120 it is second to be measured
Data snapshot.This 120 the second testing data snapshots are input to and have been constructed in trained Time in labor prediction model network,
The Time in labor of prediction is finally exported, output result includes: time numerical value, data exception etc., and issues alarm signal, with prompt
Prediction result has come out.
It is obtained by above-mentioned, by using the embodiment of the present invention, pregnant woman can be made to can be clearly understood that the current of oneself
Just before giving birth state, prepare to carry out generation, timely enter delivery room, will not be insufficient because of experience and some reactions of body are done
The decision to make mistake, the insufficient young puerpera of experience also no longer need the experience for excessively relying on elder to instruct to do-it-yourself,
Very big protective action is played to the safety of pregnant woman and baby.
It is further used as the preferred embodiment of this system, the prototype network that just before giving birth identifies is walked by following model training
Suddenly it obtains:
S201, building just before giving birth identify prototype network, the second uterine contraction signal data are obtained, by the second uterine contraction signal data
It is input to and just before giving birth identifies prototype network progress network training；
Specifically, situation there are two types of, one is: the training step for first carrying out just before giving birth identifying prototype network, then carry out just before giving birth
The training step of time prediction prototype network；Another kind is: the first training step of progress Time in labor prediction model network, then into
Row just before giving birth identifies the training step of prototype network, and the embodiment of the present invention uses the training step for first carrying out just before giving birth identifying prototype network
Suddenly, then the training step of Time in labor prediction model network is carried out.
The Time in labor prediction model network is obtained by following model training step:
S202, building Time in labor prediction model network, obtain third uterine contraction signal data, by the third uterine contraction signal
Data are input to Time in labor prediction model network and carry out network training.
It is obtained by above-mentioned, by using the embodiment of the present invention, prototype network can be trained by way of science,
To obtain the just before giving birth identification prototype network and Time in labor prediction model network of science, so that prediction is more accurate, and have
It is of practical significance.
It is further used as the preferred embodiment of this system, it is described to construct the step for just before giving birth identifying prototype network S201,
It specifically includes:
S2011, building first layer prototype network, the first layer prototype network include: the first convolutional layer, the first normalization
Layer and first function excitation layer, the output end of first convolutional layer are defeated by the first normalization layer and first function excitation layer
Enter end connection；
Specifically, construct first layer prototype network, the layer network include a conv convolutional layer, BN normalization layer and
The function actuation layer that one ReLU activation primitive is constituted, wherein conv layers of convolution kernel is 7*7*120, and first layer prototype network is defeated
Value is high dimensional feature result out.
S2012, building second layer prototype network, the second layer prototype network include: the second convolutional layer, the second normalization
Layer, second function excitation layer, the first random deactivating layer and third convolutional layer, the output end of second convolutional layer successively pass through
Two normalization layers, second function excitation layer, the first random deactivating layer are connect with the input terminal of third convolutional layer, the first function
The output end of excitation layer is connect with the input terminal of the second convolutional layer；
Specifically, construct second layer prototype network, the layer network include a conv convolutional layer, BN normalization layer,
Function actuation layer, the random deactivating layer of a Dropout and another conv convolutional layer that one ReLU activation primitive is constituted, first
The convolution kernel of a conv is 5*5*120, and the inactivation rate of Dropout is set as 50%, and second convolution kernel is 3*3*60, the second layer
The input value of prototype network is the output valve of first layer prototype network, and the output valve of second layer prototype network is one-dimensional data feature
As a result.
S2013, building third layer prototype network, the third layer prototype network include being connected by several memory unit layers
The output end of the memory unit layer group being formed by connecting, the third convolutional layer is connect with the input terminal of memory unit layer group；
Specifically, third layer prototype network is constructed, the layer network is by 5 concatenated identical LSTM memory units group layer by layer
At the output dimension of LSTM is 60, and the input value of third layer prototype network is the output valve of second layer prototype network, third layer mould
The output valve of type network is the characteristic results of 60 dimensions.
S2014, the 4th layer model network of building, the 4th layer model network include: third normalization layer, third function
Excitation layer, the first full articulamentum and the 4th function actuation layer, the output end of second convolutional layer successively pass through third normalization
Layer, third function actuation layer, the first full articulamentum are connect with the input terminal of the 4th function actuation layer, the memory unit layer group
Output end is connect with the input terminal of third normalization layer；
Specifically, the 4th layer model network is constructed, which activates letter by ReLU including a BN normalization layer, one
Excitation layer, the full articulamentum of a FC and the function actuation layer being made of sigmoid activation primitive that number is constituted, the 4th layer of mould
The input value of type network is the output valve of third layer prototype network, and the output valve of the 4th layer model network is just before giving birth to identify model net
The output valve of network.
It is further used as the preferred embodiment of this system, described be input to the second uterine contraction signal data just before giving birth identifies
Prototype network carries out the step for network training S201, specifically includes:
S2015, the second uterine contraction signal data is successively carried out at the processing of the first maximum value down-sampling and first order combination
Multiple first data snapshots are obtained after reason；
Specifically, first intercepting the second uterine contraction signal data is 10 minutes length, then is carried out to the second uterine contraction signal data
The long interception of every five seconds clock does under the first maximum value each data slot to obtain the data slot after 120 interceptions
Value processing is sampled, 49 numerical value being sequentially arranged are obtained from each data slot, by each data slot
In numerical value carry out permutation and combination, the matrix of a 7*7 is successively arranged out in the way of a line by every 7 numerical value, as first
Testing data snapshot, to obtain 120 the first data snapshots.
S2016, the first data snapshot and just before giving birth label are input to just before giving birth the first training managing of identification prototype network progress；
It is described by the first data snapshot and just before giving birth label be input to just before giving birth identify prototype network carry out the first training managing this
One step includes: to be input to the first data snapshot and just before giving birth label just before giving birth to identify to obtain the first spy after prototype network is handled
Sign is as a result, judge whether fisrt feature result falls into the first just before giving birth characteristic interval, if so, the first just before giving birth characteristic interval is constant,
Conversely, then the first just before giving birth characteristic interval is expanded, in the just before giving birth characteristic interval of first after enabling fisrt feature result fall into expansion；
S2017, the first training managing step is repeated, until all the first data snapshot all training managings
It finishes；
Specifically, successively 120 the first data snapshots are input to according to time sequencing and just before giving birth identify prototype network progress
First training managing, be then further continued for execute by the second uterine contraction signal data be input to just before giving birth identify prototype network carry out net
The step for network training S201, until just before giving birth identify model output accuracy rate meet preset condition (such as: until 95%), at this time
Current first just before giving birth characteristic interval is the optimum interval parameter for just before giving birth identifying prototype network.By way of prototype network training,
So that data are gradually intended to consistent characteristic, measure that the result come is just more representative and reasonability.
The preferred embodiment of this system is further used as, the step for the building Time in labor prediction model network
S202 is specifically included:
S2021, building first layer prototype network, the first layer prototype network include: Volume Four lamination, the 4th normalization
Layer and the 5th function actuation layer, the output end of the Volume Four lamination are defeated by the 4th normalization layer and the 5th function actuation layer
Enter end connection；
Specifically, construct first layer prototype network, the layer network include a conv convolutional layer, BN normalization layer and
The function actuation layer that one ReLU activation primitive is constituted, wherein conv layers of convolution kernel is 13*13*256, first layer prototype network
Output valve is high dimensional feature result.
S2022, building second layer prototype network, the second layer prototype network include: the 5th convolutional layer, the 4th normalization
Layer, the 6th function actuation layer, the second random deactivating layer and the 6th convolutional layer, the output end of the 5th convolutional layer successively pass through
Four normalization layers, the 6th function actuation layer, the second random deactivating layer are connect with the input terminal of the 6th convolutional layer, the 5th function
The output end of excitation layer is connect with the input terminal of the 5th convolutional layer；
Specifically, construct second layer prototype network, the layer network include a conv convolutional layer, BN normalization layer,
Function actuation layer, the random deactivating layer of a Dropout and another conv convolutional layer that one ReLU activation primitive is constituted, first
The convolution kernel of a conv is 7*7*256, and Dropout is set as 50%, and second convolution kernel is 3*3*128, increases a branch
This layer network is transformed into a residual block by connection, i.e., from the quick mapping for being input to output and being a max pooling.The
The input value of two layer model networks is the output valve of first layer prototype network, and the output valve of second layer prototype network is one-dimensional data
Characteristic results.
S2023, building third layer prototype network, the third layer prototype network includes by multiple identical second residual blocks
Layer connection composition residual block layer group, wherein each second residual block layer include: the 5th normalization layer, the 7th function actuation layer,
The random deactivating layer of third, the 7th convolutional layer, the 6th normalization layer, the 8th function actuation layer, the 4th random deactivating layer and volume eight
The input terminal of lamination, the 5th normalization layer successively passes through the 7th function actuation layer, the random deactivating layer of third, the 7th convolution
Layer, the 6th normalization layer, the 8th function actuation layer, the 4th random deactivating layer are connect with the output end of the 8th convolutional layer, and described the
The output end of six convolutional layers is connect with the input terminal of residual block layer group；
Specifically, third layer prototype network is constructed, which includes 7 continuously connected identical residual blocks, each
Residual error has one to be fast mapped as max pooling, and residual block internal structure includes that BN normalization layer, a ReLU swash
It is function actuation layer that function living is constituted, the random deactivating layer of a Dropout, a conv convolutional layer, a BN normalization layer, another
Function actuation layer, the random deactivating layer of another Dropout and another conv convolutional layer that a ReLU activation primitive is constituted one by one,
The conv convolution nucleus number of same residual block is constant, and conv convolution nucleus number is every just to reduce half, conv convolution by a residual block
The size of core is 3*3, and the input value of third layer prototype network is the output valve of second layer prototype network, third layer prototype network
Output valve be higher-dimension characteristic results.
S2024, the 4th layer model network of building, the 4th layer model network include: the 7th normalization layer, the 9th function
Excitation layer, the second full articulamentum and the tenth function actuation layer, it is described normalization layer input terminal successively pass through the 7th normalization layer,
9th function actuation layer, the second full articulamentum are connect with the output end of the tenth function actuation layer, the output of the residual block layer group
It holds and is connect with the input terminal of the 7th normalization layer；
Specifically, the 4th layer model network is constructed, which includes that a BN normalizes layer, a ReLU activation primitive
The function actuation layer of composition, the full articulamentum of a FC and a Tanh activation primitive, the input value of the 4th layer model network are the
The output valve of three layer model network, the output valve of the 4th layer model network are the output valve of Time in labor prediction model network.
It is further used as the preferred embodiment of this system, it is described that the third uterine contraction signal data is input to Time in labor
Prediction model network carries out the step for network training S202, specifically includes:
S2025, third uterine contraction signal data is successively carried out at the processing of the second maximum value down-sampling and second order combination
Multiple second data snapshots are obtained after reason；
Specifically, first intercepting third uterine contraction signal data is 10 minutes length, then is carried out to third uterine contraction signal data
The long interception of every five seconds clock does under the second maximum value each data slot to obtain the data slot after 120 interceptions
Value processing is sampled, 2209 numerical value being sequentially arranged are obtained from each data slot, by each data slice
Numerical value in section carries out permutation and combination, the matrix of a 47*47 is successively arranged out in the way of a line by every 47 numerical value, as
First testing data snapshot, to obtain 120 the first data snapshots.
S2026, the second data snapshot and just before giving birth label are input to the second training of Time in labor prediction model network progress
Processing；
It is described that second data snapshot and just before giving birth label are input at the second training of Time in labor prediction model network progress
The step for reason includes: that the second data snapshot and just before giving birth label are input to after Time in labor prediction model network is handled to obtain
To second feature as a result, judging whether second feature result falls into the second just before giving birth characteristic interval, if so, the second just before giving birth characteristic area
Between it is constant, conversely, then expanding the second just before giving birth characteristic interval, second feature result is enabled to fall into the second just before giving birth characteristic area after expanding
Between in；
S2027, the second training managing step is repeated, until all the second data snapshot all training managings
It finishes；
Specifically, 120 the second data snapshots are successively input to by Time in labor prediction model network according to time sequencing
The second training managing is carried out, execution is then further continued for by the third uterine contraction signal data and is input to Time in labor prediction model net
Network carries out the step for network training S202, until the output accuracy rate for just before giving birth identifying model meets preset condition (such as: 95%)
Until, at this time current second just before giving birth characteristic interval be Time in labor prediction model network optimum interval parameter.
It is further used as the preferred embodiment of this system, the method step is further comprising the steps of:
S301, pregnant week data and amniotic fluid status data are obtained；
S302, it thus when judging that current state is just before giving birth situation, is then sent out when according to pregnant week data and amniotic fluid status data
Just before giving birth pre-warning signal out.
Specifically, doing some simplest state identifications to pregnant woman helps to exclude special circumstances, carries out convenient for model quasi-
True prediction.
It is further used as the preferred embodiment of this system, the uterine myoelectricity acquisition processing device passes through blue tooth wireless communication
Mode and intelligent terminal communication connection.By using the communication modes that bluetooth is wireless, data acquisition can be made more convenient.
Be further used as the preferred embodiment of this system, the intelligent terminal include smart phone, IPAD, laptop computer and
Desktop computer.Generality of the invention can be improved as intelligent terminal using portable computer.
It is further used as the preferred embodiment of this system, the uterine myoelectricity acquisition processing device is believed using four emg electrodes
Number acquisition device realizes the acquisition of electromyography signal.Using the available more accurate underground palace of four emg electrode signal pickup assemblies
Contracting signal is conducive to the accuracy for improving prediction.
Suitable for present apparatus embodiment, present apparatus embodiment is implemented content in the above system embodiment
Function is identical as the above system embodiment, and the beneficial effect reached and the above system embodiment beneficial effect achieved
It is identical.
It is to be illustrated to preferable implementation of the invention, but the invention is not limited to the implementation above
Example, those skilled in the art can also make various equivalent variations on the premise of without prejudice to spirit of the invention or replace
It changes, these equivalent deformations or replacement are all included in the scope defined by the claims of the present application.
Claims (10)
1. a kind of intelligence just before giving birth monitors system, which is characterized in that described including uterine myoelectricity acquisition processing device and intelligent terminal
Uterine myoelectricity acquisition processing device and intelligent terminal communication connection；
The uterine myoelectricity acquisition processing device extracts from collected electromyography signal for acquiring electromyography signal and obtains the
One uterine contraction signal data；
The intelligent terminal, comprising:
At least one processor；
At least one processor, for storing at least one program；
When at least one described program is executed by least one described processor, so that at least one described processor realizes one kind
Intelligently just before giving birth prediction technique step, the method step includes:
According to the first uterine contraction signal data, just before giving birth status predication processing and/or Time in labor prediction processing are carried out；
Wherein, the just before giving birth status predication processing comprising:
The first testing data snapshot is obtained after doing the first pretreatment to the first uterine contraction signal data acquired, it is to be measured by first
Data snapshot input just before giving birth identifies prototype network and carries out prediction processing, thus output just before giving birth state；
The Time in labor prediction processing comprising:
The second testing data snapshot is obtained after doing the second pretreatment to the first uterine contraction signal data acquired, it is to be measured by second
Data snapshot input Time in labor prediction model network is predicted, to export prediction Time in labor.
2. a kind of intelligence just before giving birth monitors system according to claim 1, which is characterized in that it is described just before giving birth identify prototype network be
It is obtained by following model training step:
Building just before giving birth identifies prototype network, obtains the second uterine contraction signal data, the second uterine contraction signal data is input to and is faced
It produces and identifies prototype network progress network training；
And/or the Time in labor prediction model network is obtained by following model training step:
Time in labor prediction model network is constructed, third uterine contraction signal data is obtained, the third uterine contraction signal data is inputted
Network training is carried out to Time in labor prediction model network.
3. a kind of intelligence just before giving birth monitors system according to claim 2, which is characterized in that the building just before giving birth identifies model net
The step for network, specifically includes:
First layer prototype network is constructed, the first layer prototype network includes: the first convolutional layer, the first normalization layer and the first letter
The output end of number excitation layer, first convolutional layer is connect by the first normalization layer with the input terminal of first function excitation layer；
Second layer prototype network is constructed, the second layer prototype network includes: the second convolutional layer, the second normalization layer, the second letter
Number excitation layer, the first random deactivating layer and third convolutional layer, the output end of second convolutional layer successively pass through the second normalization
Layer, second function excitation layer, the first random deactivating layer are connect with the input terminal of third convolutional layer, the first function excitation layer
Output end is connect with the input terminal of the second convolutional layer；
Third layer prototype network is constructed, the third layer prototype network includes being connected in series by several memory unit layers
The output end of memory unit layer group, the third convolutional layer is connect with the input terminal of memory unit layer group；
Construct the 4th layer model network, the 4th layer model network includes: third normalization layer, third function actuation layer, the
One full articulamentum and the 4th function actuation layer, the output end of second convolutional layer successively pass through third normalization layer, third letter
Number excitation layers, the first full articulamentum are connect with the input terminal of the 4th function actuation layer, the output end of the memory unit layer group and
Third normalizes the input terminal connection of layer.
4. a kind of intelligence just before giving birth monitors system according to claim 2, which is characterized in that described by the second uterine contraction signal
Data, which are input to, just before giving birth identifies the step for prototype network carries out network training, specifically includes:
To the second uterine contraction signal data successively carry out the first maximum value down-sampling processing and first order combined treatment after obtain it is more
A first data snapshot；
First data snapshot and just before giving birth label are input to and just before giving birth identify prototype network the first training managing of progress；
Described be input to the first data snapshot and just before giving birth label just before giving birth identifies prototype network progress this step of the first training managing
It suddenly include: to be input to the first data snapshot and just before giving birth label just before giving birth to identify after prototype network is handled to obtain fisrt feature knot
Fruit, judges whether fisrt feature result falls into the first just before giving birth characteristic interval, if so, the first just before giving birth characteristic interval is constant, conversely,
Then the first just before giving birth characteristic interval is expanded, in the just before giving birth characteristic interval of first after enabling fisrt feature result fall into expansion；
The first training managing step is repeated, all training managing is complete up to the first all data snapshots.
5. a kind of intelligence just before giving birth monitors system according to claim 2, which is characterized in that the building Time in labor predicts mould
It the step for type network, specifically includes:
First layer prototype network is constructed, the first layer prototype network includes: Volume Four lamination, the 4th normalization layer and the 5th letter
The output end of number excitation layer, the Volume Four lamination is connect by the 4th normalization layer with the input terminal of the 5th function actuation layer；
Second layer prototype network is constructed, the second layer prototype network includes: the 5th convolutional layer, the 4th normalization layer, the 6th letter
Number excitation layer, the second random deactivating layer and the 6th convolutional layer, the output end of the 5th convolutional layer successively pass through the 4th normalization
Layer, the 6th function actuation layer, the second random deactivating layer are connect with the input terminal of the 6th convolutional layer, the 5th function actuation layer
Output end is connect with the input terminal of the 5th convolutional layer；
Third layer prototype network is constructed, the third layer prototype network includes being made of multiple identical second residual block layer connections
Residual block layer group, wherein each second residual block layer includes: that the 5th normalization layer, the 7th function actuation layer, third are lost at random
Living layer, the 7th convolutional layer, the 6th normalization layer, the 8th function actuation layer, the 4th random deactivating layer and the 8th convolutional layer, described the
The input terminal of five normalization layers successively passes through the 7th function actuation layer, the random deactivating layer of third, the 7th convolutional layer, the 6th normalization
Layer, the 8th function actuation layer, the 4th random deactivating layer are connect with the output end of the 8th convolutional layer, the output of the 6th convolutional layer
End is connect with the input terminal of residual block layer group；
Construct the 4th layer model network, the 4th layer model network includes: the 7th normalization layer, the 9th function actuation layer, the
Two full articulamentums and the tenth function actuation layer, the input terminal of the normalization layer successively pass through the 7th normalization layer, the 9th function
Excitation layer, the second full articulamentum are connect with the output end of the tenth function actuation layer, the output end and the 7th of the residual block layer group
Normalize the input terminal connection of layer.
6. a kind of intelligence just before giving birth monitors system according to claim 2, which is characterized in that described by the third uterine contraction signal
Data are input to the step for Time in labor prediction model network carries out network training, specifically include:
To third uterine contraction signal data successively carry out the second maximum value down-sampling processing and second order combined treatment after obtain it is more
A second data snapshot；
Second data snapshot and just before giving birth label are input to Time in labor prediction model network and carry out the second training managing；
It is described by the second data snapshot and just before giving birth label be input to Time in labor prediction model network carry out the second training managing this
One step includes: that the second data snapshot and just before giving birth label are input to after Time in labor prediction model network is handled to obtain
Two characteristic results, judge whether second feature result falls into the second just before giving birth characteristic interval, if so, second just before giving birth characteristic interval is not
Become, conversely, then expanding the second just before giving birth characteristic interval, the just before giving birth characteristic interval of second after enabling second feature result fall into expansion
In；
The second training managing step is repeated, all training managing has finished up to the second all data snapshots.
7. a kind of intelligence just before giving birth monitors system according to claim 1, which is characterized in that the method step further includes following
Step:
Obtain pregnant week data and amniotic fluid status data；
When according to pregnant week data and amniotic fluid status data to judge that current state is just before giving birth situation, then just before giving birth early warning is issued
Signal.
8. a kind of any one of -7 intelligence just before giving birth monitor system according to claim 1, which is characterized in that the uterine myoelectricity is adopted
Collection processing unit passes through blue tooth wireless communication mode and intelligent terminal communication connection.
9. a kind of any one of -7 intelligence just before giving birth monitor system according to claim 1, which is characterized in that the intelligent terminal packet
Include smart phone, IPAD, laptop computer and/or desktop computer.
10. a kind of any one of -7 intelligence just before giving birth monitor system according to claim 1, which is characterized in that the uterine myoelectricity
Acquisition processing device realizes the acquisition of electromyography signal using four emg electrode signal pickup assemblies.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910162216.0A CN109875556B (en) | 2019-03-05 | 2019-03-05 | Intelligent in-production monitoring system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910162216.0A CN109875556B (en) | 2019-03-05 | 2019-03-05 | Intelligent in-production monitoring system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109875556A true CN109875556A (en) | 2019-06-14 |
| CN109875556B CN109875556B (en) | 2021-08-31 |
Family
ID=66930546
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910162216.0A Active CN109875556B (en) | 2019-03-05 | 2019-03-05 | Intelligent in-production monitoring system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109875556B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111667916A (en) * | 2020-06-09 | 2020-09-15 | 复旦大学 | Machine learning-based antenatal uterine contraction judging system |
| CN112613581A (en) * | 2020-12-31 | 2021-04-06 | 广州大学华软软件学院 | Image recognition method, system, computer equipment and storage medium |
| CN113100711A (en) * | 2021-04-08 | 2021-07-13 | 南通市妇幼保健院 | Method and system for real-time tracking of health condition of parturient |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030135130A1 (en) * | 2002-01-15 | 2003-07-17 | General Electric Company | Method and apparatus for uterine contraction monitoring using linear predictive modeling of abdominal surface EMG signals |
| US20120238894A1 (en) * | 2006-12-11 | 2012-09-20 | Principe Jose C | System and Method for Analyzing Progress of Labor and Preterm Labor |
| CN104812314A (en) * | 2012-11-26 | 2015-07-29 | 辉凌公司 | Method and system for diagnosing uterine contraction levels using image analysis |
| CN107307845A (en) * | 2016-04-27 | 2017-11-03 | 广州翠格医疗技术有限公司 | A kind of natural labor probabilistic forecasting devices and methods therefor |
| WO2019016759A1 (en) * | 2017-07-19 | 2019-01-24 | Bloom Technologies NV | Monitoring uterine activity and assessing pre-term birth risk |
-
2019
- 2019-03-05 CN CN201910162216.0A patent/CN109875556B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030135130A1 (en) * | 2002-01-15 | 2003-07-17 | General Electric Company | Method and apparatus for uterine contraction monitoring using linear predictive modeling of abdominal surface EMG signals |
| US20120238894A1 (en) * | 2006-12-11 | 2012-09-20 | Principe Jose C | System and Method for Analyzing Progress of Labor and Preterm Labor |
| CN104812314A (en) * | 2012-11-26 | 2015-07-29 | 辉凌公司 | Method and system for diagnosing uterine contraction levels using image analysis |
| CN107307845A (en) * | 2016-04-27 | 2017-11-03 | 广州翠格医疗技术有限公司 | A kind of natural labor probabilistic forecasting devices and methods therefor |
| WO2019016759A1 (en) * | 2017-07-19 | 2019-01-24 | Bloom Technologies NV | Monitoring uterine activity and assessing pre-term birth risk |
Non-Patent Citations (1)
| Title |
|---|
| BASSAM MOSLEM等: "CLASSIFICATION OF MULTICHANNEL UTERINE EMG SIGNALS BY USING UNSUPERVISED COMPETITIVE LEARNING", 《2011 IEEE WORKSHOP ON SIGNAL PROCESSING SYSTEMS (SIPS)》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111667916A (en) * | 2020-06-09 | 2020-09-15 | 复旦大学 | Machine learning-based antenatal uterine contraction judging system |
| CN112613581A (en) * | 2020-12-31 | 2021-04-06 | 广州大学华软软件学院 | Image recognition method, system, computer equipment and storage medium |
| CN112613581B (en) * | 2020-12-31 | 2023-10-13 | 广州大学华软软件学院 | Image recognition method, system, computer equipment and storage medium |
| CN113100711A (en) * | 2021-04-08 | 2021-07-13 | 南通市妇幼保健院 | Method and system for real-time tracking of health condition of parturient |
| CN113100711B (en) * | 2021-04-08 | 2021-11-16 | 南通市妇幼保健院 | Method and system for real-time tracking of health condition of parturient |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109875556B (en) | 2021-08-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109875556A (en) | A kind of intelligence just before giving birth monitors system | |
| CN112904220B (en) | UPS (uninterrupted Power supply) health prediction method and system based on digital twinning and machine learning, electronic equipment and storable medium | |
| KR102362532B1 (en) | Method and apparatus for predicting state of battery health based on neural network | |
| CN106295799B (en) | A kind of implementation method of deep learning multilayer neural network | |
| CN109545385A (en) | A kind of medical big data analysis processing system and its method based on Internet of Things | |
| Lee et al. | Maximum likelihood estimation of two-level latent variable models with mixed continuous and polytomous data | |
| CN110334802A (en) | A kind of construction method of neural network model, device, equipment and storage medium | |
| CN108847280A (en) | The smart cloud medical treatment real-time management system of case-based reasioning | |
| CN100472555C (en) | Method and system for detecting artifacts in ICU patient records by data fusion and hypothesis testing | |
| CN110208702A (en) | A kind of lithium battery ultrasound information feature extraction method neural network based | |
| Kadrolkar et al. | Variable-word-length coding for energy-aware signal transmission | |
| CN109645996B (en) | Uterine contraction debilitation monitoring method and system, intelligent terminal and storage medium | |
| Zong et al. | Embedded software fault prediction based on back propagation neural network | |
| CN116011682A (en) | Meteorological data prediction method and device, storage medium and electronic device | |
| CN110223780A (en) | A kind of the fetal weight prediction technique and system of adaptive PSO-GRNN | |
| CN109495113A (en) | A kind of compression method and device of EEG signals | |
| Irungu et al. | ML-ECG-COVID: A Machine Learning-Electrocardiogram Signal Processing Technique for COVID-19 Predictive Modeling | |
| CN110890157A (en) | Analytical model establishing method, auxiliary diagnosis device, auxiliary diagnosis equipment and medium | |
| Ahmad et al. | Water quality monitoring using wireless sensor networks | |
| CN111104298A (en) | LSTM-based power grid server running state prediction device | |
| JP7455892B2 (en) | Abnormality prediction system, method and computer program | |
| CN113130070B (en) | Comprehensive diagnosis method and system for craniocerebral diseases | |
| CN213787356U (en) | Health detection device and health management system | |
| CN113505925B (en) | Laboratory hazardous chemical anomaly information early warning method based on ANFIS | |
| Boukhennoufa et al. | An IoT-Based framework for elderly remote monitoring |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |