CN113456064B - Intelligent interpretation method for prenatal fetal heart monitoring signals - Google Patents
Intelligent interpretation method for prenatal fetal heart monitoring signals Download PDFInfo
- Publication number
- CN113456064B CN113456064B CN202110798011.9A CN202110798011A CN113456064B CN 113456064 B CN113456064 B CN 113456064B CN 202110798011 A CN202110798011 A CN 202110798011A CN 113456064 B CN113456064 B CN 113456064B
- Authority
- CN
- China
- Prior art keywords
- signal
- fetal heart
- fetal
- heart rate
- prenatal
- 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.)
- Active
Links
- 210000002458 fetal heart Anatomy 0.000 title claims abstract description 102
- 238000012544 monitoring process Methods 0.000 title claims abstract description 48
- 238000000034 method Methods 0.000 title claims abstract description 38
- 230000001605 fetal effect Effects 0.000 claims abstract description 74
- 208000036029 Uterine contractions during pregnancy Diseases 0.000 claims abstract description 50
- 238000012545 processing Methods 0.000 claims abstract description 24
- 230000011218 segmentation Effects 0.000 claims abstract description 22
- 238000007781 pre-processing Methods 0.000 claims abstract description 15
- 230000002457 bidirectional effect Effects 0.000 claims abstract description 13
- 230000004927 fusion Effects 0.000 claims abstract description 9
- 230000006835 compression Effects 0.000 claims abstract description 5
- 238000007906 compression Methods 0.000 claims abstract description 5
- 230000002159 abnormal effect Effects 0.000 claims description 24
- 239000011159 matrix material Substances 0.000 claims description 17
- 238000012217 deletion Methods 0.000 claims description 9
- 230000037430 deletion Effects 0.000 claims description 9
- 125000004122 cyclic group Chemical group 0.000 claims description 6
- 239000012634 fragment Substances 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 4
- 238000013136 deep learning model Methods 0.000 abstract description 10
- 230000006870 function Effects 0.000 abstract description 3
- 239000000523 sample Substances 0.000 description 25
- 210000003754 fetus Anatomy 0.000 description 11
- 238000012795 verification Methods 0.000 description 11
- 238000010835 comparative analysis Methods 0.000 description 9
- 230000036541 health Effects 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 4
- 230000006378 damage Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 230000002427 irreversible effect Effects 0.000 description 3
- 230000000306 recurrent effect Effects 0.000 description 3
- 208000027418 Wounds and injury Diseases 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000007621 cluster analysis Methods 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 208000014674 injury Diseases 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000000877 morphologic effect Effects 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 206010067484 Adverse reaction Diseases 0.000 description 1
- 238000012935 Averaging Methods 0.000 description 1
- 241000288105 Grus Species 0.000 description 1
- 206010021143 Hypoxia Diseases 0.000 description 1
- 230000006838 adverse reaction Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000012850 discrimination method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000001146 hypoxic effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000008774 maternal effect Effects 0.000 description 1
- 230000015654 memory Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 230000035935 pregnancy Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/1118—Determining activity level
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
- A61B5/024—Detecting, measuring or recording pulse rate or heart rate
- A61B5/02411—Detecting, measuring or recording pulse rate or heart rate of foetuses
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/43—Detecting, measuring or recording for evaluating the reproductive systems
- A61B5/4306—Detecting, measuring or recording for evaluating the reproductive systems for evaluating the female reproductive systems, e.g. gynaecological evaluations
- A61B5/4343—Pregnancy and labour monitoring, e.g. for labour onset detection
- A61B5/4356—Assessing uterine contractions
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/43—Detecting, measuring or recording for evaluating the reproductive systems
- A61B5/4306—Detecting, measuring or recording for evaluating the reproductive systems for evaluating the female reproductive systems, e.g. gynaecological evaluations
- A61B5/4343—Pregnancy and labour monitoring, e.g. for labour onset detection
- A61B5/4362—Assessing foetal parameters
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B2503/00—Evaluating a particular growth phase or type of persons or animals
- A61B2503/02—Foetus
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Animal Behavior & Ethology (AREA)
- Veterinary Medicine (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Public Health (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Physiology (AREA)
- Signal Processing (AREA)
- Psychiatry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Gynecology & Obstetrics (AREA)
- Pregnancy & Childbirth (AREA)
- Reproductive Health (AREA)
- Cardiology (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Fuzzy Systems (AREA)
- Pediatric Medicine (AREA)
- Dentistry (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)
Abstract
The invention discloses an intelligent interpretation method of prenatal fetal heart monitoring signals, which comprises the steps of preprocessing fetal heart rate signals, uterine contraction pressure signals and fetal movement signals to obtain a fusion multi-signal data set; then carrying out segmentation processing to construct a data set to be classified; sequentially inputting an embedded layer, a splicing layer, a bidirectional gating circulating unit layer and a full-connection layer of the prenatal fetal monitoring intelligent interpretation model, and performing sigmoid function compression on the output result to obtain a classification discrimination result; the intelligent interpretation method of the invention integrates the uterine contraction pressure signal and fetal movement signal into the prenatal fetal monitoring intelligent interpretation model based on the bidirectional gating circulation unit, and compared with other deep learning models, the intelligent interpretation method of the invention has the advantages of less time consumption and better classification discrimination capability.
Description
Technical Field
The invention relates to a deep learning method, in particular to an intelligent interpretation method of prenatal fetal heart monitoring signals, which is used for intelligently classifying and discriminating prenatal fetal condition assessment.
Background
Prenatal fetal monitoring is widely used to assess the healthy development of the fetus and to treat it in a timely manner before adverse reactions occur in the fetus. Fetal heart and uterine contractility monitoring (CTG) is an important tool for fetal prenatal health monitoring, recording the changes in fetal heart rate during pregnancy and its time relationship with uterine contractions. The role of which includes, but is not limited to, determining whether the fetus is hypoxic, determining whether delivery of the maternal delivery by caesarean section is required, and guiding other assessment of the health of the fetus.
At this stage, the interpretation of the fetal heart and uterine contractility monitoring Chart (CTG) is performed by the obstetrician. Interpretation is based on clinical experience primarily from prenatal fetal monitoring guidelines and the obstetrician himself. However, the conventional fetal heart and uterine contraction monitoring Chart (CTG) interpretation standard is inconsistent, so that the experience level of obstetrician is also inconsistent, the interpretation process is very subjective, the interpretation result cannot be reproduced, ambiguity is easy to occur, misdiagnosis is caused, and irrecoverable injuries are caused to pregnant women and fetuses.
Therefore, many scholars at home and abroad begin to research the intelligent classification and discrimination of prenatal fetal monitoring by using a machine learning algorithm, and the data used is mostly a fetal monitoring standard CTG data set of a public California university Euro-language calibration machine learning knowledge base (University of California Irvine machine learning repository, UCI), and 21 clinical characteristic data are obtained by extracting features by using fetal heart rate and uterine contraction signals for classification.
However, the automatic interpretation of the prenatal fetal heart uterine contraction graph model based on machine learning at the present stage is mostly modeled by using features, and measurement errors which cannot be eliminated exist when the features are extracted. Meanwhile, after feature extraction, the interpretation results obtained according to different prenatal fetal monitoring interpretation guidelines are also inconsistent. The model using signal modeling only considers the fetal heart rate signal, ignores the relation between the fetal heart rate signal and the uterine contraction pressure signal, does not deeply study the influence of the uterine contraction pressure signal and the fetal movement signal on the health condition of the fetus, and has too simple signal preprocessing flow
In general, the prenatal fetal monitoring machine used at home and abroad at present still cannot reach the intelligent level, the obtained accuracy rate of judging suspicious types is only 45-82%, the accuracy rate of abnormal types is only 66-94%, and the prenatal fetal monitoring machine cannot be applied to clinical work. Therefore, how to further improve the accuracy of interpretation of the prenatal fetal monitoring intelligent model becomes a technical problem to be solved in the art.
Disclosure of Invention
In order to solve the defects in the prior art, the invention provides an intelligent interpretation method for prenatal fetal heart monitoring signals.
The invention adopts the following technical scheme:
an intelligent interpretation method for prenatal fetal heart monitoring signals comprises the following steps:
s1: acquiring original CTG signal data comprising a fetal heart rate signal, a uterine contraction pressure signal and a fetal movement signal;
s2: preprocessing the fetal heart rate signal, the uterine contraction pressure signal and the fetal movement signal to form a fusion multi-signal data set;
s3: the fusion multi-signal data set is subjected to segmentation processing to obtain one-dimensional fetal heart rate signals, uterine contraction pressure signals and fetal movement signal fragments with the signal length d, and a data set to be classified is constructed;
s4: inputting the data set to be classified into a pre-trained intelligent interpretation model for prenatal fetal monitoring, wherein the intelligent interpretation model for prenatal fetal monitoring comprises an embedded layer, a splicing layer, a bidirectional gating circulating unit layer and a full-connection layer;
respectively inputting the one-dimensional fetal heart rate signal, the uterine contraction pressure signal and the fetal movement signal segment with the signal length of d into an embedded layer to obtain a fetal heart rate signal, a uterine contraction pressure signal and a fetal movement signal segment of a two-dimensional matrix (d multiplied by m), wherein m is the output dimension of the embedded layer;
inputting the fetal heart rate signal, the uterine contraction pressure signal and the fetal movement signal fragments of the two-dimensional matrix (d multiplied by m) into a splicing layer for splicing to obtain a vector of the two-dimensional matrix (d multiplied by 3 m);
inputting the vector of the two-dimensional matrix (d multiplied by 3 m) to a two-way gating cyclic unit layer of 2k units for splicing to obtain a one-dimensional output vector with the length of 2 k;
inputting the one-dimensional output vector with the length of 2k to a full-connection layer of n units, performing sigmoid function compression on an output result, and outputting the probability that a sample is of a normal type and a sample is of an abnormal type;
setting corresponding class labels for the probability that the sample is of a normal class and the probability that the sample is of an abnormal class respectively;
and comparing the probability that the sample is of a normal class with the probability that the sample is of an abnormal class, and selecting a class label corresponding to the larger probability as a classification discrimination result.
Preferably, the preprocessing in step S2 includes interpolation or deletion of the fetal heart rate signal, the uterine contraction pressure signal and the fetal movement signal, respectively.
Preferably, the preprocessing in step S2 further includes normalizing the interpolated or deleted fetal heart rate signal.
Preferably, the normalization process includes calculating a baseline value of the fetal heart rate signal and subtracting the baseline value from the interpolated or deleted fetal heart rate signal to obtain a normalized fetal heart rate signal.
Preferably, the preprocessing in step S2 further includes sliding window segmentation of the interpolated or deleted fetal heart rate signal before the normalized processing of the interpolated or deleted fetal heart rate signal, so as to obtain a fetal heart rate signal segment with a signal length not less than p.
Preferably, the step S3 of segmentation processing includes synchronous sliding window segmentation of the preprocessed fetal heart rate signal, uterine contraction pressure signal and fetal movement signal.
The invention has the following beneficial effects:
firstly, the uterine contraction pressure signal and the fetal movement signal are fused into the intelligent interpretation model for prenatal fetal monitoring based on the bidirectional gating circulation unit, and compared with other deep learning models, the intelligent interpretation method provided by the invention is short in time consumption, has better classification discrimination capability, and can be used for rapidly and effectively providing auxiliary decision support for prenatal fetal monitoring for obstetrical medical staff.
Secondly, the invention not only preprocesses the fetal heart rate signals, the uterine contraction pressure signals and the fetal movement signals according to the data characteristics of the fetal heart rate signals, but also carries out sliding window segmentation processing on the preprocessed standard fetal heart rate signals, uterine contraction pressure signals and fetal movement signals, thereby realizing data enhancement, unifying the signal length and further improving the classification and discrimination performance of the intelligent interpretation model.
Thirdly, the prenatal fetal heart monitoring signal intelligent interpretation method provided by the invention uses signal modeling, does not need to extract the morphological characteristics of a fetal heart and uterine contraction monitoring Chart (CTG), and effectively eliminates measurement errors which cannot be eliminated when signals extract clinical morphological characteristics.
Fourth, the intelligent interpretation method of prenatal fetal heart monitoring signals provided by the invention can effectively reduce the workload of medical staff, effectively reduce the death rate and the cesarean productivity of fetuses, simultaneously avoid unnecessary medical intervention, ensure the normal growth and development of fetuses, and is beneficial to improving the birth population quality of China.
Drawings
Fig. 1 is a schematic flow chart of an algorithm of the intelligent interpretation model for prenatal fetal monitoring of the present invention.
Detailed Description
The CTG signal data adopted by the invention are central station data from a plurality of cooperative hospitals, recording instruments and model number are SRF618A Pro of a multi-bed wireless probe fetal monitoring workstation developed by the company, the recording frequency is 1.25Hz, and the collected data has the service life of 2016 to 2018. The three rounds of professional obstetrician interpretation according to the prenatal fetal monitoring guidelines "obstetrics and gynecology science" shows 16,355 cases of prenatal fetal monitoring cases with consistent results, wherein 11,998 cases are normal cases, 4,326 cases are suspicious cases, and 31 cases are abnormal cases. The invention is fetal heart rate uterine contraction signals and extracted fetal movement signals collected for pregnant women with the gestational weeks of 28 weeks to 42 weeks, and the specific demographic characteristics of the pregnant women are shown in table 1:
table 1 demographic characteristics of the central dataset pregnant women of the invention
In order to make the objects, technical solutions and advantageous effects of the present invention more apparent, the present invention will be further described with reference to the accompanying drawings and specific embodiments of the present invention.
Example 1
The invention provides an intelligent interpretation method for prenatal fetal heart monitoring signals, which comprises the following steps: s1: acquiring original CTG signal data comprising a fetal heart rate signal, a uterine contraction pressure signal and a fetal movement signal;
s2: preprocessing the fetal heart rate signal, the uterine contraction pressure signal and the fetal movement signal to form a fusion multi-signal data set;
the preprocessing comprises interpolation or deletion processing of fetal heart rate signals, uterine contraction pressure signals and fetal movement signals respectively;
the preprocessing further comprises the step of carrying out standardization processing on the fetal heart rate signals subjected to interpolation or deletion processing;
the preprocessing further comprises sliding window segmentation of the interpolated or deleted fetal heart rate signal before the standardized processing of the interpolated or deleted fetal heart rate signal, so as to obtain a fetal heart rate signal segment with a signal length of not less than p; wherein p is 750;
the normalization processing comprises calculating a baseline value of the fetal heart rate signal, and subtracting the baseline value from the fetal heart rate signal subjected to interpolation or deletion processing to obtain a standard fetal heart rate signal;
s3: the fusion multi-signal data set is subjected to segmentation processing to obtain one-dimensional fetal heart rate signals, uterine contraction pressure signals and fetal movement signal fragments with the signal length d, and a data set to be classified is constructed; wherein d is 1125;
the segmentation processing comprises synchronous sliding window segmentation of the preprocessed fetal heart rate signal, the uterine contraction pressure signal and the fetal movement signal.
S4, inputting the data set to be classified into a pre-trained intelligent interpretation model for prenatal fetal monitoring, and classifying and judging, wherein the intelligent interpretation model for prenatal fetal monitoring comprises an embedding layer, a splicing layer, a bidirectional gating circulating unit layer and a full-connection layer as shown in the figure 1;
respectively inputting the one-dimensional fetal heart rate signal, the uterine contraction pressure signal and the fetal movement signal segment with the signal length of d into an embedded layer to obtain a fetal heart rate signal, a uterine contraction pressure signal and a fetal movement signal segment of a two-dimensional matrix (d multiplied by m), wherein m is the output dimension of the embedded layer; wherein m is 4;
inputting the fetal heart rate signal, the uterine contraction pressure signal and the fetal movement signal fragments of the two-dimensional matrix (d multiplied by m) into a splicing layer for splicing to obtain a vector of the two-dimensional matrix (d multiplied by 3 m);
inputting the vector of the two-dimensional matrix (d multiplied by 3 m) to a two-way gating cyclic unit layer of 2k units for splicing to obtain a one-dimensional output vector with the length of 2 k; k is the number of GRU units in one direction, and 2k is the number of GRU units in the two-way gating cycle unit layer; wherein k is 512 and 2k is 1024;
inputting the one-dimensional output vector with the length of 2k to a full-connection layer of n units, performing sigmoid function compression on an output result, and outputting the probability that a sample is of a normal type and a sample is of an abnormal type; wherein n is 2;
setting corresponding class labels for the probability that the sample is of a normal class and the probability that the sample is of an abnormal class respectively; the label of the sample as normal class is 0, and the label of the sample as abnormal class is 1.
And comparing the probability that the sample is of a normal class with the probability that the sample is of an abnormal class, and selecting a class label corresponding to the larger probability as a classification discrimination result.
According to the common knowledge in the art, the problems of abnormal values and missing values of the original fetal heart rate signal, the uterine contraction pressure signal and the fetal movement signal can be solved by interpolation or deletion of missing segments. The fetal heart rate signal and the uterine contraction pressure signal have consistency, and interpolation or deletion processing can be synchronously carried out on the uterine contraction pressure signal when the fetal heart rate signal is preprocessed. Because the fetal movement signals are extracted from the uterine contraction pressure signals point by point, the fetal movement signals have consistency, and interpolation or deletion processing can be synchronously carried out on the fetal movement signals. The interpolation of the fetal heart rate signal and the uterine contraction pressure signal is neutral, and the sequence fluctuation after interpolation is stable, so that the interpolation value of fetal movement is 0 when the fetal heart rate signal and the uterine contraction pressure signal are processed.
Since the signal length in the original CTG signal data is changed from 20min (1500 points) to 10min (750 points) to 20min (1500 points) after the preprocessing, the invention discards the signal with the signal length less than 10min (750 points) by sliding window segmentation on the preprocessed fetal heart rate signal.
The invention performs sliding window segmentation by employing the distance between maximum points. Specifically, firstly, the pretreated fetal heart rate signal is subjected to smooth noise reduction through an SG filtering (Savitzky-Golay) digital filter, and the time corresponding to the extreme point of the signal is not changed; then finding out the set of all maximum points in the fetal heart rate signal, wherein the maximum points are judged by the first-order difference of two adjacent points; then taking the signal length of 10min as a section to carry out window sliding treatment, namely taking the first point in the set of maximum value points as an end point, taking the front 10min as the first section, then starting to slide a window, taking the next extreme value point as the end point, taking the front 10min as a new section, and carrying out circulation; and adding the head and the tail of the data after sliding to obtain a sliding segmentation result on the signal.
The definition of Baseline (Baseline) d is defined as the mean of the fetal heart rate with an amplitude stabilized within 5bpm over 10min, according to common knowledge in the art. According to the invention, the base line points are extracted from all extreme points by carrying out cluster analysis on the extreme points of the fetal heart rate curve, and the base line values of the fetal heart rate signals are obtained by averaging the base line points. Specifically, firstly, finding out a set of all extreme points, decentering all the extreme points, and then adopting a K-Means algorithm to perform cluster analysis on the obtained decentered extreme points; marking and distinguishing a base line part and a non-base line part of the obtained result, then taking an average tm of time abscissas of all base line points and taking an average fm of time abscissas of fetal heart rate as base line values (tm, fm) of the segment of fetal heart rate signals respectively through the base line part; calculating all fetal heart rate segmentation baseline values, and then performing baseline fitting; and (3) putting all the baseline values and the fetal heart rate curve in the same coordinate system for data fitting, and directly adopting an interpolation method for data fitting to obtain the baseline value of the fetal heart rate signal.
In addition, the invention can also make the fetal heart rate signal curve closer to the baseline part by using empirical mode decomposition (Empirical Mode Decomposition, EMD) treatment, reduce the number of extreme points, and facilitate distinguishing the baseline part from the non-baseline part by using the extreme points.
It should be noted that, those skilled in the art may use other methods to preprocess the acquired raw CTG signal data, or use other methods to extract the baseline point of the fetal heart rate signal, so as to obtain the baseline value of the fetal heart rate signal.
The intelligent interpretation model for prenatal fetal monitoring comprises an embedded layer, a splicing layer, a bidirectional gate control circulating unit layer (namely a BiGRU layer) and a full-connection layer. The embedded layer performs OneHot processing on each signal point of a one-dimensional fetal heart rate signal, a uterine contraction pressure signal and a fetal movement signal segment with the signal length of d, multiplies a result by a two-dimensional matrix W, transforms the signal into a 4 x d two-dimensional matrix, W is a 4 row and a 4 column two-dimensional matrix, d is the dimension of OneHot, the formula is OneHot (t) x W, t E (1, d), and the spliced layer performs second-by-second splicing on the output result of the embedded layer to form a two-dimensional signal with the result of 12 x d.
A bi-directional gating cyclic unit network of 2k GRU units in a bi-directional gating cyclic unit layer, the calculation formula at this layer is as follows:
update door: z t ＝σ(W (z) xt+U (z) h t-1 )
Reset gate: r is (r) t ＝σ(W (z) xt+U (z) h t-1 )
The currently stored content: h is a 1 ′＝tanh(Wxt+r⊙Uh t-1 )
Final memory (output) of current time step h t ＝z t ⊙h t-1 +(1-z t )⊙h t ′
It should be noted that, in the bidirectional gate control cyclic unit layer, each GRU unit outputs a one-dimensional vector, and finally, the one-dimensional output vector with the length of 2k is combined and output, and then, the one-dimensional output vector is input into the full-connection layer of the two units for calculation, and the calculation formula of the ith unit of the full-connection layer is as follows: f (f) ci ＝W i1 *h 1 +W i2 *h 2 +…W ik *h k +W ik+1 *h 1 ′+W ik+2 *h 2 ′+…W i2 k*h k ′。h k Indicating the result of the forward GRU, h k ' indicates the result of the reverse GRU. Then performing sigmoid compression on the two unit results, and respectively outputting the probability of the sample being normalsigmoid(f c) And a probability sigmoid (f) that the sample is of an abnormal class c2 )。
Verification example 1
In order to verify the discrimination capability of the invention on the prenatal fetal condition, a RNN, LSTM, GRU, biRNN, biLSTM, CNN and DNN deep learning model is selected for comparison analysis, and the discrimination capability comparison analysis results of the deep learning models are shown in table 2.
The gated loop units (Gated Recurrent Unit, GRU) described above are variants of the loop neural network (Recurrent Neural Network, RNN). Long Short-Term Memory (LSTM) is a variant of RNN. The bidirectional gating circulation unit (Bidirectional Gated Recurrent Unit, biGRU) consists of GRUs in the positive and negative directions, and the CTG signals are subjected to contextual characteristic acquisition by using the upper and lower channels in parallel to jointly determine the output at the same moment.
Each deep learning model learns the same training set, then uses the same verification set to verify and output a convergence curve of the model learning process, finally carries out classification and discrimination on the same test set, takes a label corresponding to the maximum value of the output probability as an output result, and divides the label into a normal class and an abnormal class. Verification example 1 comparative analysis the accuracy, precision, recall, specificity, F1 value, kappa coefficient, MCC coefficient, AUC value and time of the above deep learning model.
Accuracy, namely, accuracy (Accuracy), is the most common judgment index in deep learning. Accuracy (Precision) indicates the correct positive class data duty ratio to predict among the data predicted as positive class. The Recall (Recall) indicates the correct positive class data duty to predict from the data that is actually positive class. Specificity (Specificity) means the predicted correct negative class data duty ratio in the data that is actually negative. The positive class related to the accuracy rate, the recall rate and the specificity is the abnormal class of the invention, and the negative class is the normal class of the invention.
The F1 value (F1-score) represents a comprehensive index considering the accuracy and recall, and the F1 value is more representative when there is an imbalance phenomenon in the data. The Kappa coefficient represents an index describing judgment of consistency, and the larger the value thereof is, the higher the consistency level is. The MCC coefficient (Matthews Correlation Coefficient) and Ma Xiusi related coefficient are key indexes for measuring the most abundant information in the quality of the two classifier models.
Further, verification example 1 introduced a receiver operating characteristic curve (Receiver Operating Characteristic Curve, ROC) to evaluate the performance of the deep learning model. To measure the results of the ROC, the Area of the ROC was defined as the AUC value (Area opening Curve) ranging from [0,1]. The larger the AUC value, the better the classification effect of the model is.
TABLE 2 discrimination capability vs. analysis results for each deep learning model
The results in table 2 show that, compared with the general gating circulation unit and the mainstream deep learning algorithm, the embodiment 1 performs intelligent classification and discrimination through the bi-directional gating circulation unit (biglu) deep learning model, which not only consumes less time, but also synthesizes various evaluation indexes, and the embodiment 1 has the optimal discrimination capability, thereby rapidly and effectively providing auxiliary decision support for prenatal fetal monitoring for obstetrical medical staff.
Verification example 2
In order to verify the influence of the difference of the data sets to be classified on the discrimination capability of the intelligent interpretation method. Verification example 2 four control groups were set up, control group a: the data set to be classified contained only standard fetal heart rate signals (FHR); a control group B, wherein the data set to be classified only contains uterine contraction signals (UC); control group C, data set to be classified contains fetal movement signals (FM) only; control group D: the data set to be classified contains only the standard fetal heart rate and uterine contraction combined signal (F-U). The remaining method steps for the four control groups were identical to those of example 1. Comparative analysis example 2 was validated for accuracy, precision, recall, specificity, F1 value, kappa coefficient, MCC coefficient, and AUC value for example 1 and four control groups, and the comparative analysis results are shown in table 3.
Table 3 results of performance comparison analysis of four control groups with example 1
The results in table 3 show that, compared with the four control groups, the data set to be classified in example 1 includes a standard fetal heart rate signal, a uterine contraction pressure signal and a fetal movement signal, and the intelligent interpretation model in example 1 has the optimal classification and discrimination capability by comprehensively comparing various evaluation indexes.
Verification example 3
In order to verify the influence of synchronization of the fusion multi-signal data set on the discrimination capability of the intelligent interpretation method provided by the invention through sliding window segmentation processing. The fused multisignal dataset of control E was not subjected to sliding window segmentation processing, the remaining method steps remained consistent with example 1. Verification example 3 comparative analysis example 1 and control group E were analyzed for accuracy, precision, recall, specificity, F1 value, kappa coefficient, MCC coefficient, and AUC value, and comparative analysis results are shown in table 4.
Table 4 results of comparative analysis of control group B with example 1
The results in table 4 show that, compared with the control group E, the fusion multi-signal data set is processed synchronously by sliding window segmentation, so that the accuracy, recall, F1 value, kappa coefficient, MCC coefficient and AUC value of the embodiment 1 reach higher degree, which means that the classification discrimination performance of the intelligent interpretation model of the embodiment 1 is improved, the probability of misjudging abnormal samples as normal samples can be effectively reduced, and irreversible damage to the health of pregnant women and fetuses caused by missing treatment time is avoided.
Verification example 4
In order to verify the influence of the standardized processing of the fetal heart rate signals on the discrimination capability of the intelligent discrimination method. The original fetal heart rate signal of control group F was pretreated without normalization, and the remaining method steps were consistent with example 1. Verification example 4 comparative analysis example 1 and control group F were analyzed for accuracy, precision, recall, specificity, F1 value, kappa coefficient, MCC coefficient, and AUC value, and comparative analysis results are shown in table 4.
TABLE 5 comparative analysis results of control group A and example 1
The results in table 5 show that, compared with the control group F, the original fetal heart rate signal is subjected to the normalization treatment after being preprocessed, so that the recall rate, the F1 value, the Kappa coefficient, the MCC coefficient and the AUC value of the embodiment 1 reach a higher degree, which means that the classification discrimination performance of the intelligent interpretation model of the embodiment 1 is improved, the probability of misjudging an abnormal sample as a normal sample can be effectively reduced, and irreversible damage to the health of pregnant women and fetuses caused by missing treatment time is avoided.
Verification example 5
The following confusion matrix is adopted to verify the discrimination capability of the prenatal fetal monitoring intelligent interpretation model based on the bidirectional gating circulation unit:
| predicting Yunnan solid | Positve | Negative |
| Postive | TP(TuePositve) | FP(True Negative) |
| Negative | FN False Positve | TN(False Negative) |
TABLE 6 confusion matrix of example 1
| Prediction/realism | Normal state | Abnormal state |
| Normal state | 84.36％ | 13.85％ |
| Abnormal state | 15.64％ | 86.15％ |
The results in table 6 show that the accuracy of the abnormal sample in example 1 is highest and reaches 86.15%, the accuracy of the normal sample reaches 84.36%, and meanwhile, the probability of misjudging the real abnormal sample as the normal sample is 13.85%, so that the possibility of misjudging the normal sample as the abnormal sample is reduced, and unnecessary interference to the fetus caused by excessive production and detection can be avoided.
In summary, the intelligent interpretation method of the invention blends the uterine contraction pressure signal and the fetal movement signal into the intelligent interpretation model of the prenatal fetal monitoring based on the bidirectional gating circulation unit, compared with other deep learning models, the time consumption is shorter, the intelligent interpretation method has better classification discrimination capability, unexpected technical effects are obtained, the auxiliary decision support of the prenatal fetal monitoring can be rapidly and effectively provided for obstetrical medical staff, the probability of misjudging abnormal samples as normal samples is effectively reduced, and the irreversible injury to pregnant women and fetal health caused by missing treatment opportunities is avoided.
Furthermore, the invention not only preprocesses the fetal heart rate signal, the uterine contraction pressure signal and the fetal movement signal according to the data characteristics, but also carries out sliding window segmentation processing on the preprocessed standard fetal heart rate signal, uterine contraction pressure signal and fetal movement signal, thereby realizing data enhancement, unifying the signal length, further improving the classification discrimination performance of the prenatal fetal monitoring intelligent interpretation model based on the bidirectional gating circulation unit and obtaining unexpected technical effects.
The foregoing is merely a preferred embodiment of the present invention, but the present invention is not limited to the embodiment, and those skilled in the art, within the scope of the present disclosure, can apply equivalent modifications or substitutions according to the concept of the present technical solution, and are included in the scope of the present invention.
Claims (6)
1. An intelligent interpretation method for prenatal fetal heart monitoring signals comprises the following steps:
s1: acquiring original CTG signal data comprising a fetal heart rate signal, a uterine contraction pressure signal and a fetal movement signal;
s2: preprocessing the fetal heart rate signal, the uterine contraction pressure signal and the fetal movement signal to form a fusion multi-signal data set;
s3: the fusion multi-signal data set is subjected to segmentation processing to obtain one-dimensional fetal heart rate signals, uterine contraction pressure signals and fetal movement signal fragments with the signal length d, and a data set to be classified is constructed;
s4: inputting the data set to be classified into a pre-trained intelligent interpretation model for prenatal fetal monitoring, wherein the intelligent interpretation model for prenatal fetal monitoring comprises an embedded layer, a splicing layer, a bidirectional gating circulating unit layer and a full-connection layer;
respectively inputting the one-dimensional fetal heart rate signal, the uterine contraction pressure signal and the fetal movement signal segment with the signal length of d into an embedded layer to obtain a fetal heart rate signal, a uterine contraction pressure signal and a fetal movement signal segment of a two-dimensional matrix (d multiplied by m), wherein m is the output dimension of the embedded layer;
inputting the fetal heart rate signal, the uterine contraction pressure signal and the fetal movement signal fragments of the two-dimensional matrix (d multiplied by m) into a splicing layer for splicing to obtain a vector of the two-dimensional matrix (d multiplied by 3 m);
inputting the vector of the two-dimensional matrix (d multiplied by 3 m) to a two-way gating cyclic unit layer of 2k units for splicing to obtain a one-dimensional output vector with the length of 2 k;
inputting the one-dimensional output vector with the length of 2k to a full-connection layer of n units, performing sigmoid function compression on an output result, and outputting the probability that a sample is of a normal type and a sample is of an abnormal type;
setting corresponding class labels for the probability that the sample is of a normal class and the probability that the sample is of an abnormal class respectively;
and comparing the probability that the sample is of a normal class with the probability that the sample is of an abnormal class, and selecting a class label corresponding to the larger probability as a classification discrimination result.
2. The method for intelligently interpreting prenatal fetal heart monitoring signals according to claim 1, wherein the method comprises the following steps: the preprocessing of the step S2 comprises interpolation or deletion processing of the fetal heart rate signal, the uterine contraction pressure signal and the fetal movement signal respectively.
3. The method for intelligently interpreting prenatal fetal heart monitoring signals according to claim 2, wherein the method comprises the following steps: the preprocessing of the step S2 further comprises the step of normalizing the fetal heart rate signal subjected to interpolation or deletion.
4. The method for intelligently interpreting prenatal fetal heart monitoring signals according to claim 3, wherein the method comprises the following steps: the normalization process includes calculating a baseline value of the fetal heart rate signal and subtracting the baseline value from the interpolated or deleted fetal heart rate signal to obtain a standard fetal heart rate signal.
5. The method for intelligently interpreting prenatal fetal heart monitoring signals according to claim 3, wherein the method comprises the following steps: the preprocessing in the step S2 further comprises sliding window segmentation of the interpolated or deleted fetal heart rate signal before the interpolated or deleted fetal heart rate signal is subjected to standardization processing, so that a fetal heart rate signal segment with the signal length not less than p is obtained.
6. The method for intelligently interpreting prenatal fetal heart monitoring signals according to claim 1, wherein the method comprises the following steps: the segmentation processing of the step S3 comprises the step of synchronously sliding window segmentation of the preprocessed fetal heart rate signal, the uterine contraction pressure signal and the fetal movement signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110798011.9A CN113456064B (en) | 2021-07-15 | 2021-07-15 | Intelligent interpretation method for prenatal fetal heart monitoring signals |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110798011.9A CN113456064B (en) | 2021-07-15 | 2021-07-15 | Intelligent interpretation method for prenatal fetal heart monitoring signals |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113456064A CN113456064A (en) | 2021-10-01 |
| CN113456064B true CN113456064B (en) | 2024-04-16 |
Family
ID=77880341
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110798011.9A Active CN113456064B (en) | 2021-07-15 | 2021-07-15 | Intelligent interpretation method for prenatal fetal heart monitoring signals |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113456064B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114159039A (en) * | 2021-11-12 | 2022-03-11 | 广州三瑞医疗器械有限公司 | Intelligent antenatal fetal heart monitoring model |
| CN114724720B (en) * | 2022-06-10 | 2022-09-09 | 北京大学第三医院（北京大学第三临床医学院） | Prenatal electronic fetal heart monitoring automatic identification system based on deep learning |
| CN118196520A (en) * | 2024-03-27 | 2024-06-14 | 世济其美(广州)技术服务有限公司 | Method and system for identifying acceleration and deceleration of fetal heart monitoring chart during birth based on neural network |
| CN118319321B (en) * | 2024-05-08 | 2024-10-08 | 中国人民解放军总医院第八医学中心 | Electrocardiogram data optimization processing method for trauma patient |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1993025971A1 (en) * | 1992-06-09 | 1993-12-23 | University Of Plymouth | Medical signal analyzer |
| CN102210586A (en) * | 2011-06-03 | 2011-10-12 | 泰安市迈迪医疗电子有限公司 | Automatic analysis method for fetus monitoring device |
| CN107997741A (en) * | 2018-01-08 | 2018-05-08 | 北京大学 | A kind of coupling analytical method of Fetal Heart Rate and uterine contraction signal |
| CN108836307A (en) * | 2018-05-14 | 2018-11-20 | 广东工业大学 | A kind of intelligent ECG detection device, equipment and mobile terminal |
| CN109567868A (en) * | 2018-10-15 | 2019-04-05 | 广东宝莱特医用科技股份有限公司 | A kind of CTG fetal rhythm methods of marking and system |
| CN112971799A (en) * | 2021-02-04 | 2021-06-18 | 北京理工大学 | Non-stimulation fetal heart monitoring classification method based on machine learning |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140378855A1 (en) * | 2013-06-25 | 2014-12-25 | The Research Foundation For The State University Of New York | Apparatus and method for feature extraction and classification of fetal heart rate |
-
2021
- 2021-07-15 CN CN202110798011.9A patent/CN113456064B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1993025971A1 (en) * | 1992-06-09 | 1993-12-23 | University Of Plymouth | Medical signal analyzer |
| CN102210586A (en) * | 2011-06-03 | 2011-10-12 | 泰安市迈迪医疗电子有限公司 | Automatic analysis method for fetus monitoring device |
| CN107997741A (en) * | 2018-01-08 | 2018-05-08 | 北京大学 | A kind of coupling analytical method of Fetal Heart Rate and uterine contraction signal |
| CN108836307A (en) * | 2018-05-14 | 2018-11-20 | 广东工业大学 | A kind of intelligent ECG detection device, equipment and mobile terminal |
| CN109567868A (en) * | 2018-10-15 | 2019-04-05 | 广东宝莱特医用科技股份有限公司 | A kind of CTG fetal rhythm methods of marking and system |
| CN112971799A (en) * | 2021-02-04 | 2021-06-18 | 北京理工大学 | Non-stimulation fetal heart monitoring classification method based on machine learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113456064A (en) | 2021-10-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113456064B (en) | Intelligent interpretation method for prenatal fetal heart monitoring signals | |
| Cömert et al. | Fetal hypoxia detection based on deep convolutional neural network with transfer learning approach | |
| CN108389201B (en) | Lung nodule benign and malignant classification method based on 3D convolutional neural network and deep learning | |
| Rahmayanti et al. | Comparison of machine learning algorithms to classify fetal health using cardiotocogram data | |
| Petrozziello et al. | Deep learning for continuous electronic fetal monitoring in labor | |
| Cömert et al. | Cardiotocography signals with artificial neural network and extreme learning machine | |
| US20200111212A1 (en) | Chromosome Abnormality Detecting Model, Detecting System Thereof, And Method For Detecting Chromosome Abnormality | |
| CN113077434A (en) | Method, device and storage medium for lung cancer identification based on multi-modal information | |
| CN118285774A (en) | Intelligent judging method for fetal monitoring during birth based on dual-attention multi-mode fusion | |
| CN115618284A (en) | Multi-center fetal heart monitoring intelligent interpretation method based on unsupervised field self-adaption | |
| CN111696670B (en) | Intelligent interpretation method for prenatal fetal monitoring based on deep forest | |
| Peng et al. | Preliminary study on the efficient electrohysterogram segments for recognizing uterine contractions with convolutional neural networks | |
| Magenes et al. | Comparison of data mining techniques applied to fetal heart rate parameters for the early identification of IUGR fetuses | |
| CN118213072A (en) | Obstetrical and gynecological information analysis method based on obstetrical examination data flow | |
| Zhou et al. | Improvement of accuracy and resilience in fhr classification via double trend accumulation encoding and attention mechanism | |
| Liang et al. | Fully automatic classification of cardiotocographic signals with 1D-CNN and bi-directional GRU | |
| CN117942038A (en) | Sleep stage identification system integrating cross prediction and discrimination tasks | |
| CN114078137A (en) | Colposcope image screening method and device based on deep learning and electronic equipment | |
| Yang et al. | Unsupervised clustering and analysis of contraction-dependent fetal heart rate segments | |
| US6941288B2 (en) | Online learning method in a decision system | |
| CN117352164A (en) | Multi-mode tumor detection and diagnosis platform based on artificial intelligence and processing method thereof | |
| CN114159039A (en) | Intelligent antenatal fetal heart monitoring model | |
| CN116228731A (en) | Multi-contrast learning coronary artery high-risk plaque detection method, system and terminal | |
| Mendis et al. | The Effect of Fetal Heart Rate Segment Selection on Deep Learning Models for Fetal Compromise Detection | |
| de Vries et al. | Contrastive predictive coding for anomaly detection of fetal health from the cardiotocogram |
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 |