A method and system for data quality identification and correction of a fetal heart rate signal

Abstract

The application discloses a data quality identification correction method and system for a fetal heart rate signal, wherein the data quality identification correction method for the fetal heart rate signal comprises the following steps: acquiring a fetal heart rate signal, wherein the fetal heart rate signal comprises fetal heart rate data; dividing the fetal heart rate data by using a sliding window to obtain a plurality of fetal heart rate segments; classifying and marking the plurality of fetal heart rate segments, and marking a fetal heart rate segment that does not meet a preset data quality condition as a low-quality segment; acquiring a starting position of the low-quality segment; calculating a filling segment of the low-quality segment according to the starting position, replacing the low-quality segment by using the filling segment, and obtaining corrected fetal heart rate data. The technical scheme of the application can solve the problems in the prior art that it is difficult to identify low-quality data in fetal heart rate data in time and correct the low-quality data, signal false alarms are easily generated, and inappropriate obstetric intervention is caused.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for identifying and correcting the data quality of fetal heart rate signals. Background Technology

[0002] In the field of data processing, fetal health data is an important indicator for monitoring fetal health; assessing fetal health during pregnancy and delivery is a crucial approach to ensuring healthy fetal development and a fundamental requirement for maternal care. Measuring fetal heart rate is one of the most commonly used methods for assessing fetal health; by monitoring fetal heart rate, we can assess fetal movement, identify whether the fetus is hypoxic, and reduce the incidence of complications such as fetal acidosis.

[0003] Although fetal heart rate monitoring technology is widely used, it still has significant limitations, primarily in the quality of the acquired signals. Typically, fetal heart rate signals are recorded using skin electrodes or probes placed on the pregnant woman's abdomen, and the recording time is usually around half an hour. Because the movement of the pregnant woman or fetus can affect the contact stability of the skin electrodes or probes, the fetal heart rate signal is easily affected by interference and noise contamination, leading to a significant decrease in signal quality and the presence of low-quality segments. These low-quality segments reduce the accuracy of fetal heart rate measurements, potentially generating numerous false positives, causing incorrect alarms about fetal health or hypoxia, exaggerating the risk of fetal acidosis, and erroneously contributing to a sharp increase in cesarean section rates.

[0004] Therefore, how to identify and correct these low-quality data in a timely manner is an important requirement for reducing false alarms of fetal heart rate signals and reducing inappropriate obstetric interventions, and it is also a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] This invention provides a method and system for identifying and correcting the data quality of fetal heart rate signals, aiming to solve the problem in the prior art that it is difficult to identify and correct low-quality data in fetal heart rate data in a timely manner, which easily leads to false alarms and inappropriate obstetric interventions.

[0006] To achieve the above objectives, according to a first aspect of the present invention, a method for identifying and correcting the data quality of fetal heart rate signals is proposed, comprising:

[0007] Acquire fetal heart rate signals, which include fetal heart rate data;

[0008] The fetal heart rate data was segmented using a sliding window to obtain multiple fetal heart rate segments.

[0009] Based on preset data quality conditions, multiple fetal heart rate segments are classified into two categories, and fetal heart rate segments that do not meet the preset data quality conditions are identified as low-quality segments.

[0010] Find the starting position of the low-quality segment;

[0011] The low-quality segment is calculated based on the starting position, and the low-quality segment is replaced with the filler segment to obtain the corrected fetal heart rate data.

[0012] Preferably, in the above data quality identification and correction method, the step of using a sliding window to segment fetal heart rate data to obtain multiple fetal heart rate segments includes:

[0013] Select a sliding window of a predetermined length;

[0014] Fetal heart rate data is sequentially segmented using a sliding window according to a predetermined step size.

[0015] Obtain multiple segmented fetal heart rate fragments.

[0016] Preferably, in the above data quality identification and correction method, the step of binary classification of multiple fetal heart rate segments includes:

[0017] Multiple fetal heart rate segments are validated using preset data quality conditions, and each fetal heart rate segment is determined to meet the preset data quality conditions.

[0018] If the fetal heart rate segment meets the preset data quality conditions, then the fetal heart rate segment is marked as a high-quality segment; or,

[0019] If the fetal heart rate segment does not meet the preset data quality conditions, the fetal heart rate segment is marked as a low-quality segment.

[0020] Preferably, in the above data quality identification and correction method, the step of obtaining the starting position of the low-quality segment includes:

[0021] Acquire every sampled heart rate from the fetal heart rate data;

[0022] For any one of the sampled heart rates, obtain all fetal heart rate segments containing that one sampled heart rate;

[0023] Calculate the proportion of low-quality segments in the fetal heart rate data;

[0024] Determine whether the proportion of low-quality segments in the fetal heart rate segment is greater than or equal to a preset proportion threshold;

[0025] If the proportion of low-quality segments is greater than or equal to a preset proportion threshold, then the starting position of the fetal heart rate segment is taken as the starting position of the low-quality segments.

[0026] Preferably, in the above data quality identification and correction method, the step of calculating the imputation fragment for low-quality fragments based on the starting position includes:

[0027] Use linear interpolation to calculate linear interpolated segments for low-quality segments;

[0028] A forward time series prediction model is constructed, and the model is used to predict forward time series fragments of low-quality segments; and...

[0029] A backward time series prediction model is constructed, and the backward time series prediction fragments of low-quality segments are predicted using the backward time series prediction model.

[0030] The incomplete segment is obtained by averaging the linear interpolation segment, the forward time series prediction segment, and the backward time series prediction segment.

[0031] Preferably, in the above data quality identification and correction method, the step of constructing a forward time series prediction model and using the forward time series prediction model to predict the forward time series prediction fragment of the low-quality fragment includes:

[0032] Construct a forward time series prediction model and train it using a predetermined number of forward sampled heart rates from low-quality segments;

[0033] When the forward time series prediction model has finished training, input the previous sampled heart rate of the low-quality segment.

[0034] The forward time series prediction model is used to iteratively predict the forward time series segment based on the previous sampled heart rate, thus obtaining all heart rate data for the forward time series prediction segment.

[0035] According to a second aspect of the present invention, the present invention also provides a data quality identification and correction system for fetal heart rate signals, comprising:

[0036] The signal acquisition module is used to acquire the fetal heart rate signal, which includes fetal heart rate data.

[0037] The data segmentation module is used to segment fetal heart rate data using a sliding window to obtain multiple fetal heart rate segments.

[0038] The classification and labeling module is used to perform binary classification on multiple fetal heart rate segments according to preset data quality conditions, and to label fetal heart rate segments that do not meet the preset data quality conditions as low-quality segments.

[0039] The location acquisition module is used to obtain the starting position of low-quality segments;

[0040] The fragment replacement module is used to calculate the filler fragment for the low-quality fragment based on the starting position, and replace the low-quality fragment with the filler fragment to obtain the corrected fetal heart rate data.

[0041] Preferably, in the above-mentioned data quality identification and correction system, the data segmentation module includes:

[0042] The window selection submodule is used to select a sliding window of a predetermined length;

[0043] The sliding segmentation module is used to sequentially slide and segment fetal heart rate data using a sliding window according to a predetermined step size, and obtain multiple segmented fetal heart rate segments.

[0044] Preferably, in the above data quality identification and correction system, the classification labeling module includes:

[0045] The condition judgment submodule is used to verify multiple fetal heart rate segments using preset data quality conditions, and to determine whether each fetal heart rate segment meets the preset data quality conditions.

[0046] The first segment labeling submodule is used to label the fetal heart rate segment as a high-quality segment if it meets preset data quality conditions; or...

[0047] The second segment marking submodule is used to mark the fetal heart rate segment as a low-quality segment if the fetal heart rate segment does not meet the preset data quality conditions.

[0048] Preferably, in the above data quality identification and correction system, the location acquisition module includes:

[0049] The heart rate acquisition submodule is used to acquire all sampled heart rates from the fetal heart rate data;

[0050] The proportion calculation submodule, for any sampled heart rate among all the sampled heart rates, obtains all fetal heart rate segments containing that any sampled heart rate, and calculates the proportion of low-quality segments among the fetal heart rate segments;

[0051] The threshold determination submodule is used to determine whether the proportion of low-quality segments in the fetal heart rate segment is greater than or equal to a preset proportion threshold.

[0052] The starting position acquisition submodule is used to take the starting position of the fetal heart rate segment as the starting position of the low-quality segment if the proportion of low-quality segments is greater than or equal to a preset proportion threshold.

[0053] Preferably, in the above data quality identification and correction system, the fragment replacement module includes:

[0054] The interpolation fragment calculation submodule is used to calculate linear interpolation fragments of low-quality fragments using linear interpolation.

[0055] The prediction model construction submodule is used to construct a forward time series prediction model, which is used to predict forward time series fragments of low-quality segments; and to construct a backward time series prediction model, which is used to predict backward time series fragments of low-quality segments.

[0056] The segment averaging calculation submodule is used to calculate the average of linear interpolation segments, forward time series prediction segments, and backward time series prediction segments to obtain the filled segments.

[0057] In summary, the aforementioned fetal heart rate signal data quality identification and correction scheme acquires fetal heart rate data and segments it using a sliding window to obtain multiple fetal heart rate segments. Each segment contains fetal heart rate data within a specific time period. These segments are then binary-classified according to preset data quality conditions. Segments meeting these conditions are classified as high-quality segments, while those not meeting them are marked as low-quality segments. The starting position of each low-quality segment is then determined, and a replacement heart rate is calculated from this position to fill in the sampled heart rate. This results in a filler segment containing all replacement heart rates. Replacing the low-quality segment with this filler segment corrects the low-quality fetal heart rate data, yielding corrected, high-quality fetal heart rate data. In conclusion, this method addresses the difficulty of timely identification and correction of low-quality fetal heart rate data in existing technologies, which can lead to false alarms and inappropriate obstetric interventions. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0059] Figure 1 This is a flowchart illustrating a method for identifying and correcting the data quality of fetal heart rate signals according to an embodiment of the present invention.

[0060] Figure 2 yes Figure 1 The illustrated embodiment provides a flowchart of a method for segmenting fetal heart rate data;

[0061] Figure 3 yes Figure 1 The illustrated embodiment provides a flowchart of a method for classifying and labeling fetal heart rate segments;

[0062] Figure 4 yes Figure 3 The illustrated embodiment provides a flowchart of a method for determining fetal heart rate segments;

[0063] Figure 5 yes Figure 1 The illustrated embodiment provides a flowchart of a method for obtaining the starting position of a low-quality segment;

[0064] Figure 6 yes Figure 1 The illustrated embodiment provides a flowchart of a method for calculating a missing segment;

[0065] Figure 7 yes Figure 6 The illustrated embodiment provides a flowchart of a method for predicting forward time series segments.

[0066] Figure 8 This is a schematic diagram of the structure of a data quality identification and correction system for fetal heart rate signals provided in an embodiment of the present invention;

[0067] Figure 9 yes Figure 8 The illustrated embodiment provides a schematic diagram of the structure of a data segmentation module;

[0068] Figure 10 yes Figure 8 The illustrated embodiment provides a structural diagram of a classification tagging module;

[0069] Figure 11 yes Figure 8 The illustrated embodiment provides a structural diagram of a location acquisition module;

[0070] Figure 12 yes Figure 8 The illustrated embodiment provides a structural schematic diagram of a fragment filling module;

[0071] Figure 13-a This is a schematic diagram of raw fetal heart rate data provided in an embodiment of the present invention;

[0072] Figure 13-b This is a schematic diagram of a low-quality segment provided in an embodiment of the present invention;

[0073] Figure 13-c This is a schematic diagram of corrected fetal heart rate data provided in an embodiment of the present invention.

[0074] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0075] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0076] The main technical problem solved by the embodiments of the present invention is:

[0077] Fetal heart rate signals are easily affected by surrounding signal interference and noise contamination, leading to a significant decrease in signal quality and the presence of low-quality segments. These low-quality segments reduce the accuracy of fetal heart rate signal measurements, potentially generating numerous false positives, causing incorrect alarms about fetal health or hypoxia, exaggerating the risk of fetal acidosis, and erroneously contributing to a sharp increase in cesarean section rates.

[0078] To address the aforementioned issues, this invention provides a data quality identification and correction scheme for fetal heart rate signals. The scheme involves segmenting fetal heart rate data into multiple segments, classifying these segments according to preset data quality conditions, identifying and labeling low-quality segments. This allows for the timely identification of all low-quality segments within the fetal heart rate signal. Then, by obtaining the starting position of each low-quality segment, a replacement segment is calculated and used to replace the original low-quality segment, resulting in corrected fetal heart rate data. This improves the quality of the fetal heart rate data, reduces false alarms, and consequently minimizes inappropriate obstetric interventions.

[0079] This invention enables the use of a neural network model to execute the following fetal heart rate signal data quality identification and correction scheme. The input of the neural network model is the original fetal heart rate data, and the output is the starting position of the low-quality segment and the corrected fetal heart rate data. The above data quality identification and correction scheme mainly includes two key steps: (1) identifying the low-quality segment, and (2) correcting the low-quality segment.

[0080] To achieve the above objectives, please refer to Figure 1 , Figure 1 This is a flowchart illustrating the first method for identifying and correcting the data quality of fetal heart rate signals provided in this embodiment of the invention. Figure 1 As shown, the data quality identification and correction method for the fetal heart rate signal includes the following steps:

[0081] S110: Acquire fetal heart rate signal, wherein the fetal heart rate signal includes fetal heart rate data. In this embodiment, a neural network model can be used. The above-mentioned fetal heart rate data is input, and the neural network model outputs the starting position of the low-quality segment and the corrected fetal heart rate data. The fetal heart rate signal acquired in this embodiment is fetal heart rate data over a period of time. Specifically, when inputting into the neural network model, this fetal heart rate data can be serialized to form a fetal heart rate sequence. Specifically, let the input original fetal heart rate data be X∈R. n R n This represents a sequence of real numbers of length n.

[0082] S120: Use a sliding window to segment fetal heart rate data to obtain multiple fetal heart rate segments; the neural network model includes a sliding window, which is used to segment the serialized fetal heart rate data to obtain multiple fetal heart rate segments.

[0083] As a preferred embodiment, such as Figure 2 As shown, in the above data quality identification and correction method, step S120: using a sliding window to segment the fetal heart rate data to obtain multiple fetal heart rate segments includes:

[0084] S121: Select a sliding window of a predetermined window length.

[0085] S122: The fetal heart rate data is sequentially segmented using a sliding window according to a predetermined step size.

[0086] S123: Obtain multiple segmented fetal heart rate fragments.

[0087] Specifically, a sliding window is used to segment the original fetal heart rate data. A window length of d and a step size of s are selected to obtain k fetal heart rate segments of length d. The i-th segment among these k segments of length d is denoted as X. i ∈R d .

[0088] After segmenting the fetal heart rate data to obtain multiple fetal heart rate segments, Figure 1 The data quality identification and correction methods shown also include:

[0089] S130: Based on preset data quality conditions, perform binary classification on multiple fetal heart rate segments. Fetal heart rate segments that do not meet the preset data quality conditions are identified and marked as low-quality segments, thus completing the evaluation of low-quality segments. The original fetal heart rate data includes, for example: Figure 13-a As shown; the identified low-quality fragments are as follows Figure 13-b As shown.

[0090] In this embodiment, a neural network model can be used to perform convolution operations on the above-mentioned multiple fetal heart rate segments, and then the convolved fetal heart rate segments are input into the fully connected layer MLP of the neural network model. The above-mentioned fetal heart rate segments are then classified and labeled in two ways according to the above-mentioned preset data quality conditions. Specifically, fetal heart rate segments that meet the preset data quality conditions are high-quality segments, and fetal heart rate segments that do not meet the above-mentioned preset data quality conditions are low-quality segments.

[0091] Specifically, as a preferred embodiment, such as Figure 3 As shown, the step of performing binary classification on multiple fetal heart rate segments according to preset data quality conditions, and identifying and marking fetal heart rate segments that do not meet the preset data quality conditions as low-quality segments includes:

[0092] S131: Use preset data quality conditions to verify multiple fetal heart rate segments, and determine whether each fetal heart rate segment meets the preset data quality conditions;

[0093] S132: If the fetal heart rate segment meets the preset data quality conditions, then mark the fetal heart rate segment as a high-quality segment; or,

[0094] S133: If the fetal heart rate segment does not meet the preset data quality conditions, then mark the fetal heart rate segment as a low-quality segment.

[0095] Multiple preset data quality conditions can be set. In this embodiment, for each fetal heart rate segment X of length d... i Perform signal quality binary classification. If X i If the following three preset data quality conditions are met simultaneously, then X is recorded as X. i A high-quality fragment is marked as 1; otherwise, if any condition is not met, it is marked as X. i This is a low-quality fragment (marked as 0).

[0096] In this embodiment of the application, the preset data quality conditions are set as follows:

[0097] (1) Whether the average heart rate is between 80 beats / minute and 200 beats / minute, i.e.: 80≤mean(X_i)≤200;

[0098] (2) Whether the minimum heart rate is greater than 60 beats / minute, i.e.: min(X_i)>60;

[0099] (3) Whether the maximum heart rate / minimum heart rate is less than 1.5, i.e.: max(X_i) / min(X_i)<1.5.

[0100] In the overall fetal population, generally, (1) the possible range of heart rate is 80 to 200; (2) the instantaneous heart rate should not be lower than 60 beats / minute; (3) the change in heart rate within the window length d should not be too drastic. Therefore, the maximum heart rate is less than 1.5 times the minimum heart rate.

[0101] The specific process for determining whether each fetal heart rate segment meets the preset data quality conditions is as follows: Figure 4 As shown:

[0102] S1331: Input: A fetal heart rate segment of length d.

[0103] S1332: Determine if the average heart rate is between 80 beats / minute and 200 beats / minute; if yes, proceed to step S1333; if no, output: low quality.

[0104] S1333: Determine if the minimum heart rate is greater than 60 beats / minute; if yes, proceed to step S1334; otherwise, output: low quality.

[0105] S1334: Determine if the ratio of maximum heart rate to minimum heart rate is less than 1.5; if yes, output: high quality; if no, output: low quality.

[0106] like Figure 13-b As shown, the embodiments of this application identify the following: Figure 13-b The fetal heart rate data in the shaded area is a low-quality segment.

[0107] After binary classification of multiple fetal heart rate segments, K classification results can be obtained, where K is a real number. This embodiment of the application needs to integrate these K classification results to obtain the starting position of the low-quality segment; therefore... Figure 1 The data quality identification and correction methods shown include:

[0108] S140: Obtain the starting position of the low-quality segment. Specifically, for the original fetal heart rate data X∈R n For each sampling point X[i], calculate the proportion P[i] of all fetal heart rate segments containing X[i] that are marked as low-quality segments, obtaining P∈[0,1]. n Given a preset proportion threshold thresh, fetal heart rate segments with a proportion P of low-quality segments greater than the threshold thresh are identified. The starting positions corresponding to these fetal heart rate segments are then the starting positions of the low-quality segments. In this embodiment of the invention, the threshold thresh is set to 2 / 3.

[0109] As a preferred embodiment, such as Figure 5 As shown, the steps for obtaining the starting position of the low-quality segment include:

[0110] S141: Obtain all sampled heart rates from the fetal heart rate data. Each sampled heart rate is the aforementioned sample point X[i], and the original fetal heart rate data is X∈R. n .

[0111] S142: For any sampled heart rate among all the above sampled heart rates, obtain all fetal heart rate segments containing that sampled heart rate, and calculate the proportion of low-quality segments in the fetal heart rate segments; the proportion or ratio of the low-quality segments is P[i], where P∈[0,1]. n .

[0112] S143: Determine whether the proportion of low-quality segments in the fetal heart rate segment is greater than or equal to a preset proportion threshold;

[0113] S144: If the proportion of low-quality segments is greater than or equal to the preset proportion threshold, then the starting position of the fetal heart rate segment is taken as the starting position of the low-quality segments.

[0114] Given a preset proportion threshold thresh, we find fetal heart rate segments where the proportion P of low-quality segments is greater than the threshold thresh. The starting positions of these fetal heart rate segments are then the starting positions of the low-quality segments. The preset proportion threshold thresh can be set to 2 / 3. This method can accurately and comprehensively find low-quality segments and pinpoint their starting positions, facilitating correction of the low-quality segments from these starting positions.

[0115] In this embodiment, the starting position of the low-quality fragment output is T = {(s_1,t_1),(s_2,t_2),...}, where the tuple (s_i,t_i) represents that the starting position of the i-th low-quality fragment is s_i and the ending position is t_i, and T is a set containing the tuple (s_i,t_i), the number of tuples is not fixed.

[0116] After obtaining the starting position of the low-quality segment, Figure 1 The data quality identification and correction method shown also includes the following steps:

[0117] S150: Calculate a filler segment for the low-quality segment based on the starting position, and replace the low-quality segment with the filler segment to obtain corrected fetal heart rate data. This filler segment is obtained by combining the sampled heart rate from the non-low-quality segment of the aforementioned fetal heart rate data with the predicted heart rate corresponding to the low-quality segment. Therefore, this filler segment can replace the aforementioned low-quality segment, resulting in higher-quality fetal heart rate data. The corrected fetal heart rate data is as follows: Figure 13-c As shown.

[0118] In this embodiment, the filler segment for the low-quality segment is calculated by averaging the linear interpolation segment, the forward time series prediction segment, and the backward time series prediction segment of the low-quality segment. Let the low-quality segment be X. j Low-quality fragment X j linear interpolation segment X j 0 Forward time series prediction segment X j 1 Backward time series prediction segment X j 2 Note that the length of each segment is related to X. j same.

[0119] As a preferred embodiment, such as Figure 6 As shown, in the above data quality identification and correction method, step S150: calculating the filling fragment for the low-quality fragment based on the starting position includes:

[0120] S151: Using linear interpolation, calculate the linear interpolated fragment for the low-quality segment. For the linear interpolated fragment X... j 0 The calculation is performed using linear interpolation with left endpoint value X[s-1], right endpoint value x[t+1], and length ts. Specifically, X... j 0 The i-th value X j 0 [i] is x[t+1]+i / (ts)(x[t+1]-X[s-1]).

[0121] S152: Construct a forward time series prediction model and use it to predict the forward time series fragments of low-quality segments. For the forward time series prediction fragment X... j 1 The forward time series prediction model F is constructed to calculate the forward time series prediction segment for low-quality segments. The model F takes the value at time i-1 as input and predicts the value at time i. F can be any type of time series prediction model, such as linear regression model, AR (autoregressive model), MA (moving average model), ARIMA (differential autoregressive moving average model), deep neural network, etc., without specific restrictions.

[0122] as well as,

[0123] S153: Construct a backward time series prediction model and use the backward time series prediction model to predict the backward time series fragments of low-quality fragments.

[0124] Similar to the forward time series forecasting method, for the backward time series forecast segment X... j2 , it is calculated by constructing a backward time series prediction model G. G takes the value at time i as input and predicts the value at time i - 1. G can be any time series prediction model, such as a linear regression model, an AR (autoregressive model), an MA (moving average model), an ARIMA (autoregressive integrated moving average model), a deep neural network, etc., and no specific limitation is made here.

[0125] S154: Calculate the average of the linear interpolation segment, the forward time series prediction segment, and the backward time series prediction segment to obtain the filling segment. Specifically, average the three parts of the linear interpolation segment, the forward time series prediction segment, and the backward time series prediction segment to obtain the filling segment X j * =(X j 0 +X j 1 +X j 2 ) / 3.

[0126] Use this filling segment X j * to replace the above low-quality segment X j .

[0127] In summary, for the technical solution provided in the embodiment of the present application, without loss of generality for each low-quality segment, here the jth low-quality segment is denoted as X j ={X[s], X[s + 1],..., X[t]}, s < t; calculate the linear interpolation segment X j of the low-quality segment X j 0 , the forward time series prediction segment X j 1 , and the backward time series prediction segment X j 2 , note that the length of each segment is the same as that of X j ;

[0128] For the linear interpolation segment X j 0 , use the linear interpolation method with the left endpoint value X[s - 1], the right endpoint value x[t + 1], and the length t - s for calculation. Specifically, the ith value X j 0 of X j 0 [i] is x[t + 1]+i / (t - s)(x[t + 1]-X[s - 1]).

[0129] For the forward time series prediction segment X j 1The forward time series prediction model F is constructed to calculate the forward time series prediction segment for low-quality segments. The model F takes the value at time i-1 as input and predicts the value at time i. F can be any type of time series prediction model, such as linear regression model, AR (autoregressive model), MA (moving average model), ARIMA (differential autoregressive moving average model), deep neural network, etc., without specific restrictions.

[0130] Specifically, such as Figure 7 As shown, step S152 above: constructing a forward time series prediction model and using the forward time series prediction model to predict the forward time series prediction fragments of low-quality fragments specifically includes:

[0131] S1521: Construct a forward time series prediction model, and train the forward time series prediction model using a predetermined number of sampled heart rates from low-quality segments;

[0132] S1522: When the forward time series prediction model has finished training, input the previous sampled heart rate of the low-quality segment;

[0133] S1523: Use a forward time series prediction model to iteratively predict the forward time series segment based on the previous sampled heart rate, and obtain all heart rate data of the forward time series prediction segment.

[0134] During the training phase, the training is performed using the K data points preceding the sampling point X[s], i.e., {X[sK], x[s-K+1], ..., x[s-1]}. K is a hyperparameter that can be selected according to the actual situation.

[0135] In the prediction phase, X[s-1] is first input and X is predicted. 1 [s], then enter X 1 [s]Predicts X 1 [s+1], repeat this step until X is predicted. 1 [t]. Then the forward time series prediction segment X j 1 ={X 1 [s],X 1 [s+1],...,X 1 [t]}.

[0136] Similarly, the method for predicting low-quality segments using a backward time series prediction model is as described above:

[0137] During the training phase, the training is performed using the K data points following X[t], i.e., {X[t+1], x[t+2], ..., x[t+K]}. K is a hyperparameter that can be selected according to the actual situation.

[0138] In the prediction phase, X[t] is first input to predict X. 2 [t-1], then input X 2 [t-1] predicts X 2 [t-2], repeat this step until X is predicted. 2 [s]. Then X j 2 ={X 2 [s],X 2 [s+1],...,X 2 [t]}.

[0139] In this embodiment, the prediction results from the linear interpolation method, the forward time series prediction model, and the backward time series prediction model are averaged to obtain the filled fragment: X j * =(X j 0 +X j 1 +X j 2 ) / 3. .

[0140] Use fill fragment X j * Replace low-quality fragment X j .

[0141] In summary, the fetal heart rate signal data quality identification and correction method provided in the above embodiments of this application acquires fetal heart rate data and segments it using a sliding window to obtain multiple fetal heart rate segments. Each fetal heart rate segment contains the fetal heart rate within a certain time period. Then, the fetal heart rate segments are binary classified according to preset data quality conditions. Fetal heart rate segments that meet the preset data quality conditions are high-quality segments, while those that do not are marked as low-quality segments. At this point, the starting position of the low-quality segment is obtained, and the replacement heart rate for the sampled heart rate of the low-quality segment is calculated from the starting position. This yields a replacement segment containing all the replacement heart rates for the low-quality heart rate. Replacing the low-quality segment with this replacement segment corrects the low-quality fetal heart rate data, resulting in corrected high-quality fetal heart rate data. In summary, this method solves the problem that existing technologies struggle to identify and correct low-quality fetal heart rate data in a timely manner, and that low-quality fetal heart rate data easily leads to false alarms and inappropriate obstetric interventions.

[0142] Based on the same concept of the above method embodiments, the present invention also proposes a data quality identification and correction system for fetal heart rate signals to implement the above method of the present invention. Since the principle and method of solving the problem in this system embodiment are similar, it has at least all the beneficial effects brought about by the technical solutions of the above embodiments, and will not be described in detail here.

[0143] Specifically, such as Figure 8 As shown, Figure 8 This is a schematic diagram of the structure of a first type of fetal heart rate signal data quality identification and correction system provided in an embodiment of this application. Figure 8 As shown, the data quality identification and correction system for the fetal heart rate signal includes:

[0144] The signal acquisition module 110 is used to acquire the fetal heart rate signal, wherein the fetal heart rate signal includes fetal heart rate data;

[0145] The data segmentation module 120 is used to segment fetal heart rate data using a sliding window to obtain multiple fetal heart rate segments;

[0146] The classification and labeling module 130 is used to perform binary classification and identification of multiple fetal heart rate segments according to preset data quality conditions, and to identify and label fetal heart rate segments that do not meet the preset data quality conditions as low-quality segments.

[0147] The position acquisition module 140 is used to acquire the starting position of the low-quality segment;

[0148] The fragment replacement module 150 is used to calculate the filling fragment of the low-quality fragment based on the starting position, and replace the low-quality fragment with the filling fragment to obtain the corrected fetal heart rate data.

[0149] In summary, the fetal heart rate signal data quality identification and correction system provided in the above embodiments of this application acquires fetal heart rate data through the signal acquisition module 110 and segments it using a sliding window through the data segmentation module 120 to obtain multiple fetal heart rate segments. Each fetal heart rate segment contains the fetal heart rate within a certain time period. Then, the classification and labeling module 130 performs binary classification on the above fetal heart rate segments according to preset data quality conditions. Fetal heart rate segments that meet the above preset data quality conditions are high-quality segments, while fetal heart rate segments that do not meet the above preset data quality conditions are marked as low-quality segments. At this time, the starting position of the low-quality segment is obtained through the position acquisition module 140, and the segment replacement module 150 calculates the filler replacement heart rate of the sampled heart rate of the low-quality segment from the starting position to obtain the filler segment containing all the above filler replacement heart rates. Using the filler segment to replace the above low-quality segment can correct the low-quality fetal heart rate data and obtain corrected high-quality fetal heart rate data. In summary, the above methods can solve the problem that existing technologies cannot promptly identify and correct low-quality fetal heart rate data, which can easily lead to false alarms and inappropriate obstetric interventions.

[0150] As a preferred embodiment, such as Figure 9 As shown, the data segmentation module 120 in the above data quality identification and correction system includes:

[0151] The window selection submodule 121 is used to select a sliding window of a predetermined window length;

[0152] The sliding segmentation module 122 is used to sequentially slide and segment fetal heart rate data according to a predetermined step size using a sliding window to obtain multiple segmented fetal heart rate segments.

[0153] As a preferred embodiment, such as Figure 10 As shown, in the above data quality identification and correction system, the classification labeling module 130 includes:

[0154] The condition judgment submodule 131 is used to verify multiple fetal heart rate segments using preset data quality conditions, and to determine whether each fetal heart rate segment meets the preset data quality conditions.

[0155] The first segment labeling submodule 132 is used to label the fetal heart rate segment as a high-quality segment if the fetal heart rate segment meets preset data quality conditions; or...

[0156] The second segment marking submodule 133 is used to mark the fetal heart rate segment as a low-quality segment if the fetal heart rate segment does not meet the preset data quality conditions.

[0157] As a preferred embodiment, such as Figure 11As shown, in the above data quality identification and correction system, the location acquisition module 140 includes:

[0158] Heart rate acquisition submodule 141 is used to acquire each sampled heart rate in the fetal heart rate data;

[0159] The proportion calculation submodule 142, for any sampled heart rate among all the above sampled heart rates, obtains all fetal heart rate segments containing that any sampled heart rate, and calculates the proportion of low-quality segments in the fetal heart rate segments;

[0160] The threshold judgment submodule 143 is used to determine whether the proportion of low-quality segments in the fetal heart rate segment is greater than or equal to a preset proportion threshold.

[0161] The starting position acquisition submodule 144 is used to take the starting position of the fetal heart rate segment as the starting position of the low-quality segment if the proportion of low-quality segments is greater than or equal to a preset proportion threshold.

[0162] As a preferred embodiment, such as Figure 12 As shown, in the above data quality identification and correction system, the fragment replacement module 150 includes:

[0163] Interpolation fragment calculation submodule 151 is used to calculate linear interpolation fragments of low-quality fragments using linear interpolation.

[0164] The prediction model construction submodule 152 is used to construct a forward time series prediction model, and use the forward time series prediction model to predict forward time series predicted segments of low-quality segments; and to construct a backward time series prediction model, and use the backward time series prediction model to predict backward time series predicted segments of low-quality segments.

[0165] The segment averaging calculation submodule 153 is used to calculate the average of linear interpolation segments, forward time series prediction segments, and backward time series prediction segments to obtain the filled segments.

[0166] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0167] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0168] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0169] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0170] It should be noted that any reference signs placed between parentheses in the claims should not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0171] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0172] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for identifying and correcting the data quality of fetal heart rate signals, characterized in that, include: Acquire fetal heart rate signal, wherein the fetal heart rate signal includes fetal heart rate data; The fetal heart rate data was segmented using a sliding window to obtain multiple fetal heart rate segments; The multiple fetal heart rate segments are classified into two categories according to preset data quality conditions, and fetal heart rate segments that do not meet the preset data quality conditions are identified as low-quality segments. Obtain the starting position of the low-quality segment; Calculate the filler segment of the low-quality segment based on the starting position, and replace the low-quality segment with the filler segment to obtain the corrected fetal heart rate data; The step of segmenting the fetal heart rate data using a sliding window to obtain multiple fetal heart rate segments includes: Select a sliding window of a predetermined length; The fetal heart rate data is sequentially segmented using the sliding window according to a predetermined step size. Obtain multiple segmented fetal heart rate data; The step of performing binary classification and identification of multiple fetal heart rate segments according to preset data quality conditions includes: The preset data quality conditions are used to verify the multiple fetal heart rate segments, and it is determined whether each fetal heart rate segment meets the preset data quality conditions. If the fetal heart rate segment meets the preset data quality conditions, then the fetal heart rate segment is marked as a high-quality segment; or, If the fetal heart rate segment does not meet the preset data quality conditions, then the fetal heart rate segment is marked as a low-quality segment; The step of obtaining the starting position of the low-quality fragment includes: Obtain all sampled heart rates from the fetal heart rate data; For any sampled heart rate among all the sampled heart rates, obtain all fetal heart rate segments containing that any sampled heart rate, and calculate the proportion of low-quality segments in the fetal heart rate segments; Determine whether the proportion of low-quality segments in the fetal heart rate segment is greater than or equal to a preset proportion threshold; If the proportion of the low-quality segments is greater than or equal to the preset proportion threshold, then the starting position of the fetal heart rate segment is taken as the starting position of the low-quality segments.

2. The data quality identification and correction method according to claim 1, characterized in that, The step of calculating the filler fragment for the low-quality fragment based on the starting position includes: Using linear interpolation, calculate the linear interpolated segment of the low-quality segment; Construct a forward time series prediction model, and use the forward time series prediction model to predict the forward time series fragment of the low-quality fragment; and... A backward time series prediction model is constructed, and the backward time series prediction fragment of the low-quality fragment is predicted using the backward time series prediction model. The infill segment is obtained by averaging the linear interpolation segment, the forward time series prediction segment, and the backward time series prediction segment.

3. The data quality identification and correction method according to claim 2, characterized in that, The step of constructing a forward time series prediction model and using the forward time series prediction model to predict the forward time series prediction fragment of the low-quality fragment includes: Construct the forward time series prediction model, and train the forward time series prediction model using a predetermined number of sampled heart rates from the low-quality segments; When the forward time series prediction model has been trained, input the previous sampled heart rate of the low-quality segment; The forward time series prediction model is used to iteratively predict the forward time series segment based on the previous sampled heart rate, thereby obtaining all heart rate data for the forward time series prediction segment.

4. A data quality identification and correction system for fetal heart rate signals, characterized in that, include: A signal acquisition module is used to acquire fetal heart rate signals, wherein the fetal heart rate signals include fetal heart rate data; The data segmentation module is used to segment the fetal heart rate data using a sliding window to obtain multiple fetal heart rate segments. The classification and labeling module is used to perform binary classification and identification of the multiple fetal heart rate segments according to preset data quality conditions, and to identify fetal heart rate segments that do not meet the preset data quality conditions as low-quality segments. A location acquisition module is used to acquire the starting position of the low-quality segment; The fragment replacement module is used to calculate the filling fragment of the low-quality fragment based on the starting position, and use the filling fragment to replace the low-quality fragment to obtain the corrected fetal heart rate data; The classification labeling module includes: The condition judgment submodule is used to verify the multiple fetal heart rate segments using the preset data quality conditions, and to determine whether each of the fetal heart rate segments meets the preset data quality conditions. The first segment labeling submodule is used to label the fetal heart rate segment as a high-quality segment if the fetal heart rate segment meets the preset data quality conditions; or... The second segment marking submodule is used to mark the fetal heart rate segment as a low-quality segment if the fetal heart rate segment does not meet the preset data quality conditions. The location acquisition module includes: The heart rate acquisition submodule is used to acquire all sampled heart rates in the fetal heart rate data. The proportion calculation submodule, for any sampled heart rate of all the sampled heart rates, obtains all fetal heart rate segments containing that any sampled heart rate, and calculates the proportion of low-quality segments in the fetal heart rate segments; The threshold determination submodule is used to determine whether the proportion of low-quality segments in the fetal heart rate segment is greater than or equal to a preset proportion threshold. The starting position acquisition submodule is used to take the starting position of the fetal heart rate segment as the starting position of the low-quality segment if the proportion of the low-quality segment is greater than or equal to the preset proportion threshold.

5. The data quality identification and correction system according to claim 4, characterized in that, The fragment replacement module includes: The interpolation fragment calculation submodule is used to calculate the linear interpolation fragment of the low-quality fragment using the linear interpolation method; The prediction model construction submodule is used to construct a forward time series prediction model, and use the forward time series prediction model to predict the forward time series prediction segment of the low-quality segment; and to construct a backward time series prediction model, and use the backward time series prediction model to predict the backward time series prediction segment of the low-quality segment. The segment averaging calculation submodule is used to calculate the average of the linear interpolation segment, the forward time series prediction segment, and the backward time series prediction segment to obtain the filled segment.

Images