Ultrasound image-based labor progress angle measurement method, system, device, and medium

Abstract

The application discloses a kind of based on ultrasound image labor progress angle measurement method, system, equipment and medium, the method comprises: obtaining ultrasound image, the ultrasound image is the standard section image of preprocessed perinatal mother-fetus ultrasound;Ultrasound image is input into trained segmentation convolution network, and the segmentation result of pubic region and the segmentation result of fetal head region are obtained;Pubic region segmentation result and fetal head region segmentation result are fitted, and then the labor progress angle is measured to obtain.The present application realizes the full automation measurement of labor progress angle by adopting artificial intelligence technology, and then can replace ultrasound physician to complete labor progress angle measurement, so as to solve the technical problems that the workload of existing traditional manual measurement is large and time-consuming.

Description

TECHNICAL FIELD

[0001] The present application relates to an ultrasound image-based labor progress angle measurement method, system, device and medium, belonging to the field of translational medicine and artificial intelligence. BACKGROUND

[0002] Labor progress angle is an important indicator for evaluating the progress of delivery, and is closely related to the relative position of the pubic symphysis and fetal head during delivery. Labor progress angle can assist doctors in judging the position of the fetal head and evaluating the progress of labor, reducing unnecessary cesarean section. In clinical practice, doctors generally use vaginal examination to evaluate the position of the fetal head, which has the problems of strong invasiveness, high error rate and poor objectivity. With the widespread application of ultrasound technology, studies have shown that using ultrasound technology to assist in measuring labor progress angle is relatively effective. The specific process is to first obtain a perineal ultrasound image through an ultrasound device, and then manually label the labor progress angle on the image by the doctor. This method has the advantages of being intuitive, accurate and repeatable. However, manual labeling of labor progress angle has high operation requirements for the user, and there are problems of long labeling time and subjective differences between users. SUMMARY

[0003] Therefore, the present application provides an ultrasound image-based labor progress angle measurement method, system, computer device and storage medium, which solves the technical problems of existing artificial screening-based methods that have large workload and long time-consuming, and different ultrasound doctors obtaining different quality control results due to high dependence on the technology and professional level of ultrasound doctors, thereby affecting the accuracy of ultrasound examination results during delivery.

[0004] The first object of the present application is to provide an ultrasound image-based labor progress angle measurement method.

[0005] The second object of the present application is to provide an ultrasound image-based labor progress angle measurement system.

[0006] The third object of the present application is to provide a computer device.

[0007] The fourth object of the present application is to provide a storage medium.

[0008] The first object of the present application can be achieved by adopting the following technical solutions:

[0009] An ultrasound image-based labor progress angle measurement method, the method comprising:

[0010] obtaining an ultrasound image, the ultrasound image being a pre-processed standard section image of maternal-fetal ultrasound during delivery;

[0011] Input the ultrasound image into the trained segmentation convolutional network to obtain a pubic region segmentation result and a fetal head region segmentation result;

[0012] Fitting the pubic region segmentation result and the fetal head region segmentation result to measure the progress angle of labor.

[0013] Further, the segmentation convolutional network comprises an encoding structure and a decoding structure, the decoding structure comprises an upper branch structure and a lower branch structure, the upper branch structure comprises a plurality of layers of upper branches connected in sequence, and the lower branch structure comprises a plurality of layers of lower branches connected in sequence, the plurality of layers of upper branches and the plurality of layers of lower branches correspond one by one in sequence.

[0014] Among them, at least two layers of upper branches are variable convolutional layers, which are used to eliminate the influence of segmentation target deformation; the lower branch is constrained by the attention gate to the corresponding layer of the upper branch.

[0015] Further, the encoding structure comprises a first convolutional layer, a second convolutional layer, a first down-sampling layer, a third convolutional layer, a fourth convolutional layer, a second down-sampling layer, a fifth convolutional layer, a sixth convolutional layer, a third down-sampling layer, a seventh convolutional layer, an eighth convolutional layer, a fourth down-sampling layer, a ninth convolutional layer, and a tenth convolutional layer connected in sequence.

[0016] Further, the decoding structure further comprises a first up-sampling layer, the tenth convolutional layer, the first up-sampling layer, and the upper branch structure are connected in sequence, and the tenth convolutional layer and the lower branch structure are connected in sequence.

[0017] Further, the upper branch structure comprises a eleventh convolutional layer, a thirteenth convolutional layer, a fifteenth convolutional layer, a second up-sampling layer, a first variable convolutional layer, a nineteenth convolutional layer, a twenty-first convolutional layer, a third up-sampling layer, and a second variable convolutional layer connected in sequence.

[0018] Further, the lower branch structure comprises a twelfth convolutional layer, a fourteenth convolutional layer, a sixteenth convolutional layer, a seventeenth convolutional layer, an eighteenth convolutional layer, a twentieth convolutional layer, a twenty-second convolutional layer, a twenty-third convolutional layer, and a first fully connected layer connected in sequence.

[0019] Further, the decoding structure further comprises a twenty-fourth convolutional layer, a twenty-fifth convolutional layer, a fourth up-sampling layer, a third variable convolutional layer, a twenty-sixth convolutional layer, a twenty-seventh convolutional layer, and a second fully connected layer connected in sequence; and the upper branch structure and the twenty-fourth convolutional layer are connected in sequence.

[0020] Further, the fitting of the pubic region segmentation result and the fetal head region segmentation result to measure the progress angle of labor specifically comprises:

[0021] The pubic region segmentation result and the fetal head region segmentation result are subjected to ellipse fitting to obtain a pubic region ellipse and a fetal head region ellipse;

[0022] The coordinates of the left and right end points of the long axis of the pubic region ellipse are determined;

[0023] The coordinates of the left and right tangent points on the fetal head region ellipse are searched for with the right end point of the long axis of the pubic region ellipse as a support point;

[0024] The labor progress angle is calculated according to the determined coordinates of the left and right end points and the searched coordinates of the right tangent point.

[0025] The second object of the present application can be achieved by adopting the following technical solution:

[0026] An ultrasound image-based labor progress angle measurement system, the system comprising:

[0027] An acquisition unit configured to acquire an ultrasound image, the ultrasound image being a preprocessed standard section image of a mother-fetus during labor;

[0028] A segmentation unit configured to input the ultrasound image into a trained segmentation convolutional network to obtain a pubic region segmentation result and a fetal head region segmentation result;

[0029] A fitting and measurement unit configured to fit the pubic region segmentation result and the fetal head region segmentation result, and further measure a labor progress angle.

[0030] The third object of the present application can be achieved by adopting the following technical solution:

[0031] A computer device comprising a processor and a memory for storing a program executable by the processor, the processor implementing the ultrasound image-based labor progress angle measurement method described above when executing the program stored in the memory.

[0032] The fourth object of the present application can be achieved by adopting the following technical solution:

[0033] A storage medium storing a program, the program being executable by a processor to implement the ultrasound image-based labor progress angle measurement method described above.

[0034] The present application has the following beneficial effects over the prior art:

[0035] 1. The present application realizes full-automatic measurement of the labor progress angle by adopting artificial intelligence technology, and can thus replace ultrasound physicians to complete labor progress angle measurement, thereby solving the technical problems of large workload and long time consumption in the prior art manual measurement;

[0036] 2. The segmentation convolutional network constructed in the application can solve the technical problem of the existing method based on artificial measurement, which is highly dependent on the technology and professional level of doctors, and different doctors have different levels, resulting in different quality control results of different doctors, thereby affecting the accuracy of the measurement results of the progress of labor angle.

[0037] 3. The segmentation convolutional network constructed in the application can extract high-dimensional features reflecting the segmentation structure and micro-texture features of the segmentation region, so as to accurately segment the pubic bone and fetal head, and further make the measurement of the progress of labor angle more accurate. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the drawings shown.

[0039] Figure 1 A brief flowchart of the labor progress angle measurement method based on ultrasound images of the embodiment 1 of the application.

[0040] Figure 2 A detailed flowchart of the labor progress angle measurement method based on ultrasound images of the embodiment 1 of the application.

[0041] Figure 3 An example diagram of the standard section image of the mother-fetal ultrasound in the parturition of the embodiment 1 of the application.

[0042] Figure 4 An example diagram of the segmentation result of the standard section image of the mother-fetal ultrasound in the parturition of the embodiment 1 of the application.

[0043] Figure 5 An example diagram of the segmentation result of the standard section image of the mother-fetal ultrasound in the parturition of the embodiment 1 of the application.

[0044] Figure 6 An example diagram of the measurement of the progress of labor angle of the embodiment 1 of the application.

[0045] Figure 7 A structural block diagram of the labor progress angle measurement system based on ultrasound images of the embodiment 2 of the application.

[0046] Figure 8 A structural block diagram of the computer device of the embodiment 3 of the application. DETAILED DESCRIPTION

[0047] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0048] Embodiment 1

[0049] As shown in Figure 1 and Figure 2 , the present embodiment provides an ultrasound image-based labor progress angle measurement method, which comprises the following steps:

[0050] S101, acquiring an ultrasound image.

[0051] In this step, the ultrasound image is a pre-processed standard section image of maternal-fetal ultrasound in labor.

[0052] Before this step, the step of obtaining the ultrasound image is as follows:

[0053] S11, acquiring a standard section image of maternal-fetal ultrasound in labor.

[0054] As shown in Figure 3 , in the process of delivery, the standard section image of maternal-fetal ultrasound in labor is obtained by transperineal ultrasound examination, and the standard section image of maternal-fetal ultrasound in labor has pubic structure and fetal head structure.

[0055] S12, pre-processing the standard section image of maternal-fetal ultrasound in labor obtained in step S11 to obtain a pre-processed standard section image of maternal-fetal ultrasound in labor.

[0056] Specifically, the standard section image of maternal-fetal ultrasound in labor is adjusted to a resolution of 512x384, thereby completing the preprocessing.

[0057] S102, inputting the ultrasound image into a trained segmentation convolutional network to obtain pubic region segmentation results and fetal head region segmentation results.

[0058] As shown in Figure 4 , the pubic region segmentation results and the fetal head region segmentation results of the present embodiment.

[0059] In the present embodiment, the step of obtaining the trained segmentation convolutional network is as follows:

[0060] S21, acquiring a data set and dividing the data set into a training set and a validation set in a ratio of 8:2, wherein the data set is a standard section image data set of maternal-fetal ultrasound in labor.

[0061] S22, send the data set to an ultrasonographer for manual data labeling to obtain manual data labeling result data, and use the manual data labeling result data as the labeled data set, wherein the labeled data set includes a labeled training set and a labeled validation set, and each perinatal maternal-fetal ultrasound standard section image in the labeled training set corresponds to a manual data labeling result, and the validation set is the same.

[0062] S23, perform data augmentation on the labeled training set to obtain a data-augmented training set.

[0063] S24, sequentially perform standardization processing and normalization processing on the data-augmented training set to obtain a preprocessed training set.

[0064] In step S24, the standardization processing and the normalization processing are referred to as preprocessing.

[0065] S25, input the perinatal maternal-fetal ultrasound standard section image in the preprocessed training set into the segmentation convolutional network to obtain a corresponding inference output result, and input the inference output result and the perinatal maternal-fetal ultrasound standard section image after manual data labeling into a loss function of the segmentation convolutional network to obtain a loss value.

[0066] In step S25, the loss function is a set similarity measure function.

[0067] S26, use an adaptive moment estimation optimizer and the loss value to optimize the loss function of the segmentation convolutional network to iteratively update parameters in the segmentation convolutional network.

[0068] Steps S25 and S26 are repeatedly executed until the segmentation convolutional network converges to an optimal state, thereby obtaining a trained segmentation convolutional network.

[0069] In this embodiment, the specific structure of the segmentation convolutional network is as follows:

[0070] The first layer is an input layer, and the input size is 512x384x3 images.

[0071] The second layer is a first convolutional layer, which receives the image of the input layer; in this layer, the convolution kernel size is 3x3, the step is 1, the number of convolution kernels is 32, the SAME padding method is used, group normalization is performed, a leaky linear rectifier function is used as an activation function, and the final output size is a 512x384x32 matrix.

[0072] The third layer is a second convolutional layer, which receives a matrix with an output size of 512x384x32 from the second layer; in this layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 32, the SAME padding mode is used, group normalization is performed, a leaky linear rectifier function is used as the activation function, and finally a matrix with an output size of 512x384x32 is output.

[0073] The fourth layer is a first down-sampling layer, which receives a matrix with an output size of 512x384x32 from the third layer; in this layer, the pooling mode is maximum pooling, the down-sampling factor is 2x2, the step length is 2, and finally a matrix with an output size of 256x192x32 is output.

[0074] The fifth layer is a third convolutional layer, which receives a matrix with an output size of 256x192x32 from the fourth layer; in this layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 64, the SAME padding mode is used, group normalization is performed, a leaky linear rectifier function is used as the activation function, and finally a matrix with an output size of 256x192x64 is output.

[0075] The sixth layer is a fourth convolutional layer, which receives a matrix with an output size of 256x192x64 from the fifth layer; in this layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 64, the SAME padding mode is used, group normalization is performed, a leaky linear rectifier function is used as the activation function, and finally a matrix with an output size of 256x192x64 is output.

[0076] The seventh layer is a second down-sampling layer, which receives a matrix with an output size of 256x192x64 from the sixth layer; in this layer, the pooling mode is maximum pooling, the down-sampling factor is 2x2, the step length is 2, and finally a matrix with an output size of 128x96x64 is output.

[0077] The eighth layer is a fifth convolutional layer, which receives a matrix with an output size of 128x96x64 from the seventh layer; in this layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 128, the SAME padding mode is used, group normalization is performed, a leaky linear rectifier function is used as the activation function, and finally a matrix with an output size of 128x96x128 is output.

[0078] The ninth layer is a sixth convolutional layer, which receives a matrix with an output size of 128x96x128 from the eighth layer; in this layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 128, the SAME padding mode is used, group normalization is performed, a leaky linear rectifier function is used as the activation function, and finally a matrix with an output size of 128x96x128 is output.

[0079] The tenth layer is a third down-sampling layer, which receives a matrix with a size of 128x96x128 output by the ninth layer; in the present layer, the pooling manner is maximum pooling, the down-sampling factor is 2x2, the step length is 2, and finally a matrix with a size of 64x48x128 is output.

[0080] The eleventh layer is a seventh convolutional layer, which receives a matrix with a size of 64x48x128 output by the tenth layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 256, the SAME manner is used for padding, group normalization is performed, a leaky linear rectifier function is used as an activation function, and finally a matrix with a size of 64x48x256 is output.

[0081] The twelfth layer is an eighth convolutional layer, which receives a matrix with a size of 64x48x256 output by the eleventh layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 256, the SAME manner is used for padding, group normalization is performed, a leaky linear rectifier function is used as an activation function, and finally a matrix with a size of 64x48x256 is output.

[0082] The thirteenth layer is a fourth down-sampling layer, which receives a matrix with a size of 64x48x256 output by the twelfth layer; in the present layer, the pooling manner is maximum pooling, the down-sampling factor is 2x2, the step length is 2, and finally a matrix with a size of 32x24x256 is output.

[0083] The fourteenth layer is a ninth convolutional layer, which receives a matrix with a size of 32x24x256 output by the thirteenth layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 512, the SAME manner is used for padding, group normalization is performed, a leaky linear rectifier function is used as an activation function, and finally a matrix with a size of 32x24x512 is output.

[0084] The fifteenth layer is a tenth convolutional layer, which receives a matrix with a size of 32x24x512 output by the fourteenth layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 512, the SAME manner is used for padding, group normalization is performed, a leaky linear rectifier function is used as an activation function, and finally a matrix with a size of 32x24x512 is output.

[0085] The sixteenth layer is a first up-sampling layer, which receives a matrix with a size of 32x24x512 output by the fifteenth layer; in the present layer, the pooling manner is maximum pooling, the up-sampling factor is 2x2, the step length is 2, and finally a matrix with a size of 64x48x512 is output.

[0086] The upper branch of the seventeenth layer is a eleventh convolutional layer, which receives a matrix with an output size of 64x48x512 from the sixteenth layer; in this layer, the convolution kernel size is 3x3, the step is 1, the number of convolution kernels is 256, the SAME padding is used, the group normalization is performed, the leaky rectified linear function is used as the activation function, and finally a matrix with an output size of 64x48x256 is output.

[0087] The lower branch of the seventeenth layer is a twelfth convolutional layer, which receives a matrix with an output size of 32x24x512 from the fifteenth layer; in this layer, the convolution kernel size is 3x3, the step is 1, the number of convolution kernels is 256, the SAME padding is used, the group normalization is performed, the leaky rectified linear function is used as the activation function, and finally a matrix with an output size of 32x24x256 is output.

[0088] The upper branch of the eighteenth layer is a thirteenth convolutional layer, which receives a matrix with an output size of 64x48x256 from the upper branch of the seventeenth layer; in this layer, the convolution kernel size is 3x3, the step is 1, the number of convolution kernels is 128, the SAME padding is used, the group normalization is performed, the leaky rectified linear function is used as the activation function, and finally a matrix with an output size of 64x48x128 is output.

[0089] The lower branch of the eighteenth layer is a fourteenth convolutional layer, which receives a matrix with an output size of 32x24x256 from the lower branch of the seventeenth layer; in this layer, the convolution kernel size is 3x3, the step is 1, the number of convolution kernels is 256, the SAME padding is used, the group normalization is performed, the leaky rectified linear function is used as the activation function, and finally a matrix with an output size of 32x24x256 is output.

[0090] The upper branch of the nineteenth layer is a fifteenth convolutional layer, which receives a matrix with an output size of 64x48x128 from the upper branch of the eighteenth layer; in this layer, the convolution kernel size is 3x3, the step is 1, the number of convolution kernels is 128, the SAME padding is used, the group normalization is performed, the leaky rectified linear function is used as the activation function, and finally a matrix with an output size of 64x48x128 is output.

[0091] The lower branch of the nineteenth layer is a sixteenth convolutional layer, which receives a matrix with an output size of 32x24x256 from the lower branch of the eighteenth layer; in this layer, the convolution kernel size is 3x3, the step is 1, the number of convolution kernels is 128, the SAME padding is used, the group normalization is performed, the leaky rectified linear function is used as the activation function, and finally a matrix with an output size of 32x24x128 is output.

[0092] The upper branch of the twentieth layer is a second up-sampling layer, which receives a matrix with a size of 64x48x128 output by the upper branch of the nineteenth layer; in the present layer, the pooling manner is maximum pooling, the up-sampling factor is 2x2, the step length is 2, and finally a matrix with a size of 128x96x128 is output.

[0093] The lower branch of the twentieth layer is a seventeenth convolutional layer, which receives a matrix with a size of 32x24x128 output by the lower branch of the nineteenth layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 128, the SAME manner is used for padding, group normalization is performed, a leaky linear rectifier function is used as an activation function, and finally a matrix with a size of 32x24x128 is output.

[0094] The upper branch of the twenty-first layer is a first variability convolutional layer, which receives a matrix with a size of 128x96x128 output by the upper branch of the twentieth layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 128, the SAME manner is used for padding, group normalization is performed, a leaky linear rectifier function is used as an activation function, and finally a matrix with a size of 128x96x128 is output.

[0095] The lower branch of the twenty-first layer is an eighteenth convolutional layer, which receives a matrix with a size of 32x24x128 output by the lower branch of the twentieth layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 64, the SAME manner is used for padding, group normalization is performed, a leaky linear rectifier function is used as an activation function, and finally a matrix with a size of 32x24x64 is output.

[0096] The upper branch of the twenty-second layer is a nineteenth convolutional layer, which receives a matrix with a size of 128x96x128 output by the upper branch of the twenty-first layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 64, the SAME manner is used for padding, group normalization is performed, a leaky linear rectifier function is used as an activation function, and finally a matrix with a size of 128x96x64 is output.

[0097] The lower branch of the twenty-second layer is a twentieth convolutional layer, which receives a matrix with a size of 32x24x64 output by the lower branch of the twenty-first layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 64, the SAME manner is used for padding, group normalization is performed, a leaky linear rectifier function is used as an activation function, and finally a matrix with a size of 32x24x64 is output.

[0098] The upper branch of the twenty-third layer is a twenty-first convolutional layer, which receives a matrix with a size of 128x96x64 output by the upper branch of the twenty-second layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 64, the SAME padding mode is adopted, group normalization is performed, a leaky rectified linear unit function is used as the activation function, and finally a matrix with a size of 128x96x64 is output.

[0099] The lower branch of the twenty-third layer is a twenty-second convolutional layer, which receives a matrix with a size of 32x24x64 output by the lower branch of the twenty-second layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 32, the SAME padding mode is adopted, group normalization is performed, a leaky rectified linear unit function is used as the activation function, and finally a matrix with a size of 32x24x32 is output.

[0100] The upper branch of the twenty-fourth layer is a third up-sampling layer, which receives a matrix with a size of 128x96x64 output by the upper branch of the twenty-third layer; in the present layer, the pooling mode is maximum pooling, the up-sampling factor is 2x2, and the step length is 2. Finally, a matrix with a size of 256x192x128 is output.

[0101] The lower branch of the twenty-fourth layer is a twenty-third convolutional layer, which receives a matrix with a size of 32x24x32 output by the lower branch of the twenty-third layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 32, the SAME padding mode is adopted, group normalization is performed, a leaky rectified linear unit function is used as the activation function, and finally a matrix with a size of 32x24x32 is output.

[0102] The upper branch of the twenty-fifth layer is a second variability convolutional layer, which receives a matrix with a size of 256x192x128 output by the upper branch of the twenty-fourth layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 64, the SAME padding mode is adopted, group normalization is performed, a leaky rectified linear unit function is used as the activation function, and finally a matrix with a size of 256x192x64 is output.

[0103] The lower branch of the twenty-fifth layer is a first fully connected layer, which receives a matrix with a size of 32x24x32 output by the lower branch of the twenty-fourth layer; in the present layer, the number of output nodes is 3, and a nonlinear conversion is performed on the twenty-third convolutional layer to obtain a first feature vector, followed by a random inactivation operation to prevent overfitting.

[0104] The twenty-sixth layer is a twenty-fourth convolutional layer, which receives a matrix with a size of 256x192x64 output by the twenty-fifth layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 64, the SAME padding mode is adopted, group normalization is performed, a leaky rectified linear unit function is used as the activation function, and finally a matrix with a size of 256x192x64 is output.

[0105] The twenty-seventh layer is a twenty-fifth convolutional layer, which receives a matrix with a size of 256x192x64 output by the twenty-sixth layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 64, the SAME padding mode is adopted, group normalization is performed, a leaky rectified linear unit function is used as the activation function, and finally a matrix with a size of 256x192x64 is output.

[0106] The twenty-eighth layer is a fourth up-sampling layer, which receives a matrix with a size of 256x192x64 output by the twenty-seventh layer; in the present layer, the pooling mode is maximum pooling, the up-sampling factor is 2x2, the step length is 2, and finally a matrix with a size of 512x384x64 is output.

[0107] The twenty-ninth layer is a third variability convolutional layer, which receives a matrix with a size of 512x384x64 output by the twenty-eighth layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 32, the SAME padding mode is adopted, group normalization is performed, a leaky rectified linear unit function is used as the activation function, and finally a matrix with a size of 512x384x32 is output.

[0108] The thirtieth layer is a twenty-sixth convolutional layer, which receives a matrix with a size of 512x384x32 output by the twenty-ninth layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 32, the SAME padding mode is adopted, group normalization is performed, a leaky rectified linear unit function is used as the activation function, and finally a matrix with a size of 512x384x32 is output.

[0109] The thirty-first layer is a twenty-seventh convolutional layer, which receives a matrix with a size of 512x384x32 output by the thirtieth layer; in the present layer, the convolution kernel size is 3x3, the step length is 1, the number of convolution kernels is 32, the SAME padding mode is adopted, group normalization is performed, a leaky rectified linear unit function is used as the activation function, and finally a matrix with a size of 512x384x32 is output.

[0110] The thirty-second layer is a second fully connected layer, which receives a matrix with a size of 512x384x32 output by the thirty-first layer; in the present layer, the number of output nodes is 3, and a nonlinear conversion is performed on the twenty-seventh convolutional layer to obtain a second feature vector, followed by a random deactivation operation to prevent overfitting.

[0111] In the segmentation convolutional network, the first layer to the fifteenth layer are an encoding structure of the segmentation convolutional network, and the sixteenth layer to the thirty-second layer are a decoding structure of the segmentation convolutional network; wherein: the decoding structure comprises a first up-sampling layer, an upper branch structure, a lower branch structure, and a network structure connected after the upper branch structure, the upper branch structure comprises nine upper branch layers, i.e., a seventeenth layer upper branch to a twenty-fifth layer upper branch, the lower branch structure comprises nine lower branch layers, i.e., a seventeenth layer lower branch to a twenty-fifth layer lower branch, and the network structure connected after the upper branch structure is a twenty-sixth layer to a thirty-second layer.

[0112] It is worth noting that in the embodiment, the lower branch constrains the upper branch of the corresponding layer through the attention gate; the first variable convolutional layer, the second variable convolutional layer, and the third variable convolutional layer are used to eliminate the influence of the deformation of the segmentation target.

[0113] S103, fitting the pubic region segmentation result and the fetal head region segmentation result, and then measuring the progress of labor angle.

[0114] S1031, performing ellipse fitting on the pubic region segmentation result and the fetal head region segmentation result to obtain a pubic region ellipse and a fetal head region ellipse, as shown in Figure 5 .

[0115] S1032, determining the coordinates of the left and right end points of the long axis of the pubic region ellipse.

[0116] S1033, taking the right end point of the long axis of the pubic region ellipse as a support point, and finding the coordinates of the left and right tangent points on the fetal head region ellipse.

[0117] S1034, calculating the progress of labor angle according to the determined coordinates of the left and right end points and the found coordinates of the right tangent point, as shown in Figure 6 .

[0118] The performance of the method provided in the embodiment is tested, and the performance test results are as follows:

[0119] Specifically, the method provided in the embodiment is tested by using five-fold cross-validation. The Dice coefficient (D), accuracy (A), and absolute value of the difference of the progress of labor angle (B) are used as criteria to evaluate the effectiveness of the method provided in the embodiment, D is as high as 93.38%, A is as high as 98.74%, and B is 5.365 degrees, so it can meet the current clinical demand for measuring the progress of labor angle.

[0120] Those skilled in the art can understand that all or part of the steps in the method of implementing the above embodiments can be completed by programs instructing related hardware, and the corresponding programs can be stored in a computer readable storage medium.

[0121] It should be noted that, although the method operations of the above-described embodiments are described in a particular, sequential order, this order is not meant to be a requirement or a limitation, and any of the operations can be performed in any order, and not all operations can be required to be performed, or even performed at all, to achieve a desired result. The depicted steps can be changed, omitted, combined, and / or separated into further steps.

[0122] Embodiment 2:

[0123] As shown in the above embodiment, the labor progress angle measurement system based on ultrasound images comprises an acquisition unit 701, a segmentation unit 702 and a fitting and measurement unit 703, and the specific functions of each unit are as follows: Figure 7 The acquisition unit 701 is configured to acquire an ultrasound image, wherein the ultrasound image is a preprocessed standard section image of a maternal-fetal ultrasound during labor;

[0124] The segmentation unit 702 is configured to input the ultrasound image into a trained segmentation convolutional network to obtain a pubic region segmentation result and a fetal head region segmentation result;

[0125] The fitting and measurement unit 703 is configured to fit the pubic region segmentation result and the fetal head region segmentation result, and then measure a labor progress angle.

[0126] Embodiment 3:

[0127] As shown in the above embodiment, the computer device comprises a processor 802, a memory, an input device 803, a display device 804 and a network interface 805 connected through a system bus 801. The processor 802 is configured to provide computing and control capabilities, the memory comprises a non-volatile storage medium 806 and an internal memory 807, the non-volatile storage medium 806 stores an operating system, a computer program and a database, the internal memory 807 provides an environment for the operating system and the computer program in the non-volatile storage medium 806 to run, and the computer program is executed by the processor 802 to implement the labor progress angle measurement method of the above embodiment 1, as follows:

[0128] Figure 8 The ultrasound image is acquired, wherein the ultrasound image is a preprocessed standard section image of a maternal-fetal ultrasound during labor;

[0129] The ultrasound image is input into a trained segmentation convolutional network to obtain a pubic region segmentation result and a fetal head region segmentation result;

[0130] The pubic region segmentation result and the fetal head region segmentation result are fitted, and then a labor progress angle is measured.

[0131] The pubic region segmentation result and the fetal head region segmentation result are fitted, and then a labor progress angle is measured. ​

[0132] Embodiment 4:

[0133] The embodiment provides a storage medium, which is a computer-readable storage medium, and stores a computer program. When the computer program is executed by a processor, a labor progress angle measurement method in the embodiment 1 is realized, and the method comprises the following steps:

[0134] An ultrasound image is acquired, and the ultrasound image is a preprocessed standard section image of a perinatal mother-fetus ultrasound.

[0135] The ultrasound image is input into the trained segmentation convolutional network to obtain a pubic bone region segmentation result and a fetal head region segmentation result.

[0136] The pubic bone region segmentation result and the fetal head region segmentation result are fitted, and then a labor progress angle is measured.

[0137] It should be noted that the computer-readable storage medium of the embodiment can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. The computer-readable storage medium may, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination of the above. More specific examples of computer-readable storage media can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0138] In the embodiment, the computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, device or component. In the embodiment, the computer-readable signal medium can include a data signal carried in a baseband or as a part of a carrier wave, which carries computer-readable programs. Such a propagated data signal can take various forms, including but not limited to an electromagnetic signal, an optical signal or any suitable combination of the above. The computer-readable signal medium can also be any computer-readable storage medium other than the computer-readable storage medium, which can send, propagate or transmit programs for use by or in connection with an instruction execution system, device or component. The computer program contained on the computer-readable storage medium can be transmitted by any suitable medium, including but not limited to a wire, an optical cable, an RF (radio frequency) or the like, or any suitable combination of the above.

[0139] The computer readable storage medium described above can be written in one or more programming languages or combinations of languages for executing the computer program of the present embodiments, including object oriented programming languages such as Java, Python, C++ and conventional procedural programming languages such as C language or similar programming languages. The program can be executed entirely on the user computer, partially on the user computer, as an independent software package, partially on the user computer and partially on a remote computer, or entirely on a remote computer or server. In the case involving a remote computer, the remote computer can be connected to the user computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, through the Internet by using an Internet service provider).

[0140] In conclusion, the present application solves the technical problems of the prior art that the workload and time consumption of the method based on manual measurement are large, and different ultrasound doctors obtain different quality control results due to high dependence on the technology and professional level of the ultrasound doctors, thereby affecting the accuracy of the results of the perinatal ultrasound examination.

[0141] The above is only the preferred embodiment of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can make equivalent replacements or changes to the technical scheme and the inventive concept of the present application within the scope disclosed by the present application, and such replacements or changes are also within the protection scope of the present application.

Claims

1. A method for measuring the angle of labor progression based on ultrasound images, characterized in that, The method includes: Acquire ultrasound images, which are preprocessed standard cross-sectional images of maternal and fetal ultrasound during labor; The ultrasound images are input into a trained segmentation convolutional network to obtain segmentation results for the pubic region and the fetal head region. The segmented convolutional network includes an encoding structure and a decoding structure. The decoding structure includes an upper branch structure and a lower branch structure. The upper branch structure includes multiple layers of upper branches connected in sequence, and the lower branch structure includes multiple layers of lower branches connected in sequence. The multiple upper branches and multiple lower branches correspond one-to-one. Among them, at least two upper branches are variable convolutional layers used to eliminate the influence of segmentation target deformation; the lower branches constrain the upper branches of the corresponding layers through attention gates; The pubic region segmentation results and the fetal head region segmentation results are fitted together to obtain the labor progression angle, specifically including: Ellipse fitting was performed on the segmentation results of the pubic region and the fetal head region to obtain ellipses for the pubic region and the fetal head region. Determine the coordinates of the left and right endpoints of the major axis of the pubic region ellipse; Using the right endpoint of the major axis of the pubic region ellipse as the support point, find the coordinates of the left and right tangent points on the ellipse of the fetal head region; Calculate the labor progression angle based on the coordinates of the determined left and right endpoints and the coordinates of the found right tangent point.

2. The method for measuring the angle of labor progression according to claim 1, characterized in that, The encoding structure includes a first convolutional layer, a second convolutional layer, a first downsampling layer, a third convolutional layer, a fourth convolutional layer, a second downsampling layer, a fifth convolutional layer, a sixth convolutional layer, a third downsampling layer, a seventh convolutional layer, an eighth convolutional layer, a fourth downsampling layer, a ninth convolutional layer, and a tenth convolutional layer connected in sequence.

3. The method for measuring the angle of labor progression according to claim 2, characterized in that, The decoding structure further includes a first upsampling layer, the tenth convolutional layer, the first upsampling layer, and the upper branch structure are connected in sequence, and the tenth convolutional layer and the lower branch structure are connected in sequence.

4. The method for measuring the angle of labor progression according to any one of claims 1-3, characterized in that, The upper branch structure includes an eleventh convolutional layer, a thirteenth convolutional layer, a fifteenth convolutional layer, a second upsampling layer, a first variable convolutional layer, a nineteenth convolutional layer, a twenty-first convolutional layer, a third upsampling layer, and a second variable convolutional layer connected in sequence.

5. The method for measuring the angle of labor progression according to any one of claims 1-3, characterized in that, The lower branch structure includes a twelfth convolutional layer, a fourteenth convolutional layer, a sixteenth convolutional layer, a seventeenth convolutional layer, an eighteenth convolutional layer, a twentieth convolutional layer, a twenty-second convolutional layer, a twenty-third convolutional layer, and a first fully connected layer connected in sequence.

6. The method for measuring the angle of labor progression according to claim 1, characterized in that, The decoding structure further includes a 24th convolutional layer, a 25th convolutional layer, a 4th upsampling layer, a 3rd variable convolutional layer, a 26th convolutional layer, a 27th convolutional layer, and a 2nd fully connected layer connected in sequence; the upper branch structure and the 24th convolutional layer are connected in sequence.

7. A system for measuring the angle of labor progression based on ultrasound images, characterized in that, The method for measuring the angle of labor progression according to any one of claims 1-6; the system comprises: The acquisition unit is used to acquire ultrasound images, which are preprocessed standard cross-sectional images of maternal and fetal ultrasound during childbirth. The segmentation unit is used to input ultrasound images into a trained segmentation convolutional network to obtain segmentation results for the pubic region and the fetal head region. The fitting and measurement unit is used to fit the segmentation results of the pubic region and the fetal head region, and then measure the labor progression angle.

8. A storage medium storing a program, characterized in that, When the program is executed by the processor, it implements the method for measuring the labor progress angle as described in any one of claims 1-6.

Images