An ultrasound image processing method, apparatus, device and medium

Abstract

This application discloses an ultrasound image processing method, apparatus, device, and medium, comprising: identifying a standard section of the fetal head in an ultrasound video, or identifying a standard section of the fetal head in real-time ultrasound mode; inputting the standard section of the fetal head into an image segmentation model to obtain a segmentation result of the standard section of the fetal head; wherein the segmentation result includes the contours of key structures within the fetal head and the midline of the brain; and determining the measurement result of the standard section of the fetal head based on the contours of the key structures and the midline of the brain. This method can improve the accuracy of measuring the standard section of the fetal head.

Description

Technical Field

[0001] This application relates to the field of ultrasound image processing technology, and in particular to an ultrasound image processing method, apparatus, device and medium. Background Technology

[0002] During fetal ultrasound scans, it is usually necessary to measure a standard section of the fetal head to obtain the ultrasound scan results. Currently, there are two main measurement methods: one is to manually measure the standard section of the fetal head, but this method relies on the doctor's experience and is time-consuming; the other is to use deep learning technology for measurement, but when the contours of key head structures are irregular, deviations can easily occur, resulting in inaccurate measurement results. Summary of the Invention

[0003] In view of this, the purpose of this application is to provide an ultrasound image processing method, apparatus, device, and medium that can improve the accuracy of standard section measurement of the fetal head. The specific solution is as follows:

[0004] In a first aspect, this application discloses an ultrasound image processing method, comprising:

[0005] Identify the standard section of the fetal head in ultrasound video, or identify the standard section of the fetal head in real-time ultrasound mode;

[0006] The standard cross-section of the fetal head is input into the image segmentation model to obtain the segmentation result of the standard cross-section of the fetal head; wherein, the segmentation result includes the outline of key structures within the fetal head and the midline of the brain;

[0007] Measurement results of the standard section of the fetal head determined based on the key structural contour and the midline of the brain.

[0008] Optionally, the standard section of the fetal head includes a standard section of the thalamus and a standard section of the cerebellum, the key structural contours include the target thalamic contour and the target cerebellar contour, and the midline of the brain includes a first midline and a second midline of the brain; correspondingly, the measurement results of determining the standard section of the fetal head based on the key structural contours and the midline of the brain include...

[0009] The biparietal diameter measurement results of the thalamus in the standard thalamic section are determined based on the target thalamic contour and the first midline of the brain.

[0010] The cerebellar transverse diameter measurement result of the cerebellar standard section is determined based on the target cerebellar contour of the cerebellar standard section and the second brain midline.

[0011] Optionally, the measurement result of the biparietal diameter of the thalamus based on the target thalamic contour of the standard thalamic section and the first midline of the brain includes:

[0012] The slope of the biparietal diameter line is determined based on the first midline of the brain;

[0013] The bipolar diameter line is determined based on its slope.

[0014] The biparietal diameter line segment is determined based on the biparietal diameter straight line and the target thalamic contour;

[0015] The biparietal diameter measurement results of the thalamus in the standard section of the thalamus are determined based on the biparietal diameter line segment.

[0016] Optionally, determining the biparietal diameter line based on its slope; and determining the biparietal diameter line segment based on the biparietal diameter line and the target thalamic contour, includes:

[0017] Determine the centroid of the target thalamus contour, and determine the target pixel range based on the centroid;

[0018] Multiple bi-vertex lines are determined based on the target pixel range and the slope of the bi-vertex line;

[0019] Multiple line segments are determined based on the multiple biparietal diameter straight lines and the target thalamic contour;

[0020] The longest line segment among the plurality of line segments is determined as the bipolar line segment.

[0021] Optionally, the measurement result of the biparietal diameter of the thalamus based on the biparietal diameter segment to determine the standard section of the thalamus includes:

[0022] The contour of the inferior bony ring is determined from the standard section of the thalamus, and the lowest point of the inferior bony ring contour is determined as the final bottom endpoint of the biparietal diameter line segment, thus obtaining the final biparietal diameter line segment.

[0023] The biparietal diameter measurement result of the standard thalamic section is determined based on the final biparietal diameter segment.

[0024] Optionally, determining the contour of the inferior bony ring from the standard section of the thalamus includes:

[0025] The first region is defined based on the bottom endpoint of the bipolar diameter line segment;

[0026] Based on the first region range, an image of the first region is extracted from the standard section of the thalamus;

[0027] The contour of the lower bone ring is determined from the first region image.

[0028] Optionally, the measurement result of the cerebellar transverse diameter of the cerebellar standard section based on the target cerebellar contour of the cerebellar standard section and the second midline of the brain includes:

[0029] The slope of the cerebellar transverse diameter line is determined based on the second midline of the brain;

[0030] The cerebellar transverse diameter line is determined based on the slope of the cerebellar transverse diameter line.

[0031] The cerebellar transverse diameter line segment is determined based on the cerebellar transverse diameter line and the target cerebellar contour.

[0032] The cerebellar transverse diameter measurement results are determined based on the cerebellar transverse diameter segment.

[0033] Optionally, the measurement result of the cerebellar transverse diameter based on the cerebellar transverse diameter segment to determine the standard cerebellar section includes:

[0034] The extent of the second region and the extent of the third region are determined based on the two endpoints of the transverse cerebellar diameter segment, respectively.

[0035] Images of the second and third regions are extracted from the standard cerebellar section based on the ranges of the second and third regions, respectively.

[0036] The local cerebellar contours in the second and third region images were determined respectively;

[0037] Determine the intersection point of the cerebellar transverse diameter line and the local contour of the cerebellum;

[0038] The final cerebellar transverse diameter segment is determined based on the intersection point;

[0039] The cerebellar transverse diameter measurement result of the standard cerebellar section is determined based on the final cerebellar transverse diameter segment.

[0040] Optionally, the standard section of the fetal head further includes a standard section of the lateral brain, and the key structural contour further includes the contour of the target lateral ventricle; correspondingly, the method further includes:

[0041] The measurement results of the posterior horn diameter of the lateral ventricle in the standard lateral ventricle section are determined based on the target lateral ventricle contour of the standard lateral ventricle section.

[0042] Optionally, determining the measurement result of the posterior horn diameter of the lateral ventricle in the standard lateral ventricle section based on the target lateral ventricle contour of the standard lateral ventricle section includes:

[0043] The location information of the choroid plexus in the lateral ventricle is determined based on the target lateral ventricle contour;

[0044] Based on the location information, determine the straight line of the inner diameter of the posterior horn of the lateral ventricle;

[0045] The inner diameter segment of the posterior horn of the lateral ventricle is determined based on the straight line of the lateral ventricle diameter and the contour of the target lateral ventricle.

[0046] The measurement result of the posterior horn diameter of the lateral ventricle in the standard section of the lateral brain is determined based on the line segment of the posterior horn diameter of the lateral ventricle.

[0047] Optionally, determining the location information of the choroid plexus in the lateral ventricle based on the target lateral ventricle contour includes:

[0048] A local lateral ventricle image is determined from the standard cross-section of the lateral brain based on the target lateral ventricle contour;

[0049] The target lateral ventricle image is obtained by multiplying the mask of the target lateral ventricle outline with the local lateral ventricle image.

[0050] The mask of the target lateral ventricle contour is refined to obtain the first midline.

[0051] Multiply the first midline and the target lateral ventricle image to obtain the second midline of the choroid plexus;

[0052] By fitting the first central axis and the second central axis with straight lines respectively, a first line segment and a second line segment are obtained;

[0053] The location information of the choroid plexus in the lateral ventricle is determined based on the first line segment and the second line segment.

[0054] Optionally, the training process of the image segmentation model includes:

[0055] Obtain a first training sample set; the first training sample set includes standard cross-sectional samples of the fetal head and corresponding label information for the standard cross-sectional samples of the fetal head; the standard cross-sectional samples of the fetal head include standard cross-sectional samples of the thalamus, standard cross-sectional samples of the cerebellum, and standard cross-sectional samples of the lateral brain; the label information for the standard cross-sectional samples of the thalamus includes the thalamic contour line and the midline of the brain, the label information for the standard cross-sectional samples of the cerebellum includes the cerebellar contour line and the midline of the brain, and the label information for the standard cross-sectional samples of the lateral brain includes the lateral ventricle contour line;

[0056] Input the standard cross-sectional sample of the fetal head into the first initial model to obtain the thalamic contour, cerebellar contour, midline of the brain and lateral ventricle contour corresponding to the standard cross-sectional sample of the fetal head.

[0057] The training loss is calculated based on the label information of the thalamic contour, the cerebellar contour, the midline of the brain, the lateral ventricle contour, and the standard cross-sectional sample of the fetal head, and the parameters of the first initial model are adjusted based on the training loss.

[0058] If the training completion condition is met, the first initial model with adjusted parameters is determined as the image segmentation model.

[0059] Optionally, identifying the standard section of the fetal head in the ultrasound video, or identifying the standard section of the fetal head in real-time ultrasound mode, includes:

[0060] Using a standard section recognition model, the standard section of the fetal head in ultrasound video can be identified, or the standard section of the fetal head can be identified in real-time ultrasound mode.

[0061] The standard section recognition model is obtained by training a second initial model using a second training sample set, which includes standard section samples of the fetal head, standard section samples of non-fetal head, and label information.

[0062] Secondly, this application discloses an ultrasound image processing apparatus, comprising:

[0063] The standard section module is used to identify the standard section of the fetal head in ultrasound video, or to identify the standard section of the fetal head in real-time ultrasound mode.

[0064] An image segmentation module is used to input the standard cross-section of the fetal head into an image segmentation model to obtain the segmentation result of the standard cross-section of the fetal head; wherein, the segmentation result includes the outline of key structures within the fetal head and the midline of the brain;

[0065] The image measurement module is used to determine the measurement results of the standard section of the fetal head based on the key structural contour and the midline of the brain.

[0066] Thirdly, this application discloses an ultrasonic device, including a processor and a memory; wherein,

[0067] The memory is used to store computer programs;

[0068] The processor is used to execute the computer program to implement the aforementioned ultrasound image processing method.

[0069] Fourthly, this application discloses a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the aforementioned ultrasound image processing method.

[0070] As can be seen, this application identifies a standard section of the fetal head in an ultrasound video, then inputs the standard section of the fetal head into an image segmentation model to obtain a segmentation result of the standard section of the fetal head. The segmentation result includes the contours of key structures within the fetal head and the midline of the brain. Finally, the measurement result of the standard section of the fetal head is determined based on the contours of the key structures and the midline of the brain. That is, this application first uses an image segmentation model to determine the contours of key structures within the fetal head and the midline of the brain, and then uses the midline of the brain to assist in automatic measurement, ensuring the accuracy of the direction of the measurement items in the standard section of the fetal head, and thus improving the accuracy of the measurement of the standard section of the fetal head. Attached Figure Description

[0071] To more clearly illustrate the technical solutions in the embodiments of this application 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 embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0072] Figure 1 This is a flowchart of an ultrasound image processing method disclosed in this application;

[0073] Figure 2 This is a schematic diagram of a standard cross-section and segmentation of the thalamus disclosed in this application;

[0074] Figure 3 This is a schematic diagram of a standard cerebellar section annotation and segmentation disclosed in this application;

[0075] Figure 4 This is a schematic diagram of the standard cross-sectional annotation and segmentation of the lateral brain disclosed in this application;

[0076] Figure 5 This is a schematic diagram of a specific ultrasound image processing method disclosed in this application;

[0077] Figure 6 This is a schematic diagram of the structure of an ultrasound image processing device disclosed in this application;

[0078] Figure 7 This is a structural diagram of an ultrasonic device disclosed in this application. Detailed Implementation

[0079] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0080] Currently, there are two main methods for measuring the standard section of the fetal head: one is manual measurement, which relies on the doctor's experience and is time-consuming; the other is measurement using deep learning technology, but this method is prone to deviation and inaccurate results when the contours of key head structures are irregular. Therefore, this application provides an ultrasound image processing solution that can improve the accuracy of measuring the standard section of the fetal head.

[0081] See Figure 1As shown in the figure, this application discloses an ultrasound image processing method, including:

[0082] Step S11: Identify the standard section of the fetal head in the ultrasound video, or identify the standard section of the fetal head in real-time ultrasound mode.

[0083] In one embodiment, a standard section recognition model can be used to identify standard sections of the fetal head in ultrasound videos or in real-time ultrasound mode; wherein, the standard section recognition model is obtained by training a second initial model using a second training sample set, the second training sample set including fetal head standard section samples, non-fetal head standard section samples, and label information.

[0084] Furthermore, the training process for the standard section recognition model can include: annotating the acquired ultrasound video with standard sections of the thalamus, cerebellum, and lateral brain using annotation software. The initial model can use object detection algorithms such as RCNN (Region Convolutional Neural Networks), YOLO, and RetinaNet. The annotated ultrasound images and their labels are input into the initial model for training, iterative optimization, and finally, a model with the highest recognition accuracy is trained.

[0085] It is important to note that in fetal ultrasound scans, the thalamus, cerebellum, and lateral brain sections are fundamental and crucial for diagnosing whether fetal head development is normal. The standard criteria for the thalamus section are: an oval-shaped, strongly echoing ring on the skull, symmetrical bilateral cerebral hemispheres, a centered midline, and clear visualization of the cavum septum pellucidum, symmetrical bilateral thalamus, and the cleft-like third ventricle between the thalamus. The measurement item is the biparietal diameter (BPD), also known as the fetal head biparietal diameter, which is the length of the widest part between the left and right sides of the fetal head, also called the "greatest transverse diameter of the head." The standard criteria for the cerebellum section are: an oval-shaped, strongly echoing ring on the skull, clearly visible and symmetrical cerebellar hemispheres, and the cavum septum pellucidum anteriorly. The measurement item is the transverse cerebellar diameter, which is the horizontal length of the fetal cerebellum and is often used to predict gestational age, serving as an ideal indicator of actual gestational age. The standard criteria for the lateral brain section are: a clear, anechoic posterior horn of the lateral ventricle, containing the hyperechoic choroid plexus, but not completely filling the posterior horn. Both thalamus are still visible in the center of the image, and the midline of the brain is visible. The lateral walls of the frontal horn of the lateral ventricle are almost parallel to the falx cerebri, while the occipital horn separates laterally and is relatively far from the midline of the brain. The posterior horn diameter is measured by placing the vernier at the widest point of the lateral ventricle on its medial border, perpendicular to the long axis of the lateral ventricle. Measuring its maximum width can determine the presence of ventricular dilatation and hydrocephalus. Throughout the entire pregnancy, the occipital horn diameter of the fetal lateral ventricle should be less than 10 mm.

[0086] In this way, doctors no longer need to manually obtain a standard section of the fetal head; the standard section of the fetal head is automatically obtained during the scanning process, improving diagnostic efficiency.

[0087] Step S12: Input the standard section of the fetal head into the image segmentation model to obtain the segmentation result of the standard section of the fetal head; wherein, the segmentation result includes the outline of key structures within the fetal head and the midline of the brain.

[0088] The standard section of the fetal head includes a standard section of the thalamus and a standard section of the cerebellum; the key structural contours include the contours of the target thalamus and the target cerebellum; and the midline of the brain includes the first midline and the second midline of the brain. In addition, in one embodiment, the standard section of the fetal head also includes a standard section of the lateral brain, and the key structural contours also include the contour of the target lateral ventricle.

[0089] In one implementation, the training process of the image segmentation model may include:

[0090] Step 00: Obtain the first training sample set; the first training sample set includes standard cross-sectional samples of the fetal head and corresponding label information; the standard cross-sectional samples of the fetal head include standard cross-sectional samples of the thalamus, cerebellum, and lateral brain; the label information of the standard cross-sectional samples of the thalamus includes the thalamic contour line and the midline of the brain, the label information of the standard cross-sectional samples of the cerebellum includes the cerebellum contour line and the midline of the brain, and the label information of the standard cross-sectional samples of the lateral brain includes the lateral ventricle contour line.

[0091] Step 01: Input the standard cross-sectional sample of the fetal head into the first initial model to obtain the thalamic contour, cerebellar contour, midline of the brain, and lateral ventricle contour corresponding to the standard cross-sectional sample of the fetal head.

[0092] Step 02: Calculate the training loss based on the label information of the thalamic contour, the cerebellar contour, the midline of the brain, the lateral ventricle contour, and the standard cross-sectional sample of the fetal head, and adjust the parameters of the first initial model based on the training loss.

[0093] Step 03: If the training completion condition is met, the first initial model with adjusted parameters is determined as the image segmentation model.

[0094] Among them, the training completion conditions can be that the training loss is less than a preset loss threshold and the number of iterations reaches a preset number threshold.

[0095] For example, see Figure 2 , Figure 3 , Figure 4 As shown, Figure 2 This is a schematic diagram of a standard cross-section and segmentation of the thalamus disclosed in this application; Figure 3 This is a schematic diagram of a standard cerebellar section annotation and segmentation disclosed in this application; Figure 4 This is a schematic diagram illustrating the annotation and segmentation of a standard lateral brain section as disclosed in this application. The collected standard sections of the thalamus, cerebellum, and lateral brain can be annotated using annotation software, such as... Figure 2 (a) shows a schematic diagram of the standard cross-section of the thalamus, including three labels: the thalamus-cranial contour line (thalamus contour line), the midline of the brain, and the thalamus-biparietal diameter measurement line. Figure 3 (a) shows a schematic diagram of the standard cerebellar section, including three labels: the cerebellar hemisphere outline (cerebellar outline line), the midline of the brain, and the cerebellar transverse diameter measurement line. Figure 4 (a) shows a schematic diagram of the standard cross-section annotation of the lateral brain, including two labels: the lateral ventricle contour line and the posterior horn diameter measurement line of the lateral ventricle. The annotated image and its labels are input into the second initial model for training. In one embodiment, the second initial model can be an improved model based on the human posture key point detection model. The original key point detection model is used to detect multiple key points of the human body, including the nose, eyes, ears, shoulders, elbows, wrists, knees, etc. In this embodiment, the original multiple channels are changed to four channels. The first three channels are changed to semantic segmentation channels for learning the thalamic contour, cerebellar contour, and lateral ventricle contour, and the fourth channel is changed to a line detection channel for learning the midline of the brain, completing the multi-task measurement of the fetal thalamus, cerebellum, lateral brain contour mask and the midline of the brain. In other words, the model has four output channels, which output the inference results of the thalamic contour, midline of the brain, cerebellum contour, and lateral ventricle contour, respectively. The inference results are masks of the thalamic contour, midline of the brain, cerebellum contour, and lateral ventricle contour. Contour extraction is performed on the masks of the thalamic contour, cerebellum contour, and lateral ventricle contour to obtain the thalamic contour, cerebellum contour, and lateral ventricle contour. Linear fitting is performed on the mask of the midline of the brain to obtain the midline of the brain, which is then plotted on the image, as shown below. Figure 2 As shown in (b) Figure 2 (b) is a schematic diagram of the segmentation results of the standard section of the thalamus, including the outline of the thalamus and the midline of the brain; Figure 3 (b) is a schematic diagram of the segmentation results of the standard cerebellum section, including the cerebellum outline and the midline of the brain; Figure 4 (b) is a schematic diagram of the segmentation results of a standard lateral brain section, including the outline of the lateral ventricle. Iterative training and optimization are performed to ultimately train a model with the highest segmentation accuracy. In another implementation, the second initial model can be MobileNet.

[0096] It should be noted that the thalamic-biparietal diameter measurement line, cerebellar transverse diameter measurement line, and lateral ventricle posterior horn internal diameter measurement line in the label are not used for model training but are used to test the accuracy of the measurement. For example, in the testing phase of this application embodiment, the overlap between the thalamic-biparietal diameter measurement line, cerebellar transverse diameter measurement line, and lateral ventricle posterior horn internal diameter measurement line and the thalamic-biparietal diameter line segment, cerebellar transverse diameter line segment, and lateral ventricle posterior horn internal diameter line segment determined based on the segmentation results can be calculated to measure the accuracy of the measurement.

[0097] Step S13: Determine the measurement results of the standard section of the fetal head based on the key structural contour and the midline of the brain.

[0098] Specifically, step S13 may include:

[0099] Step 10: Determine the biparietal diameter measurement result of the thalamus based on the target thalamic contour of the standard thalamic section and the first midline of the brain.

[0100] In one implementation, step 10 may specifically include the following steps:

[0101] Step 100: Determine the slope of the biparietal diameter line based on the first midline of the brain.

[0102] It is understandable that the first midline of the brain and the biparietal diameter line are perpendicular, and the slope of the biparietal diameter line can be determined based on this perpendicular relationship.

[0103] Step 101: Determine the bipolar diameter line based on the slope of the bipolar diameter line.

[0104] Step 102: Determine the biparietal diameter line segment based on the biparietal diameter line and the target thalamic contour.

[0105] In one implementation, step 102 may specifically include:

[0106] Step 1020: Determine the centroid of the target thalamic contour, and determine the target pixel range based on the centroid. For example, the target pixel range can be obtained by taking an interval of 10 pixels to the left and right of the centroid.

[0107] Step 1021: Determine multiple bivertices based on the target pixel range and the slope of the bivertices. That is, determine multiple bivertices passing through each pixel within the target pixel range and according to the slope of the bivertices.

[0108] Step 1021: Determine multiple line segments based on the multiple biparietal diameter lines and the target thalamic contour.

[0109] In one implementation, the intersection points of the plurality of biparietal diameter lines and the target thalamic contour can be determined separately, and the two intersection points corresponding to each biparietal diameter line can be connected to obtain multiple line segments. In another implementation, the intersection of the plurality of biparietal diameter lines and the mask of the target thalamic contour can be determined separately to obtain multiple line segments.

[0110] Step 1022: Determine the longest line segment among the multiple line segments as the bipolar diameter line segment.

[0111] It should be noted that compared with the traditional ellipse fitting method, the method of calculating the vertical line based on the brain midline as the biparietal diameter direction has higher measurement accuracy and avoids the problem of large deviation of the fitted ellipse when the thalamus contour is irregular, which would lead to a large offset in the biparietal diameter direction.

[0112] Step 103: Determine the measurement result of the biparietal diameter of the thalamus in the standard section of the thalamus based on the biparietal diameter line segment.

[0113] Furthermore, step 103 may specifically include:

[0114] Step 1030: Determine the contour of the inferior bony ring from the standard section of the thalamus, and determine the lowest point of the inferior bony ring contour as the final bottom endpoint of the biparietal diameter segment to obtain the final biparietal diameter segment.

[0115] In one implementation, a first region can be determined based on the bottom endpoint of the biparietal diameter segment; an image of the first region is captured from a standard section of the thalamus based on the first region; and the contour of the inferior bony ring is determined from the first region image. If the contour of the inferior bony ring cannot be determined from the first region image, the bottom endpoint of the biparietal diameter segment is shifted upwards by a preset upward shift value to obtain the final bottom endpoint of the biparietal diameter segment.

[0116] For example, a square region is defined centered on the bottom endpoint of the biparietal diameter segment and with a first preset size, serving as the first region's range. The image of the square region corresponding to the standard thalamic section is extracted, thresholded, and the largest contour is taken as the inferior bony ring contour. If the inferior bony ring contour cannot be found, the bottom endpoint of the biparietal diameter segment is moved up 10 pixels to obtain the final biparietal diameter segment, which is then plotted on the image, as shown below. Figure 2 As shown in (b). It should be noted that, since the bottom end of the biparietal diameter segment cannot pass through the lower bone ring according to clinical requirements, the embodiments of this application can ensure that the bottom end of the biparietal diameter segment does not pass through the lower bone ring.

[0117] Step 1031: Determine the measurement result of the biparietal diameter of the thalamus in the standard section of the thalamus based on the final biparietal diameter segment.

[0118] This embodiment of the application can determine the final length of the biparietal diameter segment as the measurement result of the biparietal diameter of the thalamus in the standard section of the thalamus. Furthermore, the biparietal diameter measurement result is displayed on the ultrasound device interface. It should be noted that before 31 weeks of gestation, the growth rate is 3 mm / week; from 31 to 36 weeks of gestation, the growth rate is 1.5 mm / week; and after 36 weeks of gestation, the growth rate is 1 mm / week. Based on this standard, when the biparietal diameter of the thalamus is abnormal, an abnormality alert can be given on the main ultrasound interface, and the doctor can take appropriate action based on the alert.

[0119] Step 11: Determine the cerebellar transverse diameter measurement result of the cerebellar standard section based on the target cerebellar contour of the cerebellar standard section and the second midline of the brain.

[0120] In one implementation, step 11 may specifically include:

[0121] Step 110: Determine the slope of the cerebellar transverse diameter line based on the second brain midline.

[0122] It is understandable that the second midline of the brain and the transverse diameter of the cerebellum are perpendicular, and the slope of the biparietal diameter line can be determined based on this perpendicular relationship.

[0123] Step 111: Determine the cerebellar transverse diameter line based on the slope of the cerebellar transverse diameter line.

[0124] In one implementation, the highest and lowest points of the target cerebellar contour can be identified, the midpoint between the highest and lowest points can be determined, and the cerebellar transverse diameter line can be determined through the midpoint according to the slope of the cerebellar transverse diameter line.

[0125] Step 112: Determine the cerebellar transverse diameter line segment based on the cerebellar transverse diameter line and the target cerebellar contour.

[0126] In one implementation, the intersection point of the cerebellar transverse diameter line and the target cerebellar contour can be determined, and the intersection point can be connected to obtain the cerebellar transverse diameter line segment. In another implementation, the intersection of the cerebellar transverse diameter line and the mask of the target cerebellar contour can be determined to obtain the cerebellar transverse diameter line segment.

[0127] Step 113: Determine the measurement result of the cerebellar transverse diameter of the standard cerebellar section based on the cerebellar transverse diameter segment.

[0128] In one implementation, step 113 may specifically include:

[0129] Step 1130: Determine the range of the second region and the range of the third region based on the two endpoints of the cerebellar transverse diameter segment. The top endpoint corresponds to the range of the second region, and the bottom endpoint corresponds to the range of the third region.

[0130] Step 1131: Extract the second region image and the third region image from the standard cerebellar section based on the second region range and the third region range, respectively.

[0131] Step 1132: Determine the local cerebellar contours in the second and third region images, respectively.

[0132] For example, two square regions are defined with the two endpoints of the cerebellar transverse diameter segment as centers and a second preset size, serving as the second and third region ranges. The square region images corresponding to the standard cerebellar cross-sections are extracted, thresholded, and the largest contour is taken as the local contour of the cerebellum.

[0133] Step 1133: Determine the intersection point of the cerebellar transverse diameter line and the local contour of the cerebellum.

[0134] Step 1134: Determine the final cerebellar transverse diameter segment based on the intersection point.

[0135] It is understandable that the cerebellar transverse diameter line intersects each local cerebellar contour at two points. The top endpoint of the final cerebellar transverse diameter line segment is the uppermost of the two intersection points between the cerebellar transverse diameter line and the local cerebellar contour in the second region image, and the bottom endpoint is the lowermost of the two intersection points between the cerebellar transverse diameter line and the local cerebellar contour in the third region image. Then, the final cerebellar transverse diameter line segment is plotted on the image, as shown below. Figure 3 As shown in (b). This avoids the problem of the segmentation model's result being too small, leading to a shorter final cerebellar transverse diameter.

[0136] Step 1135: Determine the cerebellar transverse diameter measurement result of the standard cerebellar section based on the final cerebellar transverse diameter segment.

[0137] This embodiment of the application can determine the final length of the cerebellar transverse diameter segment as the measurement result of the cerebellar transverse diameter in the standard cerebellar section. Furthermore, the cerebellar transverse diameter measurement result is displayed on the ultrasound device interface. It should be noted that before 24 weeks of gestation, the cerebellar transverse diameter value is approximately equal to the gestational week; between 20 and 38 weeks of gestation, the growth rate is 1-2 mm / week; after 38 weeks of gestation, the growth rate is 0.7 mm / week. Based on this standard, when the cerebellar transverse diameter is abnormal, an abnormality alert can be given on the main ultrasound interface, and the doctor can take appropriate action based on the alert.

[0138] In this way, the midline information of the brain is incorporated into the calculation of the biparietal diameter and the transverse cerebellar diameter, ensuring that the directions of the biparietal diameter and the transverse cerebellar diameter are correct and will not be deviated, thus improving the measurement accuracy.

[0139] Furthermore, embodiments of this application can also determine the measurement result of the posterior horn diameter of the lateral ventricle in the standard lateral ventricle section based on the target lateral ventricle contour of the standard lateral ventricle section. Specifically, this includes the following steps:

[0140] Step 21: Determine the location information of the choroid plexus in the lateral ventricle based on the target lateral ventricle contour.

[0141] Furthermore, step 21 may specifically include:

[0142] Step 210: Determine a local lateral ventricle image from the standard cross-section of the lateral brain based on the target lateral ventricle contour.

[0143] In one implementation, the minimum bounding rectangle of the target lateral ventricle contour can be determined, and a local lateral ventricle image can be determined from the standard cross-section of the lateral brain based on the minimum bounding rectangle.

[0144] Step 211: Multiply the mask of the target lateral ventricle contour with the local lateral ventricle image to obtain the target lateral ventricle image.

[0145] Understandably, this would produce an image of the lateral ventricle where everything except the outline of the lateral ventricle is black.

[0146] Step 212: Refine the mask of the target lateral ventricle contour to obtain the first midline.

[0147] Step 213: Multiply the first midline and the target lateral ventricle image to obtain the second midline of the choroid plexus.

[0148] Step 214: Perform linear fitting on the first central axis and the second central axis respectively to obtain the first line segment and the second line segment.

[0149] Step 215: Determine the location information of the choroid plexus in the lateral ventricle based on the first line segment and the second line segment.

[0150] In one implementation, a first distance between the two left endpoints of the first line segment and the second line segment, and a second distance between the two right endpoints can be calculated respectively. If the first distance is less than the second distance, the choroid plexus is determined to be located on the left side of the lateral ventricle. If the first distance is greater than the second distance, the choroid plexus is determined to be located on the right side of the lateral ventricle.

[0151] Step 22: Determine the straight line of the inner diameter of the posterior horn of the lateral ventricle based on the location information.

[0152] In one implementation, if the choroid plexus is located on the left side of the lateral ventricle, a perpendicular line to the second line segment is determined through the right endpoint of the second line segment, which serves as the straight line of the inner diameter of the posterior horn of the lateral ventricle. If the choroid plexus is located on the right side of the lateral ventricle, a perpendicular line to the second line segment is determined through the left endpoint of the second line segment, which serves as the straight line of the inner diameter of the posterior horn of the lateral ventricle.

[0153] Step 23: Determine the inner diameter segment of the posterior horn of the lateral ventricle based on the straight line of the lateral ventricle diameter and the outline of the target lateral ventricle.

[0154] In one implementation, the intersection point of the straight line of the lateral ventricle diameter and the target lateral ventricle contour can be determined, and the intersection point is connected to obtain the posterior horn inner diameter line segment of the lateral ventricle. In another implementation, the intersection of the mask of the straight line of the lateral ventricle diameter and the target lateral ventricle contour can be determined to obtain the posterior horn inner diameter line segment of the lateral ventricle. Then, the posterior horn inner diameter line segment of the lateral ventricle is plotted on the image, as shown below. Figure 4 As shown in (b).

[0155] Step 24: Determine the measurement result of the posterior horn diameter of the lateral ventricle in the standard section of the lateral brain based on the line segment of the posterior horn diameter of the lateral ventricle.

[0156] This application embodiment can determine the length of the posterior horn diameter segment of the lateral ventricle as the measurement result of the posterior horn diameter of the lateral ventricle in the standard lateral brain section. Furthermore, the measurement result of the posterior horn diameter of the lateral ventricle is displayed on the ultrasound device interface. It should be noted that when the lateral ventricle diameter is in the range of 10-15mm, a "mild ventricular dilatation" alert can be given on the main ultrasound interface; when the lateral ventricle diameter is greater than 15mm, a "severe ventricular dilatation or hydrocephalus" alert can be given on the main ultrasound interface, and the doctor can take appropriate action based on the alert.

[0157] For example, see Figure 5 As shown, Figure 5 This is a schematic diagram of a specific ultrasound image processing method disclosed in this application. The ultrasound doctor clicks "Start Scan," and the ultrasound probe scans the pregnant woman's abdomen, obtaining an ultrasound video. In this application embodiment, the ultrasound image from the ultrasound video is input into a standard section recognition model. The model identifies and automatically stores the standard section of the fetal head in the ultrasound video. The standard section of the fetal head includes the standard section of the thalamus, the standard section of the cerebellum, and the standard section of the lateral brain. Then, the standard section of the fetal head is input into a segmentation model to obtain the target thalamic contour and midline of the brain for the thalamic standard section, the target cerebellar contour and midline of the brain for the cerebellar standard section, and the target lateral ventricle contour for the lateral brain standard section. These are then plotted on the image. Subsequently, the biparietal diameter segment, the cerebellar transverse diameter segment, and the posterior horn inner diameter segment of the lateral ventricle are determined and plotted on the image, and the measurement results are obtained.

[0158] Furthermore, in one specific implementation, a standard section of the fetal head is input into the segmentation model to obtain the reasoning results of the three sections, and each section is post-processed.

[0159] In this process, for the standard thalamic section, the inference result of the brain midline is fitted with a straight line to obtain the brain midline. Then, the slope of the biparietal diameter line is obtained based on the vertical relationship. The centroid of the thalamic contour is calculated, and a 10-pixel interval is taken to the left and right of the centroid. The biparietal diameter line is drawn within this interval. The intersection of the biparietal diameter line and the mask of the thalamic contour yields multiple line segments, and their lengths are calculated. The longest line segment is taken as the biparietal diameter line segment. Since the bottom end of the biparietal diameter line segment cannot pass through the inferior ossicular ring according to clinical requirements, this embodiment takes a square area of ​​a first preset size at the bottom endpoint of the biparietal diameter line segment, extracts the square area image of the corresponding position of the standard thalamic section, performs threshold segmentation on it, takes the largest contour to obtain the inferior ossicular ring contour, and calculates the lowest point of the inferior ossicular ring contour as the bottom endpoint of the final biparietal diameter line segment. If the inferior ossicular ring contour cannot be found, the bottom endpoint of the biparietal diameter line segment is moved up by 10 pixels to obtain the final biparietal diameter line segment.

[0160] For the standard cerebellar section, a straight line is fitted to the inference result of the cerebellar midline to obtain the cerebellar midline. Then, the slope of the cerebellar transverse diameter line is obtained based on the vertical relationship. The highest and lowest points of the cerebellar contour are calculated, and the midpoint between the highest and lowest points is calculated. The cerebellar transverse diameter line is drawn through the midpoint. The intersection of the cerebellar transverse diameter line and the mask of the cerebellar contour is taken as the cerebellar transverse diameter line segment. To avoid the segmentation network result being too small, which would lead to a short final cerebellar transverse diameter line segment, a square region is taken at each end of the cerebellar transverse diameter line segment. The corresponding square region image of the cerebellar standard section is extracted and thresholded. The maximum contour is taken to obtain the local contour of the cerebellum. The intersection of the cerebellar transverse diameter line and the local contour of the cerebellum is calculated to determine the final cerebellar transverse diameter line segment.

[0161] For the standard lateral ventricle section, the corresponding local original image of the lateral ventricle is obtained based on the lateral ventricle contour inferred from the network. This image is multiplied with the mask of the lateral ventricle contour to obtain the lateral ventricle contour map with a black area around it. The mask of the lateral ventricle contour is then refined to obtain the first central axis and a straight line is fitted to obtain the first line segment. The lateral ventricle contour map and the refined first central axis are multiplied to obtain the central axis of the choroid plexus. A straight line is fitted to obtain the second line segment. The distance between the left endpoint of the first line segment and the left endpoint of the second line segment, as well as the distance between the right endpoints, are calculated. If the distance between the left endpoints is small, it indicates that the choroid plexus is on the left side of the lateral ventricle. A perpendicular line is drawn at the right endpoint of the second line segment, which is the lateral ventricle internal diameter line. Conversely, the lateral ventricle posterior horn internal diameter line is drawn in the same way. The intersection of the lateral ventricle posterior horn internal diameter line and the lateral ventricle contour mask is the final lateral ventricle posterior horn internal diameter line segment.

[0162] In other words, in this embodiment, during a fetal examination, the ultrasound physician uses deep learning to make real-time judgments based on the key structures of the fetal thalamus, cerebellum, and lateral ventricles in standard cross-sections. The standard cross-sections are automatically acquired and stored. A segmentation model outputs the contours of the thalamus, cerebellum, and lateral ventricles, as well as the positions of the biparietal diameter, cerebellar transverse diameter, and lateral ventricle posterior horn inner diameter. Measurements such as the length of the biparietal diameter segment, the cerebellar transverse diameter segment, the lateral ventricle posterior horn inner diameter segment, and the contour perimeter and area are automatically calculated. The measurement results are displayed on the ultrasound device interface, and abnormal results are alerted. The calculation of the biparietal diameter and cerebellar transverse diameter incorporates midline information to ensure correct directionality and prevent deviation, thus improving measurement accuracy. The physician can measure multiple fetal brain indicators with a single click, making the operation convenient and improving examination efficiency.

[0163] As can be seen, this application's embodiments identify a standard section of the fetal head in an ultrasound video, then input the standard section of the fetal head into an image segmentation model to obtain a segmentation result of the standard section of the fetal head. The segmentation result includes the contours of key structures within the fetal head and the midline of the brain. Finally, the measurement result of the standard section of the fetal head is determined based on the contours of the key structures and the midline of the brain. That is, this application first uses an image segmentation model to determine the contours of key structures within the fetal head and the midline of the brain, and then uses the midline of the brain to assist in automatic measurement, ensuring the accuracy of the direction of the measurement items in the standard section of the fetal head, and thus improving the accuracy of the measurement of the standard section of the fetal head.

[0164] See Figure 6 As shown in the figure, an embodiment of this application provides an ultrasound image processing device, comprising:

[0165] The standard section module 11 is used to identify the standard section of the fetal head in the ultrasound video, or to identify the standard section of the fetal head in real-time ultrasound mode.

[0166] Image segmentation module 12 is used to input the standard cross-section of the fetal head into the image segmentation model to obtain the segmentation result of the standard cross-section of the fetal head; wherein, the segmentation result includes the outline of key structures within the fetal head and the midline of the brain;

[0167] The image measurement module 13 is used to determine the measurement results of the standard section of the fetal head based on the key structural contour and the midline of the brain.

[0168] As can be seen, this application's embodiments identify a standard section of the fetal head in an ultrasound video, then input the standard section of the fetal head into an image segmentation model to obtain a segmentation result of the standard section of the fetal head. The segmentation result includes the contours of key structures within the fetal head and the midline of the brain. Finally, the measurement result of the standard section of the fetal head is determined based on the contours of the key structures and the midline of the brain. That is, this application first uses an image segmentation model to determine the contours of key structures within the fetal head and the midline of the brain, and then uses the midline of the brain to assist in automatic measurement, ensuring the accuracy of the direction of the measurement items in the standard section of the fetal head, and thus improving the accuracy of the measurement of the standard section of the fetal head.

[0169] The standard section of the fetal head includes a standard section of the thalamus and a standard section of the cerebellum; the key structural contours include the contours of the target thalamus and the target cerebellum; and the midline of the brain includes the first midline and the second midline. Correspondingly, the image measurement module 13 includes...

[0170] The thalamic standard section measurement submodule is used to determine the thalamic biparietal diameter measurement result of the thalamic standard section based on the target thalamic contour of the thalamic standard section and the first midline of the brain;

[0171] The cerebellar standard section measurement submodule is used to determine the cerebellar transverse diameter measurement result of the cerebellar standard section based on the target cerebellar contour of the cerebellar standard section and the second midline of the brain.

[0172] In one implementation, the thalamic standard section measurement submodule specifically includes:

[0173] A biparietal diameter line slope determination unit is used to determine the slope of the biparietal diameter line based on the first midline of the brain;

[0174] A bipolar diameter line determination unit is used to determine a bipolar diameter line based on the slope of the bipolar diameter line.

[0175] A biparietal diameter segment determination unit is used to determine a biparietal diameter segment based on the biparietal diameter straight line and the target thalamic contour.

[0176] The thalamic biparietal diameter measurement result determination unit is used to determine the thalamic biparietal diameter measurement result of the standard section of the thalamus based on the biparietal diameter line segment.

[0177] Furthermore, the biparietal diameter line determination unit is specifically used to determine the centroid of the target thalamus contour and determine the target pixel range based on the centroid; the biparietal diameter line determination unit determines multiple biparietal diameter lines based on the target pixel range and the slope of the biparietal diameter lines. Correspondingly, the biparietal diameter line segment determination unit is specifically used to determine multiple line segments based on the multiple biparietal diameter lines and the target thalamus contour; and the longest line segment among the multiple line segments is determined as the biparietal diameter line segment.

[0178] Furthermore, the thalamic biparietal diameter measurement result determination unit is specifically used to determine the contour of the inferior bony ring from the standard thalamic section, and to determine the lowest point of the inferior bony ring contour as the final bottom endpoint of the biparietal diameter line segment, thereby obtaining the final biparietal diameter line segment; and to determine the thalamic biparietal diameter measurement result of the standard thalamic section based on the final biparietal diameter line segment. Moreover, determining the contour of the inferior bony ring from the standard thalamic section includes: determining a first region range based on the bottom endpoint of the biparietal diameter line segment; extracting a first region image from the standard thalamic section based on the first region range; and determining the contour of the inferior bony ring from the first region image.

[0179] In one implementation, the cerebellar standard section measurement submodule is configured to include:

[0180] The cerebellar transverse diameter line slope determination unit is used to determine the slope of the cerebellar transverse diameter line based on the second brain midline.

[0181] The cerebellar transverse diameter line determination unit is used to determine the cerebellar transverse diameter line based on the slope of the cerebellar transverse diameter line.

[0182] The cerebellar transverse diameter segment determination unit is used to determine the cerebellar transverse diameter segment based on the cerebellar transverse diameter straight line and the target cerebellar contour.

[0183] The cerebellar transverse diameter measurement result determination unit is used to determine the cerebellar transverse diameter measurement result of the standard cerebellar section based on the cerebellar transverse diameter line segment.

[0184] Furthermore, the cerebellar transverse diameter measurement result determination unit is specifically used to determine the range of a second region and a range of a third region based on the two endpoints of the cerebellar transverse diameter line segment; to extract the second region image and the third region image from the standard cerebellar section based on the second region range and the third region range, respectively; to determine the local contour of the cerebellum in the second region image and the third region image, respectively; to determine the intersection point of the cerebellar transverse diameter line and the local contour of the cerebellum; to determine the final cerebellar transverse diameter line segment based on the intersection point; and to determine the cerebellar transverse diameter measurement result of the standard cerebellar section based on the final cerebellar transverse diameter line segment.

[0185] Furthermore, the standard section of the fetal head also includes a standard section of the lateral brain, and the key structural contour also includes the contour of the target lateral ventricle; correspondingly, the image measurement module 13 also includes:

[0186] The lateral brain standard section measurement submodule is used to determine the measurement result of the posterior horn diameter of the lateral ventricle of the lateral brain standard section based on the target lateral ventricle contour of the lateral brain standard section.

[0187] In one implementation, the lateral brain standard section measurement submodule specifically includes:

[0188] The choroid plexus location determination unit is used to determine the location information of the choroid plexus in the lateral ventricle based on the target lateral ventricle contour;

[0189] A unit for determining the straight line of the inner diameter of the posterior horn of the lateral ventricle is used to determine the straight line of the inner diameter of the posterior horn of the lateral ventricle based on the location information.

[0190] The lateral ventricle posterior horn inner diameter segment determination unit is used to determine the lateral ventricle posterior horn inner diameter segment based on the lateral ventricle inner diameter straight line and the target lateral ventricle contour.

[0191] The unit for determining the measurement result of the posterior horn diameter of the lateral ventricle is used to determine the measurement result of the posterior horn diameter of the lateral ventricle in the standard section of the lateral brain based on the line segment of the posterior horn diameter of the lateral ventricle.

[0192] The choroid plexus location determination unit is specifically used to determine a local lateral ventricle image from the standard cross-section of the lateral brain based on the target lateral ventricle contour; multiply the mask of the target lateral ventricle contour and the local lateral ventricle image to obtain the target lateral ventricle image; refine the mask of the target lateral ventricle contour to obtain a first midline; multiply the first midline and the target lateral ventricle image to obtain a second midline of the choroid plexus; perform linear fitting on the first midline and the second midline respectively to obtain a first line segment and a second line segment; and determine the location information of the choroid plexus in the lateral ventricle based on the first line segment and the second line segment.

[0193] In one embodiment, the training process of the image segmentation model includes: acquiring a first training sample set; the first training sample set includes standard cross-sectional samples of the fetal head and corresponding label information; the standard cross-sectional samples of the fetal head include standard cross-sectional samples of the thalamus, cerebellum, and lateral brain; the label information of the standard cross-sectional samples of the thalamus includes thalamic contour lines and the midline of the brain, the label information of the standard cross-sectional samples of the cerebellum includes cerebellar contour lines and the midline of the brain, and the label information of the standard cross-sectional samples of the lateral brain includes the lateral ventricle contour lines; inputting the standard cross-sectional samples of the fetal head into a first initial model to obtain the thalamic contour, cerebellar contour, midline of the brain, and lateral ventricle contour corresponding to the standard cross-sectional samples of the fetal head; calculating the training loss based on the thalamic contour, cerebellar contour, midline of the brain, lateral ventricle contour, and label information of the standard cross-sectional samples of the fetal head, and adjusting the parameters of the first initial model based on the training loss; if the training completion condition is detected, the first initial model after parameter adjustment is determined as an image segmentation model.

[0194] In one embodiment, the standard section module 11 is specifically used to identify the standard section of the fetal head in the ultrasound video using a standard section recognition model, or to identify the standard section of the fetal head in real-time ultrasound mode.

[0195] The standard section recognition model is obtained by training a second initial model using a second training sample set, which includes standard section samples of the fetal head, standard section samples of non-fetal head, and label information.

[0196] See Figure 7 As shown in the figure, this application discloses an ultrasound device 20, including a processor 21 and a memory 22; wherein, the memory 22 is used to store a computer program; the processor 21 is used to execute the computer program, the ultrasound image processing method disclosed in the foregoing embodiment.

[0197] For details regarding the specific process of the ultrasound image processing method described above, please refer to the relevant content disclosed in the foregoing embodiments, which will not be repeated here.

[0198] Furthermore, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk, or optical disk, and the storage method can be temporary storage or permanent storage.

[0199] In addition, the ultrasound device 20 also includes a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26; wherein, the power supply 23 is used to provide operating voltage for each hardware device on the ultrasound device 20; the communication interface 24 can create a data transmission channel between the ultrasound device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0200] Of course, in another embodiment, an electronic device can be provided, including a processor and a memory; wherein the memory is used to store a computer program; and the processor is used to execute the computer program, the ultrasound image processing method disclosed in the foregoing embodiments. This electronic device is connected to an ultrasound device, acquires ultrasound video through communication with the ultrasound device, and performs corresponding processing.

[0201] In other words, the computer program implemented by the method of this application can be embedded into the data stream of an existing ultrasound imaging system to directly retrieve the ultrasound video to be analyzed and processed from the existing ultrasound system data stream; or it can be placed in a computer with storage and computing functions as an independent system program, and obtain the ultrasound video to be processed by communicating with the ultrasound imaging system or connecting to the data cloud, and then perform independent reasoning and calculation.

[0202] Furthermore, embodiments of this application also disclose a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the ultrasound image processing method disclosed in the foregoing embodiments.

[0203] For details regarding the specific process of the ultrasound image processing method described above, please refer to the relevant content disclosed in the foregoing embodiments, which will not be repeated here.

[0204] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0205] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0206] The above provides a detailed description of an ultrasound image processing method, apparatus, device, and medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and its core ideas. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. An ultrasound image processing method, characterized in that, include: Identify the standard section of the fetal head in ultrasound video, or identify the standard section of the fetal head in real-time ultrasound mode; The standard cross-section of the fetal head is input into the image segmentation model to obtain the segmentation result of the standard cross-section of the fetal head; wherein, the segmentation result includes the outline of key structures within the fetal head and the midline of the brain; Measurement results of the standard section of the fetal head determined based on the key structural contour and the midline of the brain; The standard section of the fetal head includes a standard section of the thalamus, the key structural contour includes the target thalamic contour, and the midline of the brain includes the first midline of the brain. Correspondingly, the measurement result of determining the standard section of the fetal head based on the key structural contour and the midline of the brain includes: determining the slope of the biparietal diameter line based on the first midline of the brain and the perpendicular relationship between the first midline of the brain and the biparietal diameter line; determining the biparietal diameter line based on the slope of the biparietal diameter line; determining the biparietal diameter line segment based on the biparietal diameter line and the target thalamic contour; and determining the measurement result of the biparietal diameter of the thalamus in the standard section of the thalamus based on the biparietal diameter line segment.

2. The ultrasound image processing method according to claim 1, characterized in that, The standard section of the fetal head also includes a standard section of the cerebellum, the key structural contour also includes a target cerebellar contour, and the midline of the brain also includes a second midline of the brain; correspondingly, the measurement result of determining the standard section of the fetal head based on the key structural contour and the midline of the brain also includes: The cerebellar transverse diameter measurement result of the cerebellar standard section is determined based on the target cerebellar contour of the cerebellar standard section and the second brain midline.

3. The ultrasound image processing method according to claim 1, characterized in that, The process of determining the biparietal diameter line based on its slope, and determining the biparietal diameter line segment based on the biparietal diameter line and the target thalamic contour, includes: Determine the centroid of the target thalamus contour, and determine the target pixel range based on the centroid; Multiple bi-vertex lines are determined based on the target pixel range and the slope of the bi-vertex line; Multiple line segments are determined based on the multiple biparietal diameter straight lines and the target thalamic contour; The longest line segment among the plurality of line segments is determined as the bipolar line segment.

4. The ultrasound image processing method according to claim 1, characterized in that, The measurement results of the biparietal diameter of the thalamus based on the biparietal diameter line segment to determine the standard section of the thalamus include: The contour of the inferior bony ring is determined from the standard section of the thalamus, and the lowest point of the inferior bony ring contour is determined as the final bottom endpoint of the biparietal diameter line segment, thus obtaining the final biparietal diameter line segment. The biparietal diameter measurement result of the standard thalamic section is determined based on the final biparietal diameter segment.

5. The ultrasound image processing method according to claim 4, characterized in that, Determining the contour of the inferior osseous ring from the standard section of the thalamus includes: The first region is defined based on the bottom endpoint of the bipolar diameter line segment; Based on the first region range, an image of the first region is extracted from the standard section of the thalamus; The contour of the lower bone ring is determined from the first region image.

6. The ultrasound image processing method according to claim 2, characterized in that, The measurement results of the cerebellar transverse diameter of the standard cerebellar section, determined based on the target cerebellar contour and the second midline of the brain, include: The slope of the cerebellar transverse diameter line is determined based on the second midline of the brain; The cerebellar transverse diameter line is determined based on the slope of the cerebellar transverse diameter line. The cerebellar transverse diameter line segment is determined based on the cerebellar transverse diameter line and the target cerebellar contour. The cerebellar transverse diameter measurement results are determined based on the cerebellar transverse diameter segment.

7. The ultrasound image processing method according to claim 6, characterized in that, The measurement results of the cerebellar transverse diameter based on the cerebellar transverse diameter segment to determine the standard cerebellar section include: The extent of the second region and the extent of the third region are determined based on the two endpoints of the transverse cerebellar diameter segment, respectively. Images of the second and third regions are extracted from the standard cerebellar section based on the second and third region ranges, respectively. The local cerebellar contours in the second and third region images were determined respectively; Determine the intersection point of the cerebellar transverse diameter line and the local contour of the cerebellum; The final cerebellar transverse diameter segment is determined based on the intersection point; The cerebellar transverse diameter measurement result of the standard cerebellar section is determined based on the final cerebellar transverse diameter segment.

8. The ultrasound image processing method according to claim 1, characterized in that, The standard section of the fetal head also includes a standard section of the lateral brain, and the key structural contours also include the contour of the target lateral ventricle; correspondingly, the method further includes: The measurement results of the posterior horn diameter of the lateral ventricle in the standard lateral ventricle section are determined based on the target lateral ventricle contour of the standard lateral ventricle section.

9. The ultrasound image processing method according to claim 8, characterized in that, The determination of the posterior horn diameter of the lateral ventricle based on the target lateral ventricle contour of the standard lateral ventricle section includes: The location information of the choroid plexus in the lateral ventricle is determined based on the target lateral ventricle contour; Based on the location information, determine the straight line of the inner diameter of the posterior horn of the lateral ventricle; The inner diameter segment of the posterior horn of the lateral ventricle is determined based on the straight line of the lateral ventricle diameter and the contour of the target lateral ventricle. The measurement result of the posterior horn diameter of the lateral ventricle in the standard section of the lateral brain is determined based on the line segment of the posterior horn diameter of the lateral ventricle.

10. The ultrasound image processing method according to claim 9, characterized in that, The determination of the choroid plexus's position within the lateral ventricle based on the target lateral ventricle contour includes: A local lateral ventricle image is determined from the standard cross-section of the lateral brain based on the target lateral ventricle contour; The target lateral ventricle image is obtained by multiplying the mask of the target lateral ventricle outline with the local lateral ventricle image. The mask of the target lateral ventricle contour is refined to obtain the first midline. Multiply the first midline and the target lateral ventricle image to obtain the second midline of the choroid plexus; By fitting the first central axis and the second central axis with straight lines respectively, a first line segment and a second line segment are obtained; The location information of the choroid plexus in the lateral ventricle is determined based on the first line segment and the second line segment.

11. The ultrasound image processing method according to claim 1, characterized in that, The training process of the image segmentation model includes: Obtain a first training sample set; the first training sample set includes standard cross-sectional samples of the fetal head and corresponding label information for the standard cross-sectional samples of the fetal head; the standard cross-sectional samples of the fetal head include standard cross-sectional samples of the thalamus, standard cross-sectional samples of the cerebellum, and standard cross-sectional samples of the lateral brain; the label information for the standard cross-sectional samples of the thalamus includes the thalamic contour line and the midline of the brain, the label information for the standard cross-sectional samples of the cerebellum includes the cerebellar contour line and the midline of the brain, and the label information for the standard cross-sectional samples of the lateral brain includes the lateral ventricle contour line; Input the standard cross-sectional sample of the fetal head into the first initial model to obtain the thalamic contour, cerebellar contour, midline of the brain and lateral ventricle contour corresponding to the standard cross-sectional sample of the fetal head. The training loss is calculated based on the label information of the thalamic contour, the cerebellar contour, the midline of the brain, the lateral ventricle contour, and the standard cross-sectional sample of the fetal head, and the parameters of the first initial model are adjusted based on the training loss. If the training completion condition is met, the first initial model with adjusted parameters is determined as the image segmentation model.

12. The ultrasound image processing method according to any one of claims 1 to 11, characterized in that, The identification of the standard section of the fetal head in ultrasound video, or the identification of the standard section of the fetal head in real-time ultrasound mode, includes: Using a standard section recognition model, the standard section of the fetal head in ultrasound video can be identified, or the standard section of the fetal head can be identified in real-time ultrasound mode. The standard section recognition model is obtained by training a second initial model using a second training sample set, which includes standard section samples of the fetal head, standard section samples of non-fetal head, and label information.

13. An ultrasonic image processing device, characterized in that, include: The standard section module is used to identify the standard section of the fetal head in ultrasound video, or to identify the standard section of the fetal head in real-time ultrasound mode. An image segmentation module is used to input the standard cross-section of the fetal head into an image segmentation model to obtain the segmentation result of the standard cross-section of the fetal head; wherein, the segmentation result includes the contours of key structures within the fetal head and the midline of the brain; wherein, the standard cross-section of the fetal head includes a standard cross-section of the thalamus, the contours of key structures include the contours of the target thalamus, and the midline of the brain includes the first midline of the brain. An image measurement module is used to determine the measurement results of the standard section of the fetal head based on the key structural contour and the midline of the brain; Specifically, the image measurement module is used to: determine the slope of the biparietal diameter line based on the first midline of the brain and the vertical relationship between the first midline of the brain and the biparietal diameter line; determine the biparietal diameter line based on the slope of the biparietal diameter line; determine the biparietal diameter line segment based on the biparietal diameter line and the target thalamic contour; and determine the thalamic biparietal diameter measurement result of the standard thalamic section based on the biparietal diameter line segment.

14. An ultrasonic device, characterized in that, Includes processor and memory; among which, The memory is used to store computer programs; The processor is configured to execute the computer program to implement the ultrasound image processing method as described in any one of claims 1 to 12.

15. A computer-readable storage medium, characterized in that, Used to store a computer program, wherein the computer program, when executed by a processor, implements the ultrasound image processing method as described in any one of claims 1 to 12.

Images