A method of automatic measurement of prenatal ultrasound images

Abstract

The application discloses an automatic measurement method of prenatal ultrasound images, and relates to an automatic measurement method of prenatal ultrasound images, a storage medium and an ultrasound device.The method comprises the following steps: acquiring an ultrasound image to be measured, and extracting all target objects to be measured carried by the ultrasound image to be measured; measuring all biomasses of the extracted target objects to be measured, and determining measurement results of the target objects to be measured according to the measured biomasses.The measurement results of the target objects are automatically measured, the workload of obtaining the measurement results is reduced, and the work efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of prenatal ultrasound image processing, more particularly, it relates to an automatic measurement method of prenatal ultrasound image. BACKGROUND

[0002] Prenatal ultrasound diagnosis is one of the important means to find congenital defects, and ultrasound diagnosis has the advantages of no radiation, convenience and real-time, and is widely used in clinical detection of prenatal fetuses. At present, in the parameter measurement process of conventional ultrasound images, the doctor needs to rely on rich clinical experience to operate the equipment to find the biological quantity to be measured in the ultrasound image, and then manually draw a measurement line and finally calculate the numerical value by the system. Therefore, in the traditional parameter measurement process of ultrasound images, the doctor needs to perform a large amount of tedious manual operation, which seriously affects the diagnosis efficiency of the doctor, especially when a frame of ultrasound image contains multiple biological quantities to be measured, which is more serious. In addition, since the operation habits of different doctors are different, certain measurement errors may be introduced artificially during the measurement process and affect the final measurement result. In addition, fetal congenital heart disease is the most common and most serious congenital malformation in the world, and is also the first factor of neonatal death. Although there has been great progress in the diagnosis and management of fetal congenital heart disease, fetal congenital heart disease is still the most common cause of death in the first year of life. Therefore, effective prenatal diagnosis and prevention and control will affect the outcome of neonates with fetal congenital heart disease, and reduce the birth defects and mortality. In actual clinical medicine, the chest transverse section image of the fetus at the four-chamber heart level is generally obtained by ultrasound, and then the cardiac axis of the fetus is obtained by manual tracing measurement to assist in screening and diagnosis of fetal congenital heart disease. However, the above-mentioned measurement method of cardiac axis is easily affected by the clinical experience and other subjective factors of the doctor, resulting in large data error of manual measurement, and causing missed diagnosis or misdiagnosis. SUMMARY

[0003] (I) Technical problems solved

[0004] In view of the problems in the prior art, the present application provides an automatic measurement method of prenatal ultrasound image to solve the technical problems mentioned in the background.

[0005] (II) Technical solutions

[0006] To achieve the above object, the present application provides the following technical scheme: an automatic measurement method of prenatal ultrasound image, comprising the following steps:

[0007] Step one: obtain an ultrasound image to be measured, extract all target objects to be measured carried by the ultrasound image to be measured, obtain a fetal heart ultrasound four-chamber heart section image to be measured, obtain a target segmentation image based on a preset target segmentation model, and obtain atrial septum contour data, interventricular septum contour data, spine contour data, and thoracic cavity contour data by performing target segmentation processing on the target four-chamber heart ultrasound section image;

[0008] Step two: measure all biomasses of all target objects to be measured extracted, determine the measurement results of each target object to be measured according to all measured biomasses, input the target segmentation image into a trained segmentation model to obtain a segmentation result image; the segmentation result image includes a segmentation region of each structure of the fetal heart, and an axis is fitted based on the atrial septum contour data and the interventricular septum contour data to obtain a first heart axis.

[0009] Step three: the trained segmentation model is obtained by training a preset segmentation model, and the measurement of all biomasses of all target objects to be measured extracted and the determination of the measurement results of each target object to be measured according to all measured biomasses specifically include: measuring all biomasses of all target objects to be measured extracted.

[0010] Step four: for each target object to be measured, the object type of the target object to be measured is determined; the biomasses of the target object to be measured are screened from all biomasses according to the object type to obtain the measurement results of the target object to be measured; the object type of the target object to be measured is determined according to a preset neural network, and when the preset neural network fails, the measurement results of the target object to be measured are re-determined according to all obtained biomasses.

[0011] The application further provides that after extracting all target objects to be measured carried by the ultrasound image to be measured, the number of extracted target objects to be measured is obtained; when the number is greater than or equal to 1, the step of measuring each target object to be measured is performed; when the number is less than 1, it is prompted that no target object to be measured is recognized, an axis is fitted based on the spine contour data and the thoracic cavity contour data to obtain a second heart axis, and a heart axis is obtained according to the included angle between the first heart axis and the second heart axis.

[0012] The application further provides that the size of each segmentation region in the segmentation result image is measured respectively, including: determining the outer boundary information of each segmentation region in the segmentation result image respectively; determining the minimum circumscribed rectangle of each segmentation region according to the outer boundary information; and measuring the minimum circumscribed rectangle of each segmentation region to obtain the size of each segmentation region.

[0013] The application is further configured to, after measuring all biomasses of all target objects to be measured and determining the measurement results of each target object to be measured according to all measured biomasses, further comprising: for each target object to be measured, drawing a measurement line corresponding to the target object to be measured according to the updated measurement result of the target object to be measured, and drawing the measurement line in the ultrasound image.

[0014] The application is further configured to, the measurement result of the biomass includes position information of the biomass and structure information of the biomass.

[0015] The application is further configured to, the acquisition of the ultrasound image to be measured specifically includes: acquiring a plurality of continuous ultrasound image frames; screening the plurality of continuous ultrasound image frames to obtain the ultrasound image to be measured.

[0016] The application is further configured to, the extraction of all target objects to be measured carried by the ultrasound image to be measured specifically includes: inputting the ultrasound image to be measured into a preset recognition model, wherein the preset recognition model is trained based on a training sample set, each training sample in the training sample set includes an ultrasound image and all target objects to be measured carried by the ultrasound image; outputting all target objects to be measured carried by the ultrasound image to be measured by the preset recognition model.

[0017] The application is further configured to, the determination of the outer boundary information of each segmentation region in the segmentation result image respectively includes: denoising the segmentation result image to obtain a denoised segmentation result image; performing edge detection on the denoised segmentation result image to obtain an edge detection image; and binarizing the edge detection image to obtain the outer boundary information of each segmentation region.

[0018] (Three) beneficial effects

[0019] Compared with the prior art, the application provides an automatic measurement method for prenatal ultrasound images, which has the following beneficial effects:

[0020] The application provides a prenatal ultrasound image automatic measurement method, a storage medium and an ultrasound device, the method comprises the following steps: acquiring an ultrasound image to be measured, and extracting all target objects to be measured carried by the ultrasound image to be measured; measuring all biomasses of all the extracted target objects to be measured, and determining the measurement results of each target object to be measured according to the measured biomasses. The measurement results of the target objects are automatically measured, the workload of obtaining the measurement results is reduced, and the work efficiency is improved. On the other hand, the measurement errors or failures of the measurement results caused by the wrong object type judgment can be avoided, so that the accuracy of obtaining the measurement results of the target objects is improved. The application can realize accurate segmentation of each structure of a fetal heart in a to-be-segmented image, so that the accuracy of the segmentation results is higher. The size of each segmented region in the segmentation result image is measured, automatic measurement is realized, so that the problem of low accuracy of the measurement results caused by the complicated measurement process and the non-standardization in the prior art is avoided. Furthermore, the application can ensure the accuracy of the measurement results, so that experts only need to diagnose according to the measurement results, and the application provides conditions for the ability of the experts to benefit more patients and regions, and the prenatal detection rate of fetal congenital heart disease is improved. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 The application discloses a prenatal ultrasound image automatic measurement method. DETAILED DESCRIPTION

[0022] Embodiment 1

[0023] As shown in the drawings, the application discloses a prenatal ultrasound image automatic measurement method, which comprises the following steps: Figure 1

[0024] Step one: acquiring an ultrasound image to be measured, extracting all target objects to be measured carried by the ultrasound image to be measured, acquiring a fetal heart ultrasound four-chamber heart section image to be measured, obtaining a to-be-segmented image, and performing target segmentation processing on the target four-chamber heart ultrasound section image based on a preset target segmentation model to obtain atrial septum contour data, interventricular septum contour data, spine contour data and thoracic cavity contour data.

[0025] Step two: measuring all biomasses of all the extracted target objects to be measured, and determining the measurement results of each target object to be measured according to the measured biomasses. The to-be-segmented image is input into a segmented model that is completed training to obtain a segmentation result image. The segmentation result image comprises segmented regions of each structure of a fetal heart. An axis is fitted based on the atrial septum contour data and the interventricular septum contour data to obtain a first heart axis.

[0026] ​Step three: the completed training segmentation model is obtained by training a preset segmentation model, and all biomasses of all target objects to be measured are measured, and the measurement results of each target object to be measured are determined according to all biomasses measured, which specifically includes: measuring all biomasses of all target objects to be measured extracted;

[0027] Step four: for each target object to be measured, the object type of the target object to be measured is determined, and the biomasses of the target object to be measured are screened from all biomasses according to the object type to obtain the measurement result of the target object to be measured; the object type of the target object to be measured is determined according to a preset neural network, and when the preset neural network fails, the measurement result of the target object to be measured is re-determined according to all biomasses obtained.

[0028] In further embodiments of the application, after extracting all target objects to be measured carried by the ultrasound image to be measured, the number of target objects to be measured extracted is obtained; when the number is greater than or equal to 1, the step of measuring each target object to be measured is performed; when the number is less than 1, it is prompted that no target object to be measured is identified, the second heart axis is obtained by performing axis fitting based on the spine contour data and the chest contour data, and the heart axis is obtained according to the included angle between the first heart axis and the second heart axis.

[0029] The application further provides that the size of each segmentation region in the segmentation result image is measured respectively, including: determining the outer boundary information of each segmentation region in the segmentation result image respectively; determining the minimum circumscribed rectangle of each segmentation region according to the outer boundary information; and measuring the minimum circumscribed rectangle of each segmentation region to obtain the size of each segmentation region.

[0030] Embodiment 2:

[0031] An automatic measurement method of prenatal ultrasound images, comprising the following steps:

[0032] Step one: obtaining an ultrasound image to be measured, extracting all target objects to be measured carried by the ultrasound image to be measured, obtaining a fetal heart ultrasound four-chamber heart section image to be measured, obtaining a target segmentation image, and performing target segmentation processing on the target four-chamber heart ultrasound section image based on a preset target segmentation model to obtain atrial septum contour data, interventricular septum contour data, spine contour data and chest contour data;

[0033] Step two: measure all biomasses of all target objects to be measured extracted, and determine the measurement results of each target object to be measured according to all measured biomasses, input the segmented image into the completed training segmentation model to obtain a segmentation result image; the segmentation result image includes segmented areas of each structure of the fetal heart, and the first heart axis is obtained by fitting the axis based on the atrial septum contour data and the interventricular septum contour data;

[0034] Step three: the completed training segmentation model is obtained by training a preset segmentation model, and the measurement of all biomasses of all target objects to be measured extracted and the determination of the measurement results of each target object to be measured according to all measured biomasses specifically include: measuring all biomasses of all target objects to be measured extracted; for each target object to be measured, determining the object type of the target object to be measured; filtering the biomasses of the target object to be measured from all biomasses according to the object type to obtain the measurement results of the target object to be measured; the object type of the target object to be measured is determined according to a preset neural network, and when the preset neural network fails, the measurement results of the target object to be measured are re-determined according to all obtained biomasses.

[0035] In further embodiments of the application, after measuring all biomasses of all target objects to be measured extracted and determining the measurement results of each target object to be measured according to all measured biomasses, it further includes: for each target object to be measured, drawing a corresponding measurement line according to the updated measurement results of the target object to be measured, and drawing the measurement line in the ultrasound image; the measurement results of the biomasses include position information and structure information of the biomasses; the ultrasound image to be measured is obtained by: obtaining a plurality of continuous ultrasound image frames; filtering the plurality of continuous ultrasound image frames to obtain the ultrasound image to be measured; the extraction of all target objects to be measured carried by the ultrasound image to be measured specifically includes: inputting the ultrasound image to be measured into a preset recognition model, wherein the preset recognition model is trained based on a training sample set, and each training sample in the training sample set includes an ultrasound image and all target objects to be measured carried by the ultrasound image; outputting all target objects to be measured carried by the ultrasound image to be measured by the preset recognition model; the determination of the outer boundary information of each segmentation area in the segmentation result image includes: denoising the segmentation result image to obtain a denoised segmentation result image; edge detection is performed on the denoised segmentation result image to obtain an edge detection image; the edge detection image is binarized to obtain the outer boundary information of each segmentation area.

[0036] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but cannot be listed as the limitation of the patent scope of the present application. It should be pointed out that, for ordinary skilled in the art, several modifications and improvements can be made without departing from the concept of the present application, which all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A method of automatic measurement of prenatal ultrasound images, characterized in that, The method comprises the following steps: Step one: obtaining an ultrasound image to be measured, extracting all target objects carried by the ultrasound image to be measured, obtaining a fetal heart ultrasound four-chamber heart section image to be measured, obtaining a target segmentation image based on a preset target segmentation model, and obtaining atrial septum contour data, interventricular septum contour data, spine contour data, and thoracic cavity contour data; Step two: measuring all biomasses of all extracted target objects to be measured, determining the measurement results of each target object to be measured according to the measured biomasses, inputting the target segmentation image into a completed training segmentation model to obtain a segmentation result image; the segmentation result image comprises a segmentation region of each structure of the fetal heart, an axis is fitted based on the atrial septum contour data and the interventricular septum contour data to obtain a first heart axis; the size of each segmentation region in the segmentation result image is further measured, which further comprises: determining the outer boundary information of each segmentation region in the segmentation result image respectively; determining the minimum circumscribed rectangle of each segmentation region according to the outer boundary information; and measuring the minimum circumscribed rectangle of each segmentation region to obtain the size of each segmentation region; Step three: the completed training segmentation model is obtained by training a preset target segmentation model; the measurement of all biomasses of all extracted target objects to be measured and the determination of the measurement results of each target object to be measured according to the measured biomasses specifically comprise: measuring all biomasses of all extracted target objects to be measured; Step four: for each target object to be measured, the object type of the target object to be measured is determined; the biomasses of the target object to be measured are screened from all biomasses according to the object type to obtain the measurement results of the target object to be measured; the object type of the target object to be measured is determined according to a preset neural network; when the preset neural network is incorrect, the measurement results of the target object to be measured are re-determined according to all obtained biomasses.

2. A method of automatic measurement of prenatal ultrasound images according to claim 1, characterized in that: After extracting all target objects carried by the ultrasound image to be measured, the method further comprises: obtaining the number of extracted target objects to be measured; when the number is greater than or equal to 1, the step of measuring each target object to be measured is performed; when the number is less than 1, it is prompted that no target object to be measured is identified, an axis is fitted based on the spine contour data and the thoracic cavity contour data to obtain a second heart axis; and a heart axis is obtained according to the included angle between the first heart axis and the second heart axis.

3. The method of claim 1, wherein: After measuring all biomasses of all extracted target objects to be measured and determining the measurement results of each target object to be measured according to the measured biomasses, the method further comprises: for each target object to be measured, a corresponding measurement line is drawn according to the updated measurement results of the target object to be measured, and the measurement line is drawn in the ultrasound image.

4. The method of claim 1, wherein: The measurement results of the biomasses comprise position information and structure information of the biomasses.

5. A method of automatic measurement of prenatal ultrasound images according to claim 4, characterized in that: The acquiring the to-be-measured ultrasound image specifically comprises: acquiring a plurality of continuous ultrasound image frames; and screening the plurality of continuous ultrasound image frames to obtain the to-be-measured ultrasound image.

6. A method of automatic measurement of prenatal ultrasound images according to claim 5, characterized in that: The extracting all to-be-measured target objects carried by the to-be-measured ultrasound image specifically comprises: inputting the to-be-measured ultrasound image into a preset identification model, wherein the preset identification model is trained based on a training sample set, each training sample in the training sample set comprises an ultrasound image and all to-be-measured target objects carried by the ultrasound image; and outputting, by the preset identification model, all to-be-measured target objects carried by the to-be-measured ultrasound image.

7. A method of automatic measurement of prenatal ultrasound images according to claim 3, characterized in that: The determining the outer boundary information of each segmentation region in the segmentation result image specifically comprises: denoising the segmentation result image to obtain a denoised segmentation result image; performing edge detection on the denoised segmentation result image to obtain an edge detection image; and performing binarization on the edge detection image to obtain the outer boundary information of each segmentation region.

Images