Quality evaluation and biological measurement method and system for cross section of cerebellar

Abstract

The invention discloses a transcerebellar cross section quality evaluation and biological measurement method and system, and belongs to the technical field of medical image analysis and processing. The method comprises the following steps: acquiring a transcalar cross section sample in a fetal ultrasound image, performing data preprocessing, constructing a target detection and semantic segmentation data set, training to obtain a target detection and semantic segmentation network as a prediction model, and constructing a corresponding quality evaluation system and a biological index post-processing system according to a result output by the prediction model. And the prediction model and an evaluation processing system are effectively combined together, so that evaluation and measurement of the end-to-end transcerebellar cross section ultrasound image are realized. According to the method, section quality evaluation and biological index measurement are combined on the cross section of the cerebellum for the first time, a scientific and objective quantitative evaluation system and a rapid and accurate biological index measurement method are established, and a doctor can conveniently judge the section quality.

Description

Technical Field

[0001] The invention relates to a method and system for evaluating the quality of a cerebellar cross section and for measuring biological measurements, and belongs to the technical field of medical image analysis and processing. Background Art

[0002] With the rapid development of deep learning, there are more and more methods for analyzing and processing medical images. The quality assessment of ultrasound images is not only an important prerequisite for the construction of medical image datasets, but also plays a vital role in helping doctors' diagnosis. Traditional medical image quality assessment requires rich clinical experience and a large amount of human resources, which is a great challenge for some areas with insufficient medical resources. Medical images that lack the correct quality assessment process are often prone to misjudgment of patients' symptoms, resulting in immeasurable consequences. In fetal ultrasound images, the cerebellar cross-section is an important basis for doctors to judge the development of the fetal cerebellum. How to accurately and reasonably evaluate the sections collected by the instrument plays a vital role in the subsequent diagnosis of doctors. It is very necessary to propose a scientific and objective section quality assessment method.

[0003] Deep neural network is the most popular image processing technology now. It has a good processing effect on medical images with unclear features, high noise and complex structure. Among them, the target detection task is an important research direction in the field of computer vision in deep learning. Its purpose is to automatically identify and locate the target or lesion area of ​​interest in medical images. This task has a wide range of applications in medical image analysis, especially in disease screening, diagnosis and treatment planning. The target area tissue in medical images is often slightly different from the surrounding tissue, so the target detection model needs to have very high accuracy to ensure reliable diagnosis results. Medical image annotation usually needs to be performed by professional medical personnel, and the annotation process is time-consuming and labor-intensive, so the annotated data is usually small and the data volume is limited. However, the target detection task in medical images can greatly assist clinicians in their diagnosis work, especially for clinicians with less experience, and can improve their work efficiency and accuracy. With the advancement of deep learning technology, this field is still developing. However, the results of target detection alone cannot directly provide the quantitative evaluation results of the section. How to establish an objective and scientific evaluation system to link the results of target detection with the quality of the transcerebellar cross-section in the fetal ultrasound image to achieve end-to-end section quality evaluation is also the focus of scholars' research. At the same time, on this basis, establishing a set of fetal biological index measurement methods based on section quality evaluation results is also very critical to assist doctors in diagnosis.

[0004] Therefore, there is a need for a method that can perform quality assessment on transcerebellar cross-sections in fetal ultrasound images and perform segmentation and biomarker measurement based on the assessment results. Summary of the invention

[0005] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a method for evaluating the quality of a transcerebellar cross section and measuring biological indicators, thereby establishing an evaluation system by connecting the section quality evaluation with the biological indicator measurement in series, thereby enhancing the scientificity and interpretability of the quality evaluation and the biological indicator measurement; In order to achieve the above-mentioned purpose / solve the above-mentioned technical problem, the present invention is implemented by adopting the following technical scheme:.

[0006] A first aspect: A method for assessing and biometrically evaluating cerebellar cross-section quality, the method comprising: Inputting a cerebellar cross-section image in the fetal ultrasound image to be tested into a pre-trained target detection network model, and outputting an ultrasound image with a target detection frame and a corresponding confidence level, wherein the target detection network model is based on a YOLOv11 model structure improved from YOLOv5 and YOLOv8; The ultrasound images with the target detection frame and the corresponding confidence level are passed through the quality assessment system to evaluate the ultrasound images with the section quality meeting the standard; Inputting the ultrasound image with qualified section quality into a pre-trained segmentation model to obtain the segmentation mask of the cerebellum and the cistern magnum, wherein the objective function of the segmentation model adopts a combination of Dice loss and cross entropy loss; The segmentation masks of the cerebellum and cisterna magna were processed through a post-processing system to obtain two key biometric indicators through the cerebellar cross-section.

[0007] Optionally, the training method of the target detection network model includes: Obtain transcerebellar cross-section samples from fetal ultrasound images and construct a target detection dataset; Perform data preprocessing on target detection dataset; Based on the preprocessed target detection data set, the target is to minimize the value of the objective function, and the trained target detection network model is obtained according to the final weight parameters of the encoder and the decoder.

[0008] Optionally, acquiring a transcerebellar cross-section sample in a fetal ultrasound image and constructing a target detection data set includes: According to professional books and expert opinions, the key tissues in the transverse section of the cerebellum were studied, and the five target tissues required for the section were obtained: the strong echo ring of the skull, the cavity of the septum pellucidum, the thalamus, the cerebellum, and the cistern magnum; A target box is drawn in the cross-section sample according to the target tissue, and the original image is matched one-to-one with the image containing the target box to construct a target detection dataset, wherein the target detection dataset is used as a training set.

[0009] Optionally, the target detection network model includes an input layer for data preprocessing and data enhancement, a backbone network for feature extraction, an intermediate layer composed of a convolutional neural network and fusing multi-level information to help the model identify targets of different scales, and an output layer for predicting target bounding boxes, classifications, and confidence levels.

[0010] Optionally, the quality assessment system is constructed based on the confidence results of each part of the target detection result obtained by processing the cerebellar cross-section by the target detection network model, and combined with medical prior knowledge to establish a section scoring mechanism.

[0011] Optionally, the section scoring mechanism is as follows: (1) In each slice, there is one and only one target detection result for each target tissue; (2) The relative positions of different target tissues are fixed. The strong echo ring of the skull needs to include the cavum septum pellucidum, thalamus, cerebellum and cistern magnum. The four tissues are arranged in sequence along the midline of the brain and maintain a symmetrical relationship with the midline of the brain. The final section score is obtained based on the confidence level of the five target tissues, and a quantitative evaluation is performed on the cross-section through the cerebellum.

[0012] Optionally, the section scoring mechanism uses a 10-point scoring system, where the initial score of each target tissue is two points, which is then multiplied by the confidence of the target box of the corresponding tissue in the target detection result to obtain the final score of the tissue, and then the scores of the five target tissues are summed to obtain the quantitative result of the quality assessment of the section.

[0013] Optionally, the post-processing system is established by combining the segmentation model with the segmentation mask of the cerebellum and the cistern magnum in the segmentation result obtained by processing the cerebellum cross-section and medical prior knowledge.

[0014] Optionally, the steps for establishing a post-processing system are as follows: (1) Based on the segmentation mask of the cerebellum, the convex hull of the region is first calculated using an algorithm, and the distance between the farthest point pairs on the convex hull is calculated using the rotating calculus method as the measurement value of the transverse diameter of the cerebellum; (2) According to the segmentation mask of the cerebellum, the corresponding ellipse is fitted, and the short axis direction of the ellipse is used as a straight line. Then, according to the segmentation mask of the cistern magna, the distance between the intersection of the extended short axis of the ellipse and the segmented area of ​​the cistern magna is obtained, which is used as the width of the cistern magna. The two key biometric indicators, the transverse diameter of the cerebellum through the cross section of the cerebellum and the width of the cistern magna, are obtained.

[0015] A second aspect: A transcerebellar cross-section quality assessment and biometry system, the system comprising: The target detection module is configured to input the cerebellum cross-section image in the fetal ultrasound image to be tested into a pre-trained target detection network model, and output an ultrasound image with a target detection frame and a corresponding confidence, wherein the target detection network model is based on a YOLOv11 model structure improved from YOLOv5 and YOLOv8; The quality assessment module is configured to pass the ultrasound image with the target detection frame and the corresponding confidence through the quality assessment system to assess the ultrasound image with the section quality meeting the standard; A semantic segmentation module is configured to input the ultrasound image with qualified section quality into a pre-trained segmentation model to obtain a segmentation mask of the cerebellum and the cistern magna, wherein the objective function of the segmentation model adopts a combination of Dice loss and cross entropy loss; The biometric module is configured to process the segmentation masks of the cerebellum and the cisterna magna through a post-processing system to obtain two key biometric indicators through a cross-section of the cerebellum.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention combines the section quality assessment with the biological index measurement on the cerebellar cross section for the first time, establishes a scientific and objective quantitative evaluation system and a fast and accurate biological index measurement method, which facilitates doctors to judge the section quality. It has the characteristics of high precision and high efficiency, and well solves the difficulties of large noise and complex structure in medical images, providing reliable results for subsequent quality assessment. It has the characteristics of high adaptability and can be easily expanded to other medical ultrasound image quality assessment and biological index measurement tasks; Compared with the prior art, the target detection model and semantic segmentation model adopted by the present invention have the characteristics of high precision and high efficiency, which well solves the difficulties of large noise and complex structure in medical images, and provides reliable results for subsequent quality evaluation; The present invention has the characteristics of high adaptability, can effectively solve the quality assessment problem of the cerebellar cross section in fetal ultrasound images, and can be easily extended to other quality assessment tasks of medical ultrasound images. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Shown is a flow chart of the cerebellar cross-section quality assessment and biometry method of the present invention; Figure 2 Shown is a schematic diagram of the method flow in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0019] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0020] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood by specific circumstances. Example

[0021] See also Figure 1-Figure 2 A method for evaluating the quality of cerebellar cross-sections and biometrics can be developed by combining deep learning with object detection and semantic segmentation techniques to establish a set of scientific and objective quality evaluation annotations, and segment and post-process based on the evaluation results to help doctors evaluate the sections and measure biometric indicators, thereby assisting doctors in their subsequent diagnosis. The steps are as follows: Step 1: obtain transcerebellar cross-section samples in fetal ultrasound images, construct a target detection data set, perform data preprocessing, select a deep neural network model and objective function for target detection, and obtain a trained target detection network model.

[0022] Step 2, inputting the cerebellum cross-section image in the fetal ultrasound image to be tested into a pre-trained target detection network model, and outputting an ultrasound image with a target detection frame and a corresponding confidence level, wherein the target detection network model is based on a YOLOv11 model structure improved from YOLOv5 and YOLOv8; Step 3: The ultrasound image with the target detection frame and the corresponding confidence level is passed through a quality assessment system to assess the ultrasound image with the section quality meeting the standard; Step 4, taking samples of the cerebellar cross-section in the fetal ultrasound image, constructing a semantic segmentation data set, performing data preprocessing, selecting a deep neural network model and objective function for semantic segmentation, and obtaining a trained segmentation model; Step 5, inputting the ultrasound image with qualified section quality into a pre-trained segmentation model to obtain the segmentation mask of the cerebellum and the cistern magnum, wherein the objective function of the segmentation model adopts a combination of Dice loss and cross entropy loss; In step 6, the segmentation masks of the cerebellum and the cistern magnum are processed by a post-processing system to obtain two key biometric indicators through the cross-section of the cerebellum.

[0023] In the specific implementation process of this embodiment, step 1: select a certain proportion of transcerebellar cross-section samples from fetal ultrasound images. First, based on medical prior knowledge, target boxes are marked and drawn for key tissues (strong echo rings of the skull, cavum septum pellucidum, thalamus, cerebellum, and cistern magnum) in this batch of samples. If the target tissue is incomplete or very blurred on the section, there is no need to draw the corresponding target box. After processing, the pre-processed training data is obtained.

[0024] The target detection model is constructed. The overall model structure is the YOLOv11 model structure improved based on YOLOv5 and YOLOv8. Its core structure can be divided into the following parts: an input layer for data preprocessing and data enhancement; a backbone network for feature extraction, which is composed of a convolutional neural network; an intermediate layer that further integrates multi-level information to help the model identify targets of different scales; and an output layer that predicts target bounding boxes, classifications, and confidence levels.

[0025] The objective function is a combined loss function that aims to optimize three key tasks: bounding box regression, classification, and confidence prediction. Through the optimization of these structures and objective functions, YOLOv11 has become an advanced model with both high accuracy and high efficiency in current object detection tasks.

[0026] The training method for the target detection network model described in step 1 includes: Obtain transcerebellar cross-section samples from fetal ultrasound images and construct a target detection dataset; Perform data preprocessing on the target detection dataset; Based on the preprocessed target detection data set, the target is to minimize the value of the objective function, and the trained target detection network model is obtained according to the final weight parameters of the encoder and the decoder.

[0027] The data preprocessing process of the target detection dataset is as follows: First, the key tissues in the cerebellar cross section are studied according to professional books and expert opinions, and it is concluded that the tissues required by the cross section are mainly the strong echo ring of the skull, the cavity of the septum pellucidum, the thalamus, the cerebellum and the occipital cistern. Then, the target box is drawn in the cross section sample according to the target tissue, and the original image is matched with the image containing the target box one by one to construct the training set.

[0028] In this embodiment, for step 2, based on the data set, network structure and objective function of step 1, the label feature extraction network is trained. The training process aims to minimize the value of the objective function, and the trained network model is obtained according to the final weight parameters of the discriminator and the generator. The trained network model is used as the final label feature extraction model; In the specific implementation process of this embodiment, step 3 is further described: based on the pre-trained model obtained in step 2, the target detection result through the cerebellum cross section is obtained, and the target tissue is first judged according to the following principles: (1) In each slice, there is one and only one target detection result for each target tissue; (2) The relative positions of different target tissues are fixed. The skull echo ring needs to include the other four target tissues. The cavum septum pellucidum, thalamus, cerebellum and cistern magnum need to be arranged in sequence along the midline of the brain and maintain a symmetrical relationship with the midline of the brain. If the target tissue meets the principle, the section quality is further quantitatively evaluated and scored on a 10-point scale. The initial score of each target tissue is two points, which is then multiplied by the confidence of the target box of the corresponding tissue in the target detection result to obtain the final score of the tissue. Finally, the scores of the five target tissues are summed to obtain the quantitative result of the quality assessment of the section.

[0029] In the specific implementation process of this embodiment, step 4 is further explained: take the cerebellar cross-section sample in the fetal ultrasound image, construct a semantic segmentation data set, perform data preprocessing, and study the key tissues in the cerebellar cross-section according to professional books and expert opinions. It is concluded that the tissues required for biological indicator measurement are mainly the cerebellum and the occipital cistern. Then segment the section samples according to the target tissue and construct a training set. The deep neural network model and objective function for semantic segmentation are selected. The overall network structure continues the classic U-Net framework, including encoders (downsampling) and decoders (upsampling), as well as jump connections. The objective function adopts a combination of Dice loss and cross entropy loss.

[0030] In the specific implementation process of this embodiment, for step 5: based on the data set, network structure and objective function of step 4, the label feature extraction network is trained. The training process aims to minimize the value of the objective function, and the trained network model is obtained according to the final weight parameters of the discriminator and the generator. The trained network model is used as the final semantic segmentation extraction model; In the specific implementation process of this embodiment, for step 6: constructing a post-processing system, using the segmentation results to obtain biometric indicators. Mainly based on the segmentation mask of the cerebellum and the occipital cistern in the segmentation results obtained by processing the cerebellum cross-section, combined with the medical prior knowledge provided in step 3, a post-processing system is established, and the steps are as follows: (1) Based on the segmentation mask of the cerebellum, the convex hull of the region is first calculated using an algorithm. The distance between the farthest pairs of points on the convex hull is then calculated using the rotating calculus method as the measurement of the transverse diameter of the cerebellum.

[0031] (2) According to the segmentation mask of the cerebellum, the corresponding ellipse is fitted, and the short axis direction of the ellipse is used as a straight line. Then, according to the segmentation mask of the cistern magna, the distance between the intersection of the extended short axis of the ellipse and the segmented area of ​​the cistern magna is obtained as the width of the cistern magna. In this way, two key biometric indicators of the cross-section of the cerebellum are obtained, namely the transverse diameter of the cerebellum and the width of the cistern magna.

[0032] The computer language used in the embodiment of the present invention is Python, and the development framework is Pytorch. The basics of data reading and writing, data calculation, network training, etc. are well-known technologies in the technical field and will not be elaborated here. Example

[0033] The present invention also discloses a transcerebellar cross-section quality assessment and biometry system, the system comprising: The target detection module is configured to input the cerebellum cross-section image in the fetal ultrasound image to be tested into a pre-trained target detection network model, and output an ultrasound image with a target detection frame and a corresponding confidence, wherein the target detection network model is based on a YOLOv11 model structure improved from YOLOv5 and YOLOv8; The quality assessment module is configured to pass the ultrasound image with the target detection frame and the corresponding confidence through the quality assessment system to assess the ultrasound image with the section quality meeting the standard; A semantic segmentation module is configured to input the ultrasound image with qualified section quality into a pre-trained segmentation model to obtain a segmentation mask of the cerebellum and the cistern magna, wherein the objective function of the segmentation model adopts a combination of Dice loss and cross entropy loss; The biometric module is configured to process the segmentation masks of the cerebellum and the cisterna magna through a post-processing system to obtain two key biometric indicators through a cross-section of the cerebellum.

[0034] This invention combines the section quality assessment with the measurement of biological indicators on the cerebellar cross section for the first time, establishes a scientific and objective quantitative evaluation system and a fast and accurate method for measuring biological indicators, and facilitates doctors to judge the section quality. It has the characteristics of high precision and high efficiency, and well solves the difficulties of large noise and complex structure in medical images, providing reliable results for subsequent quality assessment. It has the characteristics of high adaptability and is easy to expand to other medical ultrasound image quality assessment and biological indicator measurement tasks.

[0035] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for quality assessment and biometry of a transcerebellar cross section, characterized in that: The method comprises: Inputting the cerebellum cross-section image in the fetal ultrasound image to be tested into a pre-trained target detection network model, and outputting an ultrasound image with a target detection frame and a corresponding confidence level, wherein the target detection network model is based on a YOLOv11 model structure improved from YOLOv5 and YOLOv8; The ultrasound images with the target detection frame and the corresponding confidence level are passed through the quality assessment system to evaluate the ultrasound images with the section quality meeting the standard; Inputting the ultrasound image with qualified section quality into a pre-trained segmentation model to obtain the segmentation mask of the cerebellum and the cistern magnum, wherein the objective function of the segmentation model adopts a combination of Dice loss and cross entropy loss; The segmentation masks of the cerebellum and cisterna magna were processed through a post-processing system to obtain two key biometric indicators through the cerebellar cross-section.

2. The method for transcerebellar cross-sectional quality assessment and biometry according to claim 1, characterized in that: The training method of the target detection network model includes: Obtain transcerebellar cross-section samples from fetal ultrasound images and construct a target detection dataset; Perform data preprocessing on the target detection dataset; Based on the preprocessed target detection data set, the target is to minimize the value of the objective function, and the trained target detection network model is obtained according to the final weight parameters of the encoder and the decoder.

3. The method for transcerebellar cross-sectional quality assessment and biometry according to claim 2, characterized in that: The step of obtaining a transcerebellar cross-section sample in a fetal ultrasound image and constructing a target detection data set includes: According to professional books and expert opinions, the key tissues in the transverse section of the cerebellum were studied, and the five target tissues required for the section were obtained: the strong echo ring of the skull, the cavity of the septum pellucidum, the thalamus, the cerebellum, and the cistern magnum; A target box is drawn in the cross-section sample according to the target tissue, and the original image is matched one-to-one with the image containing the target box to construct a target detection dataset, wherein the target detection dataset is used as a training set.

4. The method for transcerebellar cross-sectional quality assessment and biometry according to claim 1, characterized in that: The target detection network model includes an input layer for data preprocessing and data enhancement, a backbone network for feature extraction, an intermediate layer composed of a convolutional neural network and integrating multi-level information to help the model identify targets of different scales, and an output layer for predicting target bounding boxes, classifications, and confidence levels.

5. The method for transcerebellar cross-sectional quality assessment and biometry according to claim 1, characterized in that: The quality assessment system is constructed based on the confidence results of each part of the target detection results obtained by processing the cerebellar cross-section using the target detection network model, and combined with medical prior knowledge to establish a cross-section scoring mechanism.

6. The method for transcerebellar cross-sectional quality assessment and biometry according to claim 5, characterized in that: The aspect scoring mechanism is as follows: (1) In each slice, there is one and only one target detection result for each target tissue; (2) The relative positions of different target tissues are fixed. The strong echo ring of the skull needs to include the cavum septum pellucidum, thalamus, cerebellum and cistern magnum. The four tissues are arranged in sequence along the midline of the brain and maintain a symmetrical relationship with the midline of the brain. The final section score is obtained based on the confidence level of the five target tissues, and a quantitative evaluation is performed on the cross-section through the cerebellum.

7. The method for transcerebellar cross-sectional quality assessment and biometry according to claim 5, characterized in that: The section scoring mechanism uses a 10-point scoring system. The initial score of each target tissue is 2 points, which is multiplied by the confidence of the target box of the corresponding tissue in the target detection result to obtain the final score of the tissue. The scores of the five target tissues are then summed to obtain the quantitative result of the quality assessment of the section.

8. The method of transcerebellar cross-sectional quality assessment and biometry according to claim 1, characterized in that: The post-processing system is established by combining the segmentation mask of the cerebellum and the cistern magnum in the segmentation result obtained by processing the cerebellum cross section according to the segmentation model and medical prior knowledge.

9. The method for transcerebellar cross-sectional quality assessment and biometry according to claim 8, characterized in that: The steps to establish a post-processing system are as follows: (1) Based on the segmentation mask of the cerebellum, the convex hull of the region is first calculated using an algorithm, and the distance between the farthest point pairs on the convex hull is calculated using the rotating calculus method as the measurement value of the transverse diameter of the cerebellum; (2) According to the segmentation mask of the cerebellum, the corresponding ellipse is fitted, and the short axis direction of the ellipse is used as a straight line. Then, according to the segmentation mask of the cistern magna, the distance between the intersection of the extended short axis of the ellipse and the segmented area of ​​the cistern magna is obtained, which is used as the width of the cistern magna. The two key biometric indicators, the transverse diameter of the cerebellum through the cross section of the cerebellum and the width of the cistern magna, are obtained.

10. A transcerebellar cross-section quality assessment and biometry system, characterized in that: The system comprises: The target detection module is configured to input the cerebellum cross-section image in the fetal ultrasound image to be tested into a pre-trained target detection network model, and output an ultrasound image with a target detection frame and a corresponding confidence, wherein the target detection network model is based on a YOLOv11 model structure improved from YOLOv5 and YOLOv8; The quality assessment module is configured to pass the ultrasound image with the target detection frame and the corresponding confidence through the quality assessment system to assess the ultrasound image with the section quality meeting the standard; A semantic segmentation module is configured to input the ultrasound image with qualified section quality into a pre-trained segmentation model to obtain a segmentation mask of the cerebellum and the cistern magna, wherein the objective function of the segmentation model adopts a combination of Dice loss and cross entropy loss; The biometric module is configured to process the segmentation masks of the cerebellum and the cisterna magna through a post-processing system to obtain two key biometric indicators through a cross-section of the cerebellum.

Images