WO2023133929A1 - 一种基于超声的人体组织对称性检测分析方法 - Google Patents

一种基于超声的人体组织对称性检测分析方法 Download PDF

Info

Publication number
WO2023133929A1
WO2023133929A1 PCT/CN2022/073355 CN2022073355W WO2023133929A1 WO 2023133929 A1 WO2023133929 A1 WO 2023133929A1 CN 2022073355 W CN2022073355 W CN 2022073355W WO 2023133929 A1 WO2023133929 A1 WO 2023133929A1
Authority
WO
WIPO (PCT)
Prior art keywords
human tissue
partitions
ultrasound
feature data
symmetry
Prior art date
Application number
PCT/CN2022/073355
Other languages
English (en)
French (fr)
Inventor
范列湘
李德来
蔡泽杭
李斌
黄彬
Original Assignee
汕头市超声仪器研究所股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 汕头市超声仪器研究所股份有限公司 filed Critical 汕头市超声仪器研究所股份有限公司
Publication of WO2023133929A1 publication Critical patent/WO2023133929A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G06T7/0014—Biomedical image inspection using an image reference approach
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/68—Analysis of geometric attributes of symmetry
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10132—Ultrasound image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30016—Brain
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30196—Human being; Person

Definitions

  • the invention belongs to the technical field of ultrasonic detection and analysis, and in particular relates to an ultrasonic-based method for detecting and analyzing the symmetry of human tissue.
  • the purpose of the present invention is to provide an ultrasound-based human tissue symmetry detection and analysis method that uses symmetry to judge tissue abnormalities to adapt to the analysis and diagnosis of such images.
  • an ultrasound-based method for detecting and analyzing the symmetry of human tissue including the following steps: S01, detection, performing ultrasonic detection on human tissue, and obtaining ultrasonic information corresponding to human tissue.
  • Boundary detection analyzing the grayscale of the ultrasound image to obtain the corresponding boundary of the human tissue.
  • Partitioning is performed according to the detected boundary and combined with a human tissue partitioning method, and is divided into at least two partitions.
  • the comparison of partitions includes the comparison of feature data between partitions that are symmetrical to each other and/or the comparison of partitions and reference feature data, wherein the reference feature data is the pre-stored health status of the corresponding human tissue feature data.
  • step S03 the boundary is detected using a traditional image edge detection method or a machine learning method.
  • step S04 the division of the human tissue is based on the human tissue and the detected boundary geometry, and the division is performed through the characteristics of the human tissue.
  • the feature extraction method in step S05 is to encode and combine the partition data to form corresponding feature data, wherein the encoding method adopts the direct sorting of pixels or the value vectorization of area pixel points.
  • the feature extraction method in step S05 adopts histogram statistics or gray-scale co-occurrence matrix, and forms corresponding feature data after normalization.
  • the beneficial effect of the present invention lies in: the use of computer to automatically analyze the symmetry of human tissue images to achieve the purpose of automatic identification of tissue abnormalities, by dividing human tissue into sections and comparing the sections, so as to quickly identify possible lesions in human tissue
  • the parts are quickly calibrated to provide doctors with a faster way to identify and facilitate doctors to diagnose.
  • Embodiment 1 discloses a method for detecting and analyzing the symmetry of human tissue based on ultrasound. Taking the brain as an example, the method includes the following steps: S01, detection, performing ultrasonic detection on human tissue, and obtaining ultrasonic information corresponding to human tissue; In this embodiment, ultrasonic testing is performed on the cranium, and corresponding ultrasonic information is acquired through echoes.
  • Imaging transforming ultrasound information into ultrasound images; 2D scanning will directly form a single image, while 3D scanning will directly generate images of multiple slices, which will be normalized and processed into a curved surface to form a brain model.
  • Boundary detection analyze the gray scale of the ultrasound image to obtain the corresponding boundary of human tissue; because the skull in the ultrasound brain image has a strong emission, it presents a high gray scale value in the image, and the probe is in the The position of the origin of the coordinates in the figure transmits sound waves to the brain through the acoustic window and detects echoes.
  • the edge of the image formed after detection is easy to detect.
  • the boundary between the skull and the brain tissue can be detected by using the gray scale or the gray scale plus the gray scale gradient value. In this way, the spatial position of the ultrasonic scanning plane in the cranial model can be estimated.
  • traditional image edge detection methods or machine learning methods are used to detect boundaries.
  • a traditional image edge detection method is adopted.
  • the position of the midsagittal plane and the midcoronal plane of the skull can be obtained from the three-dimensional skull model, and the junction of the midsagittal plane and the midcoronal plane is the scanning origin.
  • the model makes the left and right parts of the image symmetrical with respect to the midsagittal plane or the midfrontal plane.
  • the image is marked as an image that can be used for symmetry calculations.
  • the image is divided into two parts, preferably left and right Partitions, in this embodiment, the left and right partitions correspond to the left brain and the right brain respectively, and are divided from the middle plane of the sagittal plane.
  • Partition data processing feature extraction is performed on the divided partitions to form corresponding feature data; that is, feature extraction is performed on the left and right partitions respectively.
  • the feature extraction method in this step adopts histogram statistics or gray-scale co-occurrence matrix, And form the corresponding feature data after normalization.
  • the comparison of the partitions includes the feature data between the partitions that are symmetrical to each other
  • the comparison of feature data between partitions is used, and through the comparison of the left and right partitions, it can be preliminarily judged whether there is an abnormal gray scale value on the image.
  • the marked position is the position of the suspected lesion, which is convenient for the doctor to make a diagnosis, so as to save the doctor's diagnosis time and improve the pertinence of the diagnosis.
  • Embodiment 2 discloses a method for detecting and analyzing the symmetry of human tissue based on ultrasound, taking the kidney as an example, including the following steps: S01, detection, performing ultrasonic detection on human tissue, and obtaining ultrasonic information corresponding to human tissue; In this embodiment, ultrasonic testing is performed on the kidneys, and corresponding ultrasonic information is obtained through echoes.
  • Imaging transforming ultrasound information into ultrasound images; 2D scanning will directly form a single image, while 3D scanning will directly generate images of multiple slices, which are normalized and processed into a curved surface to form a kidney model.
  • Boundary detection analyzing the grayscale of the ultrasound image to obtain the corresponding boundary of the human tissue; during scanning, both kidneys are scanned.
  • the spine has a strong emission, and the image shows a high gray value.
  • the image formed after the detection is easy to detect the spine.
  • the spine can be detected by using gray scale or gray scale plus gray scale gradient value together. border with the kidney.
  • traditional image edge detection methods or machine learning methods are used to detect boundaries.
  • a traditional image edge detection method is adopted.
  • the human kidneys are directly distributed on both sides of the spine. Therefore, for the division of the kidneys, the division of the left and right kidneys into the left and right can be completed by directly identifying the position of the spine as the boundary.
  • Partition data processing feature extraction is performed on the divided partitions to form corresponding feature data; that is, feature extraction is performed on the left and right partitions respectively.
  • the feature extraction method in this step adopts histogram statistics or gray-scale co-occurrence matrix, And form the corresponding feature data after normalization.
  • the comparison of the partitions includes the difference between the symmetrical partitions and the partitions.
  • the feature data comparison between the partitions and the reference feature data is adopted, and by comparing the left and right partitions with the reference feature data, it can be preliminarily judged whether there is an abnormal gray scale value on the image. Since the kidneys are generally relatively symmetrical from left to right, and can be compared through a preset healthy template, the method of comparison with reference feature data is used for comparison.
  • the marked position is the position of the suspected lesion, which is convenient for the doctor to make a diagnosis, so as to save the doctor's diagnosis time and improve the pertinence of the diagnosis.
  • Embodiment 3 discloses a method for detecting and analyzing the symmetry of human tissue based on ultrasound, taking the thyroid gland as an example, including the following steps: S01, detection, performing ultrasonic detection on human tissue, and obtaining ultrasonic information corresponding to human tissue; In this embodiment, ultrasonic detection is performed on the thyroid gland, and corresponding ultrasonic information is acquired through echoes.
  • Imaging transforming ultrasound information into ultrasound images; 2D scanning will directly form a single image, while 3D scanning will directly generate images of multiple slices, which will be normalized and processed into a curved surface to form a thyroid model.
  • Boundary detection analyzing the grayscale of the ultrasound image to obtain the corresponding boundary of the human tissue; during scanning, the whole thyroid gland is scanned.
  • the trachea has weak emission, and the image shows a standard low gray scale value.
  • the trachea is easy to identify in the image formed after the detection. It can be detected by using gray scale or gray scale plus gray scale gradient value together.
  • the boundary between the trachea and the thyroid Specifically, in this step, traditional image edge detection methods or machine learning methods are used to detect boundaries. In this embodiment, a machine learning method is used for detection.
  • the human thyroid gland can be divided into the left lobe and the right lobe of the thyroid gland. Therefore, for the division of the thyroid gland, it can be completed by directly identifying the position of the trachea as the boundary, and dividing the left lobe and the right lobe into left and right regions. partition.
  • Partition data processing feature extraction is performed on the divided partitions to form corresponding feature data; that is, feature extraction is performed on the left and right partitions respectively.
  • the feature extraction method in this step is to code and combine the partition data to form corresponding features.
  • feature data where the encoding method adopts the direct sorting of pixels or the value vectorization of area pixels. This method codes and combines the data of each area and then inputs it into a trainable neural network to calculate the value of similarity.
  • the encoding method can be simple direct sorting of pixels [X1, X2, ..., XN], or vectorizing the values of pixels in each area [[x(1,1),x(2, 1),..x( N, 1)], [x(1,2),x(2, 2),..x(N, 2)], ..., [x(1,M),x(2, N),.. x(N, M)]] etc.
  • Uppercase represents data of all regions
  • lowercase represents data of one region
  • N represents the total number of regions
  • M represents the total number of pixels in the region.
  • the comparison of the partitions includes the difference between the symmetrical partitions and the partitions.
  • This embodiment adopts the comparison of characteristic data between partitions and the comparison of characteristic data between partitions and reference characteristic data at the same time. to determine whether there are abnormal grayscale values on the image. Since the left lobe and the right lobe of the thyroid gland are generally relatively symmetrical, and can be compared through a preset healthy template, the method of comparison with reference feature data is used for comparison.
  • the marked position is the position of the suspected lesion, which is convenient for the doctor to make a diagnosis, so as to save the doctor's diagnosis time and improve the pertinence of the diagnosis.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • Quality & Reliability (AREA)
  • Radiology & Medical Imaging (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Geometry (AREA)
  • Image Processing (AREA)
  • Ultra Sonic Daignosis Equipment (AREA)
  • Image Analysis (AREA)

Abstract

一种基于超声的人体组织对称性检测分析方法，包括如下步骤：检测；成像；边界检测，对超声图像的灰阶进行分析，获取人体组织的对应边界；分区，结合人体组织分区方法进行分区，至少分成两个分区；分区数据处理，形成对应的特征数据；比对，对分区的特征数据进行比对，确认特征数据是否存在差异；标记，将存在异常的灰阶值在超声图像中对应的位置进行标示并显示。有益效果在于：利用计算机自动分析人体组织图像的对称性来达到组织异常自动识别的目的，通过将人体组织分区，并且对分区进行比对，从而达到快速对人体组织中存在的病灶部位进行快速标定，提供给医生更快的识别途径以方便医生进行诊断。

Description

一种基于超声的人体组织对称性检测分析方法 技术领域
本发明属于超声检测分析技术领域，具体涉及一种基于超声的人体组织对称性检测分析方法。
 背景技术
多年来医学图像的自动分析和诊断有广泛和深入的研究，除了传统的图像处理技术，目前流行的有机器学习方法。前者利用图像中各种器官的灰度差别可以有效地区分边界，在病灶的灰阶与背景的灰阶足够大时，病灶亦可以被分拣出来，以实现自动检查的目的。后者利用大量的经过影像专家分析和标定的数据对神经网络进行训练。训练过的神经网络模型可以有效地重复影像专家分析的结果，达到自动检查的目的。
传统的图像处理技术可以有效地分割不同特性的区域，但是当组织的变化微小时或者组织的结构复杂时，其效果不好。机器学习方法需要大量的数据来做训练，而且当组织特征存在不确定性时，其结果变差。同时，人体组织的对称性不一定是空间几何一一对应的对称，譬如左右脑的回路不是对应中分线对折的对应关系，所以常规的几何变化方法不能满足要求。
 技术问题
本发明的目的在于提供一种利用对称性来判断组织异常，以适应这类图像的分析和诊断的基于超声的人体组织对称性检测分析方法。
为达到上述目的，本发明采用如下技术方案：一种基于超声的人体组织对称性检测分析方法，包括如下步骤：S01、检测，对人体组织进行超声检测，获取对应人体组织的超声信息。
S02、成像，将超声信息转化成超声图像。
S03、边界检测，对超声图像的灰阶进行分析，获取人体组织的对应边界。
S04、分区，依据检测到的边界，结合人体组织分区方法进行分区，至少分成两个分区。
S05、分区数据处理，对分成的分区分别进行特征提取，形成对应的特征数据。
S06、比对，对分区的特征数据进行比对，确认特征数据是否存在差异，计算出差异处对应的灰阶值。
S07、标记，将存在异常的灰阶值在超声图像中对应的位置进行标示并显示。
具体的，步骤S06中，分区的比对包括互相对称的分区与分区之间特征数据的比对和/或分区与参考特征数据的比对，其中，参考特征数据为对应人体组织预存的健康状态特征数据。
具体的，步骤S03中，边界的检测采用传统的图像边缘检测方法或机器学习方法进行检测。
具体的，步骤S04中，人体组织的分区是基于人体组织和检测到的边界几何，通过人体组织特性进行分区。
具体的，步骤S05中的特征提取方式为针对分区数据做编码组合从而形成对应的特征数据，其中编码方法采用像素点直接排序或区域像素点的值矢量化。
具体的，步骤S05中的特征提取方式采用直方图统计或灰阶共生矩阵，并在进行归一化后形成对应的特征数据。
本发明的有益效果在于：利用计算机自动分析人体组织图像的对称性来达到组织异常自动识别的目的，通过将人体组织分区，并且对分区进行比对，从而达到快速对人体组织中可能存在的病灶部位进行快速标定，提供给医生更快的识别途径以方便医生进行诊断。
 本发明的最佳实施方式
实施例1，本实施例公开一种基于超声的人体组织对称性检测分析方法，以颅脑为例，包括如下步骤：S01、检测，对人体组织进行超声检测，获取对应人体组织的超声信息；在本实施例中即对颅脑进行超声检测，并通过回波获取对应的超声信息。
S02、成像，将超声信息转化成超声图像；2D扫查会直接形成一副单一的图像，而3D扫查会直接生成多副切片的图像，在归一化化处理成曲面形成颅脑模型。
S03、边界检测，对超声图像的灰阶进行分析，获取人体组织的对应边界；由于超声颅脑图像的头骨有很强的发射，在图像中呈现高灰阶值，超声扫查时，探头处于图中的坐标原点的位置透过声学窗口向颅脑发射声波和检测回波。检测后形成的图像其边缘容易检测，利用灰阶或者灰阶加灰阶梯度值一起可以检测到头骨与颅脑内组织的边界，此边界在经过滤波和拟合之后得到以个头盖骨的曲面，从而推算超声扫描面在颅脑模型的空间位置。具体的，该步骤中，边界的检测采用传统的图像边缘检测方法或机器学习方法进行检测。本实施例中采用的为传统图像边缘检测方法。
S04、分区，依据检测到的边界，结合人体组织分区方法进行分区，至少分成两个分区；具体的，该步骤中，人体组织的分区是基于人体组织和检测到的边界几何，通过人体组织特性进行分区。从三维头盖骨模型可以得到头盖骨正中矢状面和正中冠状面的的位置，正中矢状面和正中冠状面的结合处为扫查原点。模型相对于正中矢状面或者正中额叶面都使得图像的左右部分呈现对称的特性。当拟合后的扫查面相对于正中矢状面或正中冠状面对称时，该图像被标定为可以用于对称性计算的图像，此时，将该图像进行分区，优选分为左右两个分区，本实施例中，左右分区分别对应左脑和右脑，从矢状面的中分面进行划分。
S05、分区数据处理，对分成的分区分别进行特征提取，形成对应的特征数据；即分别对左右分区进行特征提取，具体的，该步骤中的特征提取方式采用直方图统计或灰阶共生矩阵，并在进行归一化后形成对应的特征数据。
S06、比对，对分区的特征数据进行比对，确认特征数据是否存在差异，计算出差异处对应的灰阶值；该步骤中，分区的比对包括互相对称的分区与分区之间特征数据的比对和/或分区与参考特征数据的比对，其中，参考特征数据为对应人体组织预存的健康状态特征数据。本实施例采用分区与分区之间特征数据的比对，通过左右分区的比对，能够初步判断图像上是否存在异常的灰阶值。
S07、标记，将存在异常的灰阶值在超声图像中对应的位置进行标示并显示。此时，标记出来的位置为疑似病灶位置，方便医生进行诊断，以节省医生的诊断时间以及提高诊断的针对性。
实施例2，本实施例公开一种基于超声的人体组织对称性检测分析方法，以肾脏为例，包括如下步骤：S01、检测，对人体组织进行超声检测，获取对应人体组织的超声信息；在本实施例中即对肾脏进行超声检测，并通过回波获取对应的超声信息。
S02、成像，将超声信息转化成超声图像；2D扫查会直接形成一副单一的图像，而3D扫查会直接生成多副切片的图像，在归一化化处理成曲面形成肾脏模型。
S03、边界检测，对超声图像的灰阶进行分析，获取人体组织的对应边界；扫查时为两个肾脏均进行扫查。超声扫查时扫查到脊椎具有有很强的发射，在图像中呈现高灰阶值，检测后形成的图像其脊椎容易检测，利用灰阶或者灰阶加灰阶梯度值一起可以检测到脊椎与肾脏的边界。具体的，该步骤中，边界的检测采用传统的图像边缘检测方法或机器学习方法进行检测。本实施例中采用的为传统图像边缘检测方法。
S04、分区，依据检测到的边界，结合人体组织分区方法进行分区，至少分成两个分区；具体的，该步骤中，人体组织的分区是基于人体组织和检测到的边界几何，通过人体组织特性进行分区。人体肾脏直接分布在脊椎两侧，因此，对于肾脏的分区，直接通过识别脊椎的位置作为边界，将左右肾脏分别为左右两区，即可完成分区。
S05、分区数据处理，对分成的分区分别进行特征提取，形成对应的特征数据；即分别对左右分区进行特征提取，具体的，该步骤中的特征提取方式采用直方图统计或灰阶共生矩阵，并在进行归一化后形成对应的特征数据。
S06、比对，对分区的特征数据进行比对，确认特征数据是否存在差异，计算出差异处对应的灰阶值；具体的，该步骤中，分区的比对包括互相对称的分区与分区之间特征数据的比对和/或分区与参考特征数据的比对，其中，参考特征数据为对应人体组织预存的健康状态特征数据。本实施例采用分区与参考特征数据之间特征数据的比对，通过左右分区分别与参考特征数据的比对，能够初步判断图像上是否存在异常的灰阶值。由于肾脏一般左右相对比较对称，并且，能够通过预设的健康模板进行比对，因此，采用与参考特征数据进行比对的方式进行比对。
S07、标记，将存在异常的灰阶值在超声图像中对应的位置进行标示并显示。此时，标记出来的位置为疑似病灶位置，方便医生进行诊断，以节省医生的诊断时间以及提高诊断的针对性。
实施例3，本实施例公开一种基于超声的人体组织对称性检测分析方法，以甲状腺为例，包括如下步骤：S01、检测，对人体组织进行超声检测，获取对应人体组织的超声信息；在本实施例中即对甲状腺进行超声检测，并通过回波获取对应的超声信息。
S02、成像，将超声信息转化成超声图像；2D扫查会直接形成一副单一的图像，而3D扫查会直接生成多副切片的图像，在归一化化处理成曲面形成甲状腺模型。
S03、边界检测，对超声图像的灰阶进行分析，获取人体组织的对应边界；扫查时为甲状腺整体进行扫查。超声扫查时扫查到气管具有较弱的发射，在图像中呈现规范的低灰阶值，检测后形成的图像其气管容易识别，利用灰阶或者灰阶加灰阶梯度值一起可以检测到气管与甲状腺的边界。具体的，该步骤中，边界的检测采用传统的图像边缘检测方法或机器学习方法进行检测。本实施例中采用的为机器学习方法进行检测。
S04、分区，依据检测到的边界，结合人体组织分区方法进行分区，至少分成两个分区；具体的，该步骤中，人体组织的分区是基于人体组织和检测到的边界几何，通过人体组织特性进行分区。人体甲状腺依托于气管可以分为甲状腺左侧叶和右侧叶，因此，对于甲状腺的分区，直接通过识别气管的位置作为边界，将左侧叶和右侧叶分别为左右两区，即可完成分区。
S05、分区数据处理，对分成的分区分别进行特征提取，形成对应的特征数据；即分别对左右分区进行特征提取，具体的，该步骤中的特征提取方式为针对分区数据做编码组合从而形成对应的特征数据，其中编码方法采用像素点直接排序或区域像素点的值矢量化。此方法将各个区的数据做编码组合后输入可训练的神经网络来计算相似度的值。编码方法可以是简单的像素点直接排序[X1, X2, …，XN]、或者将各区域像素点的值矢量化[[x(1,1),x(2, 1),..x(N, 1)], [x(1,2),x(2, 2),..x(N, 2)], …, [x(1,M),x(2, N),..x(N, M)]]等。大写代表全部区域数据，小写代表区域的一个数据， N表示总的区域个数，M表示区域像素总个数。
S06、比对，对分区的特征数据进行比对，确认特征数据是否存在差异，计算出差异处对应的灰阶值；具体的，该步骤中，分区的比对包括互相对称的分区与分区之间特征数据的比对和/或分区与参考特征数据的比对，其中，参考特征数据为对应人体组织预存的健康状态特征数据。本实施例同时采用分区与分区之间特征数据的比对以及分区与参考特征数据之间特征数据的比对，通过左右分区互相比对以及左右分区分别与参考特征数据的比对，能够更好的判断图像上是否存在异常的灰阶值。由于甲状腺的左侧叶和右侧叶一般左右相对比较对称，并且，能够通过预设的健康模板进行比对，因此，采用与参考特征数据进行比对的方式进行比对。
S07、标记，将存在异常的灰阶值在超声图像中对应的位置进行标示并显示。此时，标记出来的位置为疑似病灶位置，方便医生进行诊断，以节省医生的诊断时间以及提高诊断的针对性。
当然，以上仅为本发明较佳实施方式，并非以此限定本发明的使用范围，故，凡是在本发明原理上做等效改变均应包含在本发明的保护范围内。

Claims (6)

  1. 一种基于超声的人体组织对称性检测分析方法，其特征在于，包括如下步骤：
    S01、检测，对人体组织进行超声检测，获取对应人体组织的超声信息；
    S02、成像，将超声信息转化成超声图像；
    S03、边界检测，对超声图像的灰阶进行分析，获取人体组织的对应边界；
    S04、分区，依据检测到的边界，结合人体组织分区方法进行分区，至少分成两个分区；
    S05、分区数据处理，对分成的分区分别进行特征提取，形成对应的特征数据；
    S06、比对，对分区的特征数据进行比对，确认特征数据是否存在差异，计算出差异处对应的灰阶值；
    S07、标记，将存在异常的灰阶值在超声图像中对应的位置进行标示并显示。
  2. 根据权利要求1所述基于超声的人体组织对称性检测分析方法，其特征在于：所述步骤S06中，分区的比对包括互相对称的分区与分区之间特征数据的比对和/或分区与参考特征数据的比对，其中，参考特征数据为对应人体组织预存的健康状态特征数据。
  3. 根据权利要求1所述基于超声的人体组织对称性检测分析方法，其特征在于：所述步骤S03中，边界的检测采用传统的图像边缘检测方法或机器学习方法进行检测。
  4. 根据权利要求1所述基于超声的人体组织对称性检测分析方法，其特征在于：所述步骤S04中，人体组织的分区是基于人体组织和检测到的边界几何，通过人体组织特性进行分区。
  5. 根据权利要求1-4任意一项所述基于超声的人体组织对称性检测分析方法，其特征在于，所述步骤S05中的特征提取方式为针对分区数据做编码组合从而形成对应的特征数据，其中编码方法采用像素点直接排序或区域像素点的值矢量化。
  6. 根据权利要求1-4任意一项所述基于超声的人体组织对称性检测分析方法，其特征在于：所述步骤S05中的特征提取方式采用直方图统计或灰阶共生矩阵，并在进行归一化后形成对应的特征数据。
PCT/CN2022/073355 2022-01-14 2022-01-24 一种基于超声的人体组织对称性检测分析方法 WO2023133929A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210043855.7A CN114418984A (zh) 2022-01-14 2022-01-14 一种基于超声的人体组织对称性检测分析方法
CN202210043855.7 2022-01-14

Publications (1)

Publication Number Publication Date
WO2023133929A1 true WO2023133929A1 (zh) 2023-07-20

Family

ID=81272611

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/073355 WO2023133929A1 (zh) 2022-01-14 2022-01-24 一种基于超声的人体组织对称性检测分析方法

Country Status (2)

Country Link
CN (1) CN114418984A (zh)
WO (1) WO2023133929A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117095067A (zh) * 2023-10-17 2023-11-21 山东虹纬纺织有限公司 基于人工智能的纺织品色差检测方法

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102930552A (zh) * 2012-11-22 2013-02-13 北京理工大学 基于对称结构减影的脑肿瘤自动提取方法
CN104657979A (zh) * 2014-12-24 2015-05-27 中国科学院深圳先进技术研究院 一种超声图像特征检测方法及系统
CN105997128A (zh) * 2016-08-03 2016-10-12 上海联影医疗科技有限公司 利用灌注成像识别病灶的方法及系统
US20180000461A1 (en) * 2016-06-29 2018-01-04 Niramai Health Analytix Pvt. Ltd. Thermography-based breast cancer screening using a measure of symmetry
CN107693047A (zh) * 2017-10-18 2018-02-16 飞依诺科技(苏州)有限公司 超声成像中基于对称组织的体标设置方法及系统
CN109363676A (zh) * 2018-10-09 2019-02-22 中国人民解放军第四军医大学 一种乳腺电阻抗扫描成像的双乳对称性检测方法
CN110782434A (zh) * 2019-10-17 2020-02-11 天津大学 一种脑结核mri图像病灶智能标记定位装置
CN111368586A (zh) * 2018-12-25 2020-07-03 深圳迈瑞生物医疗电子股份有限公司 超声成像方法及系统

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107767386B (zh) * 2017-10-12 2021-02-12 深圳开立生物医疗科技股份有限公司 超声图像处理方法及装置
CN109509186B (zh) * 2018-11-09 2020-09-18 北京邮电大学 基于大脑ct图像的缺血性脑卒中病灶检测方法及装置

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102930552A (zh) * 2012-11-22 2013-02-13 北京理工大学 基于对称结构减影的脑肿瘤自动提取方法
CN104657979A (zh) * 2014-12-24 2015-05-27 中国科学院深圳先进技术研究院 一种超声图像特征检测方法及系统
US20180000461A1 (en) * 2016-06-29 2018-01-04 Niramai Health Analytix Pvt. Ltd. Thermography-based breast cancer screening using a measure of symmetry
CN105997128A (zh) * 2016-08-03 2016-10-12 上海联影医疗科技有限公司 利用灌注成像识别病灶的方法及系统
CN107693047A (zh) * 2017-10-18 2018-02-16 飞依诺科技(苏州)有限公司 超声成像中基于对称组织的体标设置方法及系统
CN109363676A (zh) * 2018-10-09 2019-02-22 中国人民解放军第四军医大学 一种乳腺电阻抗扫描成像的双乳对称性检测方法
CN111368586A (zh) * 2018-12-25 2020-07-03 深圳迈瑞生物医疗电子股份有限公司 超声成像方法及系统
CN110782434A (zh) * 2019-10-17 2020-02-11 天津大学 一种脑结核mri图像病灶智能标记定位装置

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117095067A (zh) * 2023-10-17 2023-11-21 山东虹纬纺织有限公司 基于人工智能的纺织品色差检测方法
CN117095067B (zh) * 2023-10-17 2024-02-02 山东虹纬纺织有限公司 基于人工智能的纺织品色差检测方法

Also Published As

Publication number Publication date
CN114418984A (zh) 2022-04-29

Similar Documents

Publication Publication Date Title
CN110338844B (zh) 三维成像数据的显示处理方法和三维超声成像方法及系统
US9980704B2 (en) Non-invasive image analysis techniques for diagnosing diseases
CN100387196C (zh) 用于测量与疾病相关的组织变化的方法和系统
ES2914387T3 (es) Estudio inmediato
Rahmatullah et al. Quality control of fetal ultrasound images: Detection of abdomen anatomical landmarks using adaboost
Datta et al. Gray matter segmentation of the spinal cord with active contours in MR images
CN111374712B (zh) 一种超声成像方法及超声成像设备
CN110288698B (zh) 基于mri的半月板三维重建系统
Yaqub et al. Automatic detection of local fetal brain structures in ultrasound images
CN108961278B (zh) 基于影像数据的腹壁肌肉分割的方法及其系统
Sindhwani et al. Semi‐automatic outlining of levator hiatus
Wijata et al. Unbiased validation of the algorithms for automatic needle localization in ultrasound-guided breast biopsies
WO2023133929A1 (zh) 一种基于超声的人体组织对称性检测分析方法
Xia et al. Automatic plane of minimal hiatal dimensions extraction from 3D female pelvic floor ultrasound
CN111481233A (zh) 胎儿颈项透明层厚度测量方法
Aji et al. Automatic measurement of fetal head circumference from 2-dimensional ultrasound
Rahmatullah et al. Anatomical object detection in fetal ultrasound: computer-expert agreements
Chen et al. Recognition of hyperparathyroidism based on transfer learning
CN113838020B (zh) 一种基于钼靶影像的病变区域量化方法
Cao et al. Liver fibrosis identification based on ultrasound images
Ishikawa et al. Detecting a Fetus in Ultrasound Images using Grad CAM and Locating the Fetus in the Uterus.
CN115439650A (zh) 基于ct图像跨模态迁移学习的肾脏超声图像分割方法
Feng et al. Automatic fetal weight estimation using 3d ultrasonography
CN113744234A (zh) 一种基于gan的多模态脑部图像配准方法
CN111862014A (zh) 一种基于左右侧脑室分割的alvi自动测量方法及装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22919575

Country of ref document: EP

Kind code of ref document: A1