WO2019061148A1 - 超声动态图像处理方法、装置及超声摄像设备 - Google Patents

超声动态图像处理方法、装置及超声摄像设备 Download PDF

Info

Publication number
WO2019061148A1
WO2019061148A1 PCT/CN2017/103978 CN2017103978W WO2019061148A1 WO 2019061148 A1 WO2019061148 A1 WO 2019061148A1 CN 2017103978 W CN2017103978 W CN 2017103978W WO 2019061148 A1 WO2019061148 A1 WO 2019061148A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
correlation coefficient
roi
ultrasonic
ultrasound
Prior art date
Application number
PCT/CN2017/103978
Other languages
English (en)
French (fr)
Inventor
刘利鸿
Original Assignee
北京匡图医疗科技有限公司
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 北京匡图医疗科技有限公司 filed Critical 北京匡图医疗科技有限公司
Priority to PCT/CN2017/103978 priority Critical patent/WO2019061148A1/zh
Publication of WO2019061148A1 publication Critical patent/WO2019061148A1/zh

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
    • A61B8/08—Detecting organic movements or changes, e.g. tumours, cysts, swellings
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion

Definitions

  • the present invention relates to the field of image processing and analysis, and in particular to an ultrasonic dynamic image processing method, apparatus, and ultrasonic imaging apparatus.
  • B-mode ultrasound is the most widely used and most influential medical ultrasound.
  • the interface formed by each layer of tissue and the structure of the tissue are reflected back, and the intensity of the light is reflected by the light and dark, and the plurality of light spots are arranged in an orderly manner to form a corresponding slice. Image.
  • Ultrasound is becoming an effective tool to help clinicians diagnose human tissue lesions such as pneumonia due to the advantages of no radiation and short diagnostic time.
  • a big problem faced by clinicians is how to use and interpret ultrasound images to determine the region of interest (ROI), which requires familiarity with the ultrasound anatomy of the tissue, the ultrasound display of different lesions, and the operation of the ultrasound probe.
  • ROI region of interest
  • an embodiment of the present invention provides an ultrasonic dynamic image processing method, wherein the method includes: acquiring an ultrasonic dynamic image; generating a correlation coefficient image based on the acquired ultrasonic dynamic image; and determining a sense according to the generated correlation coefficient image Interest area ROI; extracting statistical features from the acquired ultrasound image based on the determined ROI.
  • the acquired ultrasound dynamic image has a plurality of consecutive image frames
  • the generating the correlation coefficient image based on the acquired ultrasound dynamic image comprises: selecting two adjacent image frames from the plurality of consecutive image frames; Correlation coefficients of the adjacent two image frames to generate the correlation coefficient image.
  • the determining, according to the generated correlation coefficient image, the ROI comprises: determining a position of a correlation coefficient change rate in the correlation coefficient image that is greater than a preset value as a boundary of the ROI.
  • the statistical feature comprises at least one of: a first-order feature, a second-order feature, and a difference.
  • the method further comprises: training the prediction model with the statistical feature; determining, by the trained prediction model, the probability that the ROI exists or the ROI exists from the newly acquired ultrasound image.
  • the method includes training the predictive model using the statistical features and a machine learning algorithm.
  • the present invention provides an ultrasound dynamic image processing apparatus, wherein the apparatus includes: an image acquisition unit configured to acquire an ultrasound dynamic image; and a correlation coefficient image generation unit configured to generate a correlation coefficient based on the acquired ultrasound dynamic image And an area determining unit, configured to determine a region of interest ROI according to the generated correlation coefficient image; and a feature extracting unit configured to extract a statistical feature from the acquired ultrasound image according to the determined ROI.
  • the acquired ultrasound dynamic image has a plurality of consecutive image frames
  • the correlation coefficient image generation unit includes: an image frame selection module, configured to select two adjacent image frames from the plurality of consecutive image frames;
  • a correlation coefficient calculation module is configured to calculate correlation coefficients of the adjacent two image frames to generate the correlation coefficient image.
  • the area determining unit is configured to determine, as the boundary of the ROI, a position in which the correlation coefficient change rate in the correlation coefficient image is greater than a preset value.
  • the statistical feature includes at least one of: a first-order feature, a second-order feature, And the difference.
  • the apparatus further includes: a model training unit, configured to train the prediction model by using the statistical feature; and a determining unit, configured to determine, by using the trained prediction model, the presence of the ROI or the presence of the ROI from the newly acquired ultrasound image The probability.
  • a model training unit configured to train the prediction model by using the statistical feature
  • a determining unit configured to determine, by using the trained prediction model, the presence of the ROI or the presence of the ROI from the newly acquired ultrasound image The probability.
  • the model training unit is configured to train the prediction model by using the statistical feature and a machine learning algorithm.
  • the present invention provides an ultrasonic imaging apparatus including the above-described ultrasonic moving image processing apparatus.
  • the present invention provides a machine readable storage medium having stored thereon instructions for causing a machine to perform the method according to the above.
  • the ROI in the ultrasound image is automatically determined by using the correlation coefficient image generated from the ultrasound image, thereby facilitating the positioning of the ROI in the ultrasound image; and using the statistical features corresponding to the ROI, the classification of the ROI is realized, thereby realizing Accurate interpretation of ultrasound images.
  • FIG. 1 is a flowchart of an ultrasonic image processing method according to an embodiment of the present invention
  • FIG. 2 is a flow chart of an ultrasonic image processing method according to a preferred embodiment of the present invention.
  • FIG. 3a is an example lung B mode ultrasound image provided by an embodiment of the present invention.
  • Figure 3b is a correlation coefficient image corresponding to Figure 3a;
  • FIG. 4a is another example lung B-mode ultrasound image provided by an embodiment of the present invention.
  • Figure 4b is a correlation coefficient image corresponding to Figure 4a;
  • FIG. 5 is a schematic diagram of the composition of an ultrasound image processing apparatus according to an embodiment of the present invention.
  • Fig. 6 is a block diagram showing the composition of an ultrasonic imaging apparatus according to an exemplary embodiment of the present invention.
  • first and second are used merely to facilitate the description of different components, and are not to be construed as indicating or implying a sequence relationship, relative importance or implicit indication.
  • features defining “first” and “second” may include at least one of the features, either explicitly or implicitly.
  • FIG. 1 is a flowchart of an ultrasonic image processing method according to an embodiment of the present invention.
  • the ultrasonic image processing method provided by the embodiment of the present invention may include:
  • the ultrasonic moving image can be obtained from the ultrasonic imaging apparatus 1 as shown in FIG. 6.
  • the ultrasonic imaging apparatus 1 forms a two-dimensional ultrasonic image for a part to be examined (for example, a lung) using a reflected echo signal capable of transmitting and receiving ultrasonic waves in a subject (for example, a human body), and can perform ultrasonic image storage and/or Or display.
  • the ultrasonic imaging apparatus 1 may include an ultrasonic probe 3 including an oscillator element that irradiates the subject and receives ultrasonic waves, an ultrasonic wave transmitting and receiving unit 4 that transmits and receives an ultrasonic signal, and a two-dimensional ultrasound image based on the received signal ( Ultrasound image composing portion 5 of B-mode ultrasonic image); display portion 6 for displaying an ultrasonic image (for example, ultrasonic moving image) constituted in the ultrasonic image constructing portion; control portion 7 for control; providing instruction to the control portion 7. a control panel 8; a storage unit 9 for storing an ultrasonic image and a control arithmetic program; and a sound generation unit 10 for generating a sound such as an alarm.
  • an ultrasonic probe 3 including an oscillator element that irradiates the subject and receives ultrasonic waves, an ultrasonic wave transmitting and receiving unit 4 that transmits and receives an ultrasonic signal, and a two-dimensional ultrasound image based on the received signal ( Ultra
  • an ultrasound image or an ultrasound dynamic image may be acquired from the ultrasound image constructing section 5. image.
  • the ultrasound image data or signals may be acquired from the ultrasound transceiver 4, in this embodiment, the above steps may be to acquire ultrasound image data.
  • the ultrasound dynamic image may be a continuous image (eg, having a plurality of consecutive image frames).
  • correlation coefficient images of the two ultrasound images may be obtained using at least two (or two frames) ultrasound images in the ultrasound dynamic image.
  • two adjacent image frames may be first selected from the plurality of consecutive image frames (or in an ultrasonic moving image); then, correlation coefficients of the adjacent two image frames are calculated to generate the correlation coefficient image.
  • the correlation coefficients of two adjacent image frames may be respectively calculated by respectively calculating correlation coefficients of pixels corresponding to positions in the two image frames, and then the correlation coefficients of the corresponding pixels are formed to form a correlation coefficient image, as shown in FIG. 3a and FIG. 3b. Shown.
  • FIG. 3a is an example lung B-mode ultrasound image provided by an embodiment of the present invention
  • FIG. 3b is a correlation coefficient image corresponding to FIG. 3a.
  • the abscissa and the ordinate of both Figures 3a and 3b are pixels
  • the right side of Figure 3b shows the colorimetric of the correlation coefficient in the figure.
  • the reflected signal of the ultrasonic wave in the region of the first tissue 31 has a high correlation coefficient, and total reflection occurs at the first interface 33 of the first tissue 31 and the air 32, resulting in the area of the air 32 being Noise, so the air 32 region has a lower correlation coefficient.
  • FIG. 4a is another example lung B-mode ultrasound image provided by an embodiment of the present invention
  • FIG. 4b is a correlation coefficient image corresponding to FIG. 4a. 4a and 4b, both the abscissa and the ordinate are pixels, and the right side of Fig. 4b shows the colorimetric of the correlation coefficient in the figure.
  • the reflected signal of the ultrasonic wave in the first tissue 31 region has a higher correlation coefficient
  • the correlation coefficient at the second interface 43 between the second tissue 42 region and the first tissue 31 region is higher.
  • the large variation e.g., the correlation coefficient on both sides of the second interface 43 reaches 50%
  • the correlation coefficient in the second tissue 42 region is much higher than the air 32 region.
  • different tissue regions or different regions of tissue can be distinguished by correlation coefficients.
  • the ROI may be determined from the trend of the correlation coefficient (eg, automatically) using the generated correlation coefficient image.
  • a position in the correlation coefficient image in which the correlation coefficient change rate is greater than a preset value (for example, 50%) may be determined as a boundary of the ROI, thereby determining an ROI according to the generated correlation coefficient image.
  • the ROI can be automatically determined from the boundary (eg, based on image recognition).
  • the ROI can be a normal tissue or a diseased tissue.
  • the correlation coefficient image since the correlation coefficient image has a corresponding pixel position with the ultrasound image, after determining the ROI position in the correlation coefficient image, it can be mapped to a position in the ultrasound image, thereby determining normal in the ultrasound image.
  • the statistical feature can include at least one of: a first order feature, a second order feature, and a difference. The extraction of statistical features can be performed according to an extraction algorithm corresponding to the statistical features.
  • the ultrasound image for which the statistical feature is extracted may be any of the two image frames used by the generated correlation coefficient image.
  • the ROI in the ultrasound image is automatically determined by using the correlation coefficient image generated from the ultrasound image, thereby facilitating the positioning of the ROI in the ultrasound image; and using the statistical features corresponding to the ROI, the classification of the ROI is realized, thereby realizing Accurate interpretation of ultrasound images.
  • the ultrasonic image processing method provided by the embodiment of the present invention may include:
  • S106 Determine, by using the trained prediction model, the probability that the ROI exists or the ROI exists from the newly acquired ultrasound image.
  • training the predictive model with statistical features can include training the predictive model with the statistical features and a machine learning algorithm.
  • machine learning algorithms may include, but are not limited to, decision trees, logistic regression, and the like. It should be noted that embodiments of the present invention may employ any machine learning algorithm or a combination of several machine learning algorithms. Different machine learning algorithms may have different performance in prediction. The combination of several machine learning algorithms may outperform any of these in predictive performance.
  • a well-trained predictive model such as a predictive model trained using more than one machine learning algorithm, may be present for newly acquired ultrasound data (eg, extracted statistical features) or ultrasound images.
  • the ROI's judgment for example, can give the probability of the presence of an ROI.
  • an ultrasonic dynamic image processing apparatus is provided.
  • FIG. 5 is a schematic diagram of the composition of an ultrasonic image processing apparatus according to an embodiment of the present invention.
  • the apparatus may include: an image acquiring unit 51, configured to acquire an ultrasonic dynamic image; a correlation coefficient image generating unit 52, configured to generate a correlation coefficient image based on the acquired ultrasonic moving image; and an area determining unit 53, Determining the ROI of the ROI according to the generated correlation coefficient image; the feature extraction unit 54 is configured to extract statistical features from the acquired ultrasound image according to the determined ROI.
  • the acquired ultrasound dynamic image may have a plurality of consecutive image frames
  • the correlation coefficient image generation unit 52 may include: an image frame selection module 521, configured to select two adjacent ones from the plurality of consecutive image frames. An image frame; a correlation coefficient calculation module 522, configured to calculate correlation coefficients of the adjacent two image frames to generate the correlation coefficient image.
  • the region determining unit 53 may be configured to determine, as the boundary of the ROI, a position in which the correlation coefficient change rate in the correlation coefficient image is greater than a preset value.
  • the statistical feature can include at least one of: a first order feature, a second order feature, and a difference.
  • the apparatus may further include: a model training unit 55, configured to train the prediction model by using the statistical feature; and a determining unit 56, configured to determine, by using the trained prediction model, the presence of the ROI from the newly acquired ultrasound image or There is a probability that the ROI exists.
  • a model training unit 55 configured to train the prediction model by using the statistical feature
  • a determining unit 56 configured to determine, by using the trained prediction model, the presence of the ROI from the newly acquired ultrasound image or There is a probability that the ROI exists.
  • model training unit 55 can be configured to train the predictive model using the statistical features and machine learning algorithms.
  • an ultrasonic imaging apparatus is provided.
  • an ultrasonic imaging apparatus may include: an ultrasonic probe 3 including an oscillator element that irradiates an object to receive ultrasonic waves; and an ultrasonic transmission/reception unit 4 that transmits and receives an ultrasonic signal; An ultrasonic image forming unit 5 such as a two-dimensional ultrasonic image (B-mode ultrasonic image); a display unit 6 that displays an ultrasonic image formed in the ultrasonic image forming unit; a control unit 7 for control; and an instruction to the control unit 7 a control panel 8; a storage unit 9 for storing an ultrasonic image and a control arithmetic program; and a sound generation unit 10 for generating a sound such as an alarm.
  • the ultrasonic image constructing portion 5 may include the ultrasonic moving image processing device provided by the above embodiment of the present invention
  • the ultrasonic moving image processing apparatus provided by the above-described embodiments of the present invention may be provided to be coupled with the ultrasonic image constructing section 5 to acquire an ultrasonic image from the ultrasonic image forming section.
  • the ultrasonic image processing apparatus provided by the above-described embodiments of the present invention may be provided to be coupled with the ultrasonic transceiver unit 4 to acquire ultrasound image data or signals from the ultrasonic transceiver unit 4.
  • the display portion 6 and the storage port 9 can be used to display and store an ultrasound image or a correlation coefficient image from the ultrasound image processing device, respectively.
  • a machine readable storage medium which is readable by a machine An instruction is stored on the storage medium for causing the machine to perform the method according to the above.
  • a machine readable storage medium provided by an embodiment of the present invention may be implemented in an apparatus such as an ultrasonic imaging apparatus.
  • An ultrasonic dynamic image processing method comprising:
  • Statistical features are extracted from the acquired ultrasound images based on the determined ROI.
  • Correlating coefficients of the adjacent two image frames are calculated to generate the correlation coefficient image.
  • determining the ROI according to the generated correlation coefficient image comprises: determining, as the location of the correlation coefficient change rate in the correlation coefficient image, a position greater than a preset value The boundary of the ROI.
  • the statistical feature comprises at least one of: a first order feature, a second order feature, and a difference.
  • Determining the presence of the ROI or presence from a newly acquired ultrasound image through a trained predictive model The probability of the ROI.
  • An ultrasonic dynamic image processing apparatus comprising:
  • An image acquisition unit configured to acquire an ultrasound dynamic image
  • a correlation coefficient image generating unit configured to generate a correlation coefficient image based on the acquired ultrasound moving image
  • a region determining unit configured to determine a region of interest ROI according to the generated correlation coefficient image
  • a feature extraction unit configured to extract a statistical feature from the acquired ultrasound image according to the determined ROI.
  • the image acquisition unit is configured to acquire the ultrasound dynamic image from an ultrasound imaging apparatus.
  • the correlation coefficient image generation unit comprises:
  • An image frame selection module configured to select two adjacent image frames from the plurality of consecutive image frames
  • a correlation coefficient calculation module is configured to calculate correlation coefficients of the adjacent two image frames to generate the correlation coefficient image.
  • the area determining unit is configured to determine a position in the correlation coefficient image that the correlation coefficient change rate is greater than a preset value as a boundary of the ROI.
  • the statistical characteristic comprises at least one of: a first order feature, a second order feature, and a difference.
  • a model training unit for training the prediction model by using the statistical feature
  • a judging unit configured to determine, by the trained prediction model, the probability that the ROI exists or the ROI exists from the newly acquired ultrasound image.
  • Apparatus according to embodiment 16 wherein said model training unit is operative to train said predictive model using said statistical features and machine learning algorithms.
  • the machine learning algorithm comprises one of: a decision tree or a logical regression.
  • An ultrasonic imaging apparatus characterized in that the ultrasonic imaging apparatus comprises the ultrasonic moving image processing apparatus according to any one of embodiments 10-18.
  • a machine readable storage medium having stored thereon instructions for causing a machine to perform the method of any of embodiments 1-9.
  • the foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and the like. .
  • any combination of various different embodiments of the embodiments of the present invention may also be performed. As long as it does not deviate from the idea of the embodiments of the present invention, it should also be regarded as the content disclosed in the embodiments of the present invention.

Abstract

本发明实施例提供一种超声动态图像处理方法、装置及超声摄像设备，属于图像处理及分析领域。所述方法包括：获取超声动态图像；基于所获取的超声动态图像生成相关系数图像；根据所生成的相关系数图像确定感兴趣区域ROI；根据所确定的ROI从所获取的超声图像中提取统计特征。通过上述技术方案，利用根据超声图像生成的相关系数图像自动确定超声图像中的ROI，便于实现对超声图像中ROI的定位；利用对应于ROI的统计特征，实现了对ROI的分类，从而能够实现对超声图像准确解读。

Description

超声动态图像处理方法、装置及超声摄像设备 技术领域
本发明涉及图像处理及分析领域，具体地，涉及一种超声动态图像处理方法、装置及超声摄像设备。
背景技术
B模式超声是应用最广、影响最大的医学用超声。这种方法是在声束穿经人体时，把各层组织所构成的介面和组织内结构的反射回声，以光点的明暗反应其强弱，由众多的光点排列有序的组成相应切面的图像。
由于无辐射，诊断时间短等优点，超声正在成为帮助临床医师诊断诸如肺炎等人体组织病变的有效工具。
本申请发明人在实现本发明的过程中发现，现有技术的上述方案具有缺陷。临床医生面临的一个大问题是如何使用及解释超声图像来确定感兴趣区域(region of interest，ROI)，这需要熟悉组织的超声解剖结构，不同病变的超声显示以及超声探头的操作。
针对上述问题，现有技术中尚无良好解决方案。
发明内容
本发明实施例的目的是提供一种方法及设备，该方法及设备能够对超声图像进行处理准确确定感兴趣区域。
为了实现上述目的，本发明实施例提供一种超声动态图像处理方法，其中，该方法包括：获取超声动态图像；基于所获取的超声动态图像生成相关系数图像；根据所生成的相关系数图像确定感兴趣区域ROI；根据所确定的ROI从所获取的超声图像中提取统计特征。
可选地，所获取的超声动态图像具有多个连续图像帧，所述基于所获取的超声动态图像生成相关系数图像包括：从所述多个连续图像帧中选择相邻两个图像帧；计算所述相邻两个图像帧的相关系数以生成所述相关系数图像。
可选地，所述根据所生成的相关系数图像确定所述ROI包括：将所述相关系数图像中相关系数变化率大于预设值的位置确定为所述ROI的边界。
可选地，所述统计特征包括以下中至少一者：一阶特征、二阶特征、以及递差。
可选地，该方法还包括：利用所述统计特征训练预测模型；通过训练的预测模型从新获取的超声图像中确定存在所述ROI或存在所述ROI的概率。
可选地，该方法包括：利用所述统计特征和机器学习算法训练所述预测模型。
另一方面，本发明提供一种超声动态图像处理装置，其中，该装置包括：图像获取单元，用于获取超声动态图像；相关系数图像生成单元，用于基于所获取的超声动态图像生成相关系数图像；区域确定单元，用于根据所生成的相关系数图像确定感兴趣区域ROI；特征提取单元，用于根据所确定的ROI从所获取的超声图像中提取统计特征。
可选地，所获取的超声动态图像具有多个连续图像帧，所述相关系数图像生成单元包括：图像帧选择模块，用于从所述多个连续图像帧中选择相邻两个图像帧；相关系数计算模块，用于计算所述相邻两个图像帧的相关系数以生成所述相关系数图像。
可选地，所述区域确定单元，用于将所述相关系数图像中相关系数变化率大于预设值的位置确定为所述ROI的边界。
可选地，所述统计特征包括以下中至少一者：一阶特征、二阶特征、 以及递差。
可选地，该装置还包括：模型训练单元，用于利用所述统计特征训练预测模型；判断单元，用于通过训练的预测模型从新获取的超声图像中确定存在所述ROI或存在所述ROI的概率。
可选地，所述模型训练单元，用于利用所述统计特征和机器学习算法训练所述预测模型。
另一方面，本发明提供一种超声摄像设备，该超声摄像设备包括上述的超声动态图像处理装置。
另一方面，本发明提供一种机器可读存储介质，该机器可读存储介质上存储有指令，该指令用于使得机器执行根据上述的方法。
通过上述技术方案，利用根据超声图像生成的相关系数图像自动确定超声图像中的ROI，便于实现对超声图像中ROI的定位；利用对应于ROI的统计特征，实现了对ROI的分类，从而能够实现对超声图像准确解读。
本发明实施例的其它特征和优点将在随后的具体实施方式部分予以详细说明。
附图说明
附图是用来提供对本发明实施例的进一步理解，并且构成说明书的一部分，与下面的具体实施方式一起用于解释本发明实施例，但并不构成对本发明实施例的限制。在附图中：
图1是本发明实施例提供的超声图像处理方法流程图；
图2是本发明一个优选实施例提供的超声图像处理方法流程图；
图3a是本发明实施例提供的一个示例肺部B模式超声图像；
图3b是对应于图3a的相关系数图像；
图4a是本发明实施例提供的另一个示例肺部B模式超声图像；
图4b是对应于图4a的相关系数图像；
图5是本发明实施例提供的超声图像处理装置组成示意图；
图6是本发明示例实施例的超声摄像设备的组成框图。
附图标记说明
1      超声摄像设备          2     被检体
3      超声波探头            4     超声波收发部
5      超声图像构成部        6     显示部
7      控制部                8     控制面板
9      存储部                10    声音产生部
31     第一组织              32    空气
33     第一界面              42    第二组织
43     第二界面              51    图像获取单元
52     相关系数图像生成单元  521   图像帧选择模块
522    相关系数计算模块      53    区域确定单元
54     特征提取单元          55    模型训练单元
56     判断单元。
具体实施方式
以下结合附图对本发明实施例的具体实施方式进行详细说明。应当理解的是，此处所描述的具体实施方式仅用于说明和解释本发明实施例，并不用于限制本发明实施例。
需要说明的是，在本发明的描述中，术语“第一”、“第二”仅用于方便描述不同的部件，而不能理解为指示或暗示顺序关系、相对重要性或者隐含指明所指示的技术特征的数量。由此，限定有“第一”、“第二”的特征可以明示或者隐含地包括至少一个该特征。
在本发明的描述中，需要理解的是，术语“上”、“下”、“内”、“外”、 “顶”、“底”等指示的方位或位置关系为基于附图所示的方位或位置关系，仅是为了便于描述本发明和简化描述，而不是指示或暗示所指的装置或元件必须具有特定的方位、以特定的方位构造和操作，因此不能理解为对本发明的限制。
本技术领域技术人员可以理解，本发明的说明书中使用的措辞“包括”及“包含”是指存在上述特征、整数、步骤、操作、元件和/或组件，但是并不排除存在或添加一个或多个其他特征、整数、步骤、操作、元件、组件和/或它们的组合。应该理解，当我们称元件被“连接”或“耦接”到另一元件时，它可以直接连接或耦接到其他元件，或者也可以存在中间元件。此外，这里使用的“连接”或“耦接”可以包括无线连接或无线耦接。这里使用的措辞“和/或”包括一个或更多个相关联的列出项的全部或任一单元和全部组合。
图1是本发明实施例提供的超声图像处理方法流程图。如图1所示，本发明实施例提供的超声图像处理方法可以包括：
S101，获取超声动态图像。
在实施方式中，超声动态图像可以从如图6所示的超声摄像设备1获得。该超声摄像设备1使用能够在被检体(例如，人体)内收发超声波的反射回波信号，针对被检部位(例如，肺部)来形成二维超声图像，并可以进行超声图像存储和/或显示。在实施方式中，超声摄像设备1可以包括：具备向被检体照射并接收超声波的振荡器元件的超声波探头3；收发超声波信号的超声波收发部4；基于接收信号来构成诸如二维超声图像(B模式超声图像)的超声图像构成部5；显示在超声图像构成部中构成的超声图像(例如，超声动态图像)的显示部6；用于控制的控制部7；向控制部7提供指示的控制面板8；用于存储超声图像以及控制运算程序的存储部9；以及产生告警等声音的声音产生部10。
在实施方式中，可以从超声图像构成部5获取超声图像或超声动态图 像。在不同的实施方式中，可以从超声波收发部4获取超声图像数据或信号，在该实施方式中，上述步骤可以为获取超声图像数据。在实施方式中，上述超声动态图像可以是连续图像(例如，具有多个连续图像帧)。
S102，基于所获取的超声动态图像生成相关系数图像。
在实施方式中，利用超声动态图像中的至少两个(或两帧)超声图像可以得到该两个超声图像的相关系数图像。在实施方式中，可以首先从所述多个连续图像帧(或超声动态图像中)中选择相邻两个图像帧；然后，计算该相邻两个图像帧的相关系数以生成所述相关系数图像。在实施方式中相邻两个图像帧的相关系数可以通过分别计算两个图像帧中对应位置的像素的相关系数，然后由得到的对应像素的相关系数构成相关系数图像，如图3a和图3b所示。
图3a是本发明实施例提供的一个示例肺部B型超声图像，图3b是对应于图3a的相关系数图像。其中，图3a和图3b中横坐标和纵坐标均为像素，图3b右侧示出相关系数在图中的比色。如图3a和图3b所示，超声波在第一组织31区域的反射信号具有较高的相关系数，而在第一组织31与空气32的第一界面33发生全反射，导致了空气32区域为噪声，因此空气32区域具有较低的相关系数。
图4a是本发明实施例提供的另一个示例肺部B型超声图像，图4b是对应于图4a的相关系数图像。其中，图4a和图4b中横坐标和纵坐标均为像素，图4b右侧示出相关系数在图中的比色。如图4a和图4b所示，超声波在第一组织31区域的反射信号具有较高的相关系数，而第二组织42区域与第一组织31区域之间的第二界面43处相关系数存在较大变化(例如，第二界面43两边相关系数变化达到50％)，在第二组织42区域的相关系数比空气32区域高很多。因此，通过相关系数能够区分不同的组织区域或组织的不同区域(例如，正常组织或病变组织)。
S103，根据所生成的相关系数图像确定感兴趣区域ROI。
在实施方式中，利用生成的相关系数图像可以根据相关系数的变化趋势(例如，自动)确定ROI。在实施方式中，可以将所述相关系数图像中相关系数变化率大于预设值(例如，50％)的位置确定为所述ROI的边界，从而根据所生成的相关系数图像确定ROI。在确定了ROI的边界(例如，图3b中的第一界面33)后，可以(例如，基于图像识别)根据该边界自动确定ROI。在实施方式中ROI可以是正常组织或病变组织。
S104，根据所确定的ROI从所获取的超声图像中提取统计特征。
在实施方式中，由于相关系数图像与超声图像具有相对应的像素位置，因此在确定了相关系数图像中的ROI位置后，可以将其对应到超声图像中的位置，从而在超声图像中确定正常组织或病变组织的范围。在确定了组织的范围后，可以对该范围内的超声图像(例如，超声图像中的纹理图案)进行统计特征提取，从而确定组织在超声图像中的统计特征。在实施方式中，统计特征可以包括以下中至少一者：一阶特征、二阶特征、以及递差。对于统计特征的提取可以根据与统计特征相对应的提取算法进行。统计特征具有非常好的优点，由于利用了图像像素的内在变化，不受图像是如何获取的影响，如超声机器的增益，聚焦点等。在实施方式中，提取统计特征所针对的超声图像可以是生成的相关系数图像所使用的两个图像帧中任一者。
通过上述技术方案，利用根据超声图像生成的相关系数图像自动确定超声图像中的ROI，便于实现对超声图像中ROI的定位；利用对应于ROI的统计特征，实现了对ROI的分类，从而能够实现对超声图像准确解读。
图2是本发明一个优选实施例提供的超声图像处理方法流程图。如图2所示，在一个优选实施方式中，本发明实施例提供的超声图像处理方法可以包括：
S101，获取超声动态图像。
S102，基于所获取的超声动态图像生成相关系数图像。
S103，根据所生成的相关系数图像确定感兴趣区域ROI。
S104，根据所确定的ROI从所获取的超声图像中提取统计特征。
S105，利用所述统计特征训练预测模型。
S106，通过训练的预测模型从新获取的超声图像中确定存在所述ROI或存在所述ROI的概率。
在实施方式中，利用统计特征训练预测模型可以包括利用所述统计特征和机器学习算法训练所述预测模型。机器学习算法的示例可以包括但不限于决策树、逻辑回归等。需要说明的是，本发明的实施例可以采用任意机器学习算法或几种机器学习算法的组合。不同的机器学习算法在预测中可能具有不同的性能。而几种机器学习算法的组合可能在预测性能中优于其中任一种。
在实施方式中，经过很好训练的预测模型，例如使用包含多于一种的机器学习算法训练的预测模型，可对新获取的超声数据(例如，提取的统计特征)或超声图像进行是否存在ROI的判断，例如，可以给出存在ROI的概率。
本发明实施例的另一方面，提供一种超声动态图像处理装置。
图5是本发明实施例提供的超声图像处理装置组成示意图。如图5所示，该装置可以包括：图像获取单元51，用于获取超声动态图像；相关系数图像生成单元52，用于基于所获取的超声动态图像生成相关系数图像；区域确定单元53，用于根据所生成的相关系数图像确定感兴趣区域ROI；特征提取单元54，用于根据所确定的ROI从所获取的超声图像中提取统计特征。
在实施方式中，所获取的超声动态图像可以具有多个连续图像帧，相关系数图像生成单元52可以包括：图像帧选择模块521，用于从所述多个连续图像帧中选择相邻两个图像帧；相关系数计算模块522，用于计算所述相邻两个图像帧的相关系数以生成所述相关系数图像。
在实施方式中，区域确定单元53，可以用于将所述相关系数图像中相关系数变化率大于预设值的位置确定为所述ROI的边界。
在实施方式中，统计特征可以包括以下中至少一者：一阶特征、二阶特征、以及递差。
在实施方式中，上述装置还可以包括：模型训练单元55，用于利用所述统计特征训练预测模型；判断单元56，用于通过训练的预测模型从新获取的超声图像中确定存在所述ROI或存在所述ROI的概率。
在实施方式中，模型训练单元55，可以用于利用所述统计特征和机器学习算法训练所述预测模型。
本发明实施例的另一方面，提供一种超声摄像设备。
图6是本发明示例实施例的超声摄像设备的组成框图。如图6所示，本发明示例实施例提供的超声摄像设备可以包括：具备向被检体照射并接收超声波的振荡器元件的超声波探头3；收发超声波信号的超声波收发部4；基于接收信号来构成诸如二维超声图像(B模式超声图像)的超声图像构成部5；显示在超声图像构成部中构成的超声图像的显示部6；用于控制的控制部7；向控制部7提供指示的控制面板8；用于存储超声图像以及控制运算程序的存储部9；以及产生告警等声音的声音产生部10。在实施方式中，上述超声图像构成部5可以包括本发明上述实施例提供的超声动态图像处理装置。
在不同的实施方式中，本发明上述实施例提供的超声动态图像处理装置可以被设置与超声图像构成部5耦合以从该超声图像构成部获取超声图像。在可替换的实施方式中，本发明上述实施例提供的超声图像处理装置可以被设置与超声波收发部4耦合以从该超声波收发部4获取超声图像数据或信号。在上述实施方式中，显示部6和存储不9可以分别用于显示和存储来自超声图像处理装置的超声图像或相关系数图像。
本发明实施例的另一方面，提供一种机器可读存储介质，该机器可读 存储介质上存储有指令，该指令用于使得机器执行根据上述的方法。
在实施方式中，本发明实施例提供的机器可读存储介质可以在诸如超声摄像设备的设备中实施。
实施例
1、一种超声动态图像处理方法，其特征在于，该方法包括：
获取超声动态图像；
基于所获取的超声动态图像生成相关系数图像；
根据所生成的相关系数图像确定感兴趣区域ROI；
根据所确定的ROI从所获取的超声图像中提取统计特征。
2、根据实施例1所述的方法，其特征在于，从超声摄像设备获取所述超声图像。
3、根据实施例1所述的方法，其特征在于，所获取的超声动态图像具有多个连续图像帧，所述基于所获取的超声动态图像生成相关系数图像包括：
从所述多个连续图像帧中选择相邻两个图像帧；
计算所述相邻两个图像帧的相关系数以生成所述相关系数图像。
4、根据实施例1所述的方法，其特征在于，所述根据所生成的相关系数图像确定所述ROI包括：将所述相关系数图像中相关系数变化率大于预设值的位置确定为所述ROI的边界。
5、根据实施例4所述的方法，其特征在于，所述预设值为50％。
6、根据实施例4所述的方法，其特征在于，所述统计特征包括以下中至少一者：一阶特征、二阶特征、以及递差。
7、根据实施例1-6中任一项所述的方法，其特征在于，该方法还包括：
利用所述统计特征训练预测模型；
通过训练的预测模型从新获取的超声图像中确定存在所述ROI或存在 所述ROI的概率。
8、根据实施例7所述的方法，其特征在于，该方法包括：利用所述统计特征和机器学习算法训练所述预测模型。
9、根据实施例8所述的方法，其特征在于，所述机器学习算法包括以下中之一者：决策树或逻辑回归。
10、一种超声动态图像处理装置，其特征在于，该装置包括：
图像获取单元，用于获取超声动态图像；
相关系数图像生成单元，用于基于所获取的超声动态图像生成相关系数图像；
区域确定单元，用于根据所生成的相关系数图像确定感兴趣区域ROI；
特征提取单元，用于根据所确定的ROI从所获取的超声图像中提取统计特征。
11、根据实施例10所述的装置，其特征在于，所述图像获取单元，用于从超声摄像设备获取所述超声动态图像。
12、根据实施例10所述的装置，其特征在于，所获取的超声图像具有多个连续图像帧，所述相关系数图像生成单元包括：
图像帧选择模块，用于从所述多个连续图像帧中选择相邻两个图像帧；
相关系数计算模块，用于计算所述相邻两个图像帧的相关系数以生成所述相关系数图像。
13、根据实施例10所述的装置，其特征在于，所述区域确定单元，用于将所述相关系数图像中相关系数变化率大于预设值的位置确定为所述ROI的边界。
14、根据实施例13所述的装置，其特征在于，所述预设值为50％。
15、根据实施例13所述的装置，其特征在于，所述统计特征包括以下中至少一者：一阶特征、二阶特征、以及递差。
16、根据实施例10-15中任一项所述的装置，其特征在于，该装置还包 括：
模型训练单元，用于利用所述统计特征训练预测模型；
判断单元，用于通过训练的预测模型从新获取的超声图像中确定存在所述ROI或存在所述ROI的概率。
17、根据实施例16所述的装置，其特征在于，所述模型训练单元，用于利用所述统计特征和机器学习算法训练所述预测模型。
18、根据实施例17所述的装置，其特征在于，所述机器学习算法包括以下中之一者：决策树或逻辑回归。
19、一种超声摄像设备，其特征在于，该超声摄像设备包括根据实施例10-18中任一项所述的超声动态图像处理装置。
20、一种机器可读存储介质，该机器可读存储介质上存储有指令，该指令用于使得机器执行根据实施例1-9中任一项所述的方法。
以上结合附图详细描述了本发明实施例的可选实施方式，但是，本发明实施例并不限于上述实施方式中的具体细节，在本发明实施例的技术构思范围内，可以对本发明实施例的技术方案进行多种简单变型，这些简单变型均属于本发明实施例的保护范围。
另外需要说明的是，在上述具体实施方式中所描述的各个具体技术特征，在不矛盾的情况下，可以通过任何合适的方式进行组合。为了避免不必要的重复，本发明实施例对各种可能的组合方式不再另行说明。
本领域技术人员可以理解实现上述实施例方法中的全部或部分步骤是可以通过程序来指令相关的硬件来完成，该程序存储在一个存储介质中，包括若干指令用以使得单片机、芯片或处理器(processor)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质。
此外，本发明实施例的各种不同的实施方式之间也可以进行任意组合， 只要其不违背本发明实施例的思想，其同样应当视为本发明实施例所公开的内容。

Claims (14)

  1. 一种超声动态图像处理方法，其特征在于，该方法包括：
    获取超声动态图像；
    基于所获取的超声动态图像生成相关系数图像；
    根据所生成的相关系数图像确定感兴趣区域ROI；
    根据所确定的ROI从所获取的超声图像中提取统计特征。
  2. 根据权利要求1所述的方法，其特征在于，所获取的超声动态图像具有多个连续图像帧，所述基于所获取的超声动态图像生成相关系数图像包括：
    从所述多个连续图像帧中选择相邻两个图像帧；
    计算所述相邻两个图像帧的相关系数以生成所述相关系数图像。
  3. 根据权利要求1所述的方法，其特征在于，所述根据所生成的相关系数图像确定所述ROI包括：将所述相关系数图像中相关系数变化率大于预设值的位置确定为所述ROI的边界。
  4. 根据权利要求3所述的方法，其特征在于，所述统计特征包括以下中至少一者：一阶特征、二阶特征、以及递差。
  5. 根据权利要求1-4中任一项所述的方法，其特征在于，该方法还包括：
    利用所述统计特征训练预测模型；
    通过训练的预测模型从新获取的超声图像中确定存在所述ROI或存在所述ROI的概率。
  6. 根据权利要求5所述的方法，其特征在于，该方法包括：利用所述统计特征和机器学习算法训练所述预测模型。
  7. 一种超声动态图像处理装置，其特征在于，该装置包括：
    图像获取单元，用于获取超声动态图像；
    相关系数图像生成单元，用于基于所获取的超声动态图像生成相关系数图像；
    区域确定单元，用于根据所生成的相关系数图像确定感兴趣区域ROI；
    特征提取单元，用于根据所确定的ROI从所获取的超声图像中提取统计特征。
  8. 根据权利要求7所述的装置，其特征在于，所获取的超声动态图像具有多个连续图像帧，所述相关系数图像生成单元包括：
    图像帧选择模块，用于从所述多个连续图像帧中选择相邻两个图像帧；
    相关系数计算模块，用于计算所述相邻两个图像帧的相关系数以生成所述相关系数图像。
  9. 根据权利要求7所述的装置，其特征在于，所述区域确定单元，用于将所述相关系数图像中相关系数变化率大于预设值的位置确定为所述ROI的边界。
  10. 根据权利要求9所述的装置，其特征在于，所述统计特征包括以下中至少一者：一阶特征、二阶特征、以及递差。
  11. 根据权利要求7-10中任一项所述的装置，其特征在于，该装置还包括：
    模型训练单元，用于利用所述统计特征训练预测模型；
    判断单元，用于通过训练的预测模型从新获取的超声图像中确定存在所述ROI或存在所述ROI的概率。
  12. 根据权利要求11所述的装置，其特征在于，所述模型训练单元，用于利用所述统计特征和机器学习算法训练所述预测模型。
  13. 一种超声摄像设备，其特征在于，该超声摄像设备包括根据权利要求7-12中任一项所述的超声动态图像处理装置。
  14. 一种机器可读存储介质，该机器可读存储介质上存储有指令，该指令用于使得机器执行根据权利要求1-6中任一项所述的方法。
PCT/CN2017/103978 2017-09-28 2017-09-28 超声动态图像处理方法、装置及超声摄像设备 WO2019061148A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2017/103978 WO2019061148A1 (zh) 2017-09-28 2017-09-28 超声动态图像处理方法、装置及超声摄像设备

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2017/103978 WO2019061148A1 (zh) 2017-09-28 2017-09-28 超声动态图像处理方法、装置及超声摄像设备

Publications (1)

Publication Number Publication Date
WO2019061148A1 true WO2019061148A1 (zh) 2019-04-04

Family

ID=65900207

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/103978 WO2019061148A1 (zh) 2017-09-28 2017-09-28 超声动态图像处理方法、装置及超声摄像设备

Country Status (1)

Country Link
WO (1) WO2019061148A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113616945A (zh) * 2021-08-13 2021-11-09 湖北美睦恩医疗设备有限公司 一种基于聚焦超声图像识别的检测方法及美容美体装置

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040193053A1 (en) * 2003-03-27 2004-09-30 Sei Kato Ultrasonic imaging method and ultrasonic diagnostic apparatus
CN102549450A (zh) * 2009-10-07 2012-07-04 皇家飞利浦电子股份有限公司 超声无回声成像
WO2012135611A2 (en) * 2011-03-30 2012-10-04 Hitachi Aloka Medical, Inc. Methods and apparatus for ultrasound imaging
CN103377475A (zh) * 2012-04-27 2013-10-30 通用电气公司 血管内斑块的计算机诊断系统及方法
CN104203115A (zh) * 2012-03-30 2014-12-10 日立阿洛卡医疗株式会社 超声成像的方法和装置
CN105705098A (zh) * 2013-09-20 2016-06-22 透壁生物技术公司 用于诊断疾病的图像分析技术
CN106725592A (zh) * 2016-11-09 2017-05-31 飞依诺科技(苏州)有限公司 超声射频元数据的自动识别方法及系统
CN107072637A (zh) * 2014-09-25 2017-08-18 皇家飞利浦有限公司 用于自动气胸检测的设备和方法

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040193053A1 (en) * 2003-03-27 2004-09-30 Sei Kato Ultrasonic imaging method and ultrasonic diagnostic apparatus
CN102549450A (zh) * 2009-10-07 2012-07-04 皇家飞利浦电子股份有限公司 超声无回声成像
WO2012135611A2 (en) * 2011-03-30 2012-10-04 Hitachi Aloka Medical, Inc. Methods and apparatus for ultrasound imaging
CN104203115A (zh) * 2012-03-30 2014-12-10 日立阿洛卡医疗株式会社 超声成像的方法和装置
CN103377475A (zh) * 2012-04-27 2013-10-30 通用电气公司 血管内斑块的计算机诊断系统及方法
CN105705098A (zh) * 2013-09-20 2016-06-22 透壁生物技术公司 用于诊断疾病的图像分析技术
CN107072637A (zh) * 2014-09-25 2017-08-18 皇家飞利浦有限公司 用于自动气胸检测的设备和方法
CN106725592A (zh) * 2016-11-09 2017-05-31 飞依诺科技(苏州)有限公司 超声射频元数据的自动识别方法及系统

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113616945A (zh) * 2021-08-13 2021-11-09 湖北美睦恩医疗设备有限公司 一种基于聚焦超声图像识别的检测方法及美容美体装置

Similar Documents

Publication Publication Date Title
US10813595B2 (en) Fully automated image optimization based on automated organ recognition
RU2657855C2 (ru) Система трехмерной ультразвуковой визуализации
US20210177373A1 (en) Ultrasound system with an artificial neural network for guided liver imaging
JP4891038B2 (ja) 映像処理システム及び方法
CN110325119B (zh) 卵巢卵泡计数和大小确定
JP6890971B2 (ja) モデルベースセグメンテーションを用いた像撮像誘導
CN110678933B (zh) 超声临床特征检测与相关联的设备、系统和方法
US11100665B2 (en) Anatomical measurements from ultrasound data
JP6841907B2 (ja) Ｂラインを自動的に検出し、超音波スキャンの画像をスコア付けすることによる代表超音波画像の向上された視覚化および選択のための方法、システム及び非一時的コンピュータ可読媒体
US11488298B2 (en) System and methods for ultrasound image quality determination
US11069059B2 (en) Prenatal ultrasound imaging
KR102539901B1 (ko) 2차원 초음파 이미지를 음영화하는 방법 및 시스템
US11622743B2 (en) Rib blockage delineation in anatomically intelligent echocardiography
US20180161015A1 (en) Variable speed of sound beamforming based on automatic detection of tissue type in ultrasound imaging
US20120154400A1 (en) Method of reducing noise in a volume-rendered image
US11593933B2 (en) Systems and methods for ultrasound image quality determination
US20210077062A1 (en) Device and method for obtaining anatomical measurements from an ultrasound image
JP2005193017A (ja) 乳房患部分類の方法及びシステム
US20220237798A1 (en) Method and system for automatically estimating a hepatorenal index from ultrasound images
CN112773393A (zh) 用于通过基于超声图像分析自动调整波束形成器参数来提供超声图像增强的方法和系统
CN114554966A (zh) 用于图像优化的系统和方法
US20170086783A1 (en) Detecting amniotic fluid position based on shear wave propagation
WO2019061148A1 (zh) 超声动态图像处理方法、装置及超声摄像设备
US11890142B2 (en) System and methods for automatic lesion characterization
US11250564B2 (en) Methods and systems for automatic measurement of strains and strain-ratio calculation for sonoelastography

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: 17927435

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 17927435

Country of ref document: EP

Kind code of ref document: A1