Connect public, paid and private patent data with Google Patents Public Datasets

Muscle thickness measuring method and system based on ultrasonic image

Download PDF

Info

Publication number
CN103027713A
CN103027713A CN 201210563056 CN201210563056A CN103027713A CN 103027713 A CN103027713 A CN 103027713A CN 201210563056 CN201210563056 CN 201210563056 CN 201210563056 A CN201210563056 A CN 201210563056A CN 103027713 A CN103027713 A CN 103027713A
Authority
CN
Grant status
Application
Patent type
Prior art keywords
tracing
image
measuring
windows
method
Prior art date
Application number
CN 201210563056
Other languages
Chinese (zh)
Inventor
芦祎
李济舟
周永进
刘骏识
王磊
Original Assignee
中国科学院深圳先进技术研究院
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

Links

Abstract

The invention discloses a muscle thickness measuring method and system based on an ultrasonic image. The method comprises the following steps of: extracting an interesting image from the ultrasonic image; acquiring the positions of a plurality of initial tracing windows selected in the interesting image; tracing the tracing windows, and determining the positions of a plurality of tracing windows which correspond to every frame of subsequent image through a tracing algorithm; processing tracing windows which are similar to the modes of surrounding images in the tracing windows of each frame of image by taking a diagonal intersection point as a central point, and processing other tracing windows in every frame of image by adopting an edge detection method; and computing a maximum vertical distance between the position of each tracing window processed by using the central point in every frame of image and the position of each tracing window processed through the edge detection method, wherein the maximum vertical distance is taken as a muscle thickness value. Due to the adoption of the muscle thickness measuring method and the muscle thickness measuring system based on the ultrasonic image, the measuring accuracy and measuring efficiency are increased, and the aim of measuring in real time is fulfilled.

Description

基于超声图像的肌肉厚度测量方法和系统 Muscle thickness measurement method and system based on the ultrasound image

技术领域 FIELD

[0001] 本发明涉及图像处理领域,特别是涉及一种基于超声图像的肌肉厚度测量方法和系统。 [0001] The present invention relates to image processing, and more particularly to a method and system for measuring muscle thickness based on ultrasound images.

背景技术 Background technique

[0002] 骨骼肌的力学特性是和它的结构形态相关的,任何的身体活动和体育运动,都是由骨骼肌的收缩完成的,这直接影响了人体的力量和耐力。 [0002] mechanical properties of skeletal muscle and its morphology is related to any physical activity and sports, is done by the contraction of skeletal muscle, which directly affects the body's strength and endurance. 肌肉具有一定的弹性,被拉长后,当拉力解除时可自动恢复到原来的程度,肌肉的弹性可以减缓外力对人体的冲击,因而在运动中扮演着至关重要的作用。 Muscle has a certain elasticity after being stretched, when the tension is released automatically restored to its original level, muscle flexibility can slow the impact of external forces on the human body, and thus play a vital role in the movement. 而肌肉的构成又十分复杂,定量分析和评估肌肉功能状态是运动医学和运动功能康复研究中的难点和热点。 The muscle composition and is very complex, quantitative analysis and assessment of the state of muscle function is studied sports medicine and rehabilitation of motor function in difficult and hot.

[0003]目前对于肌肉的厚度的测量,如厚度的测量大部分采取的是人工手动测量,因手动测量对环境等诸多主观因素相当敏感,使得测量缺乏客观性,测量精度难以控制,并且对于测量大批量的肌肉厚度图片,操心过程费时费力,测量效率低。 [0003] At present, for measuring the thickness of the muscle, the majority of the thickness as measured is taken manually measure, due to manual measurement is quite sensitive to many environmental subjective factors, such lack of objectivity measurement, measurement accuracy is difficult to control, and the measurement large quantities of muscle thickness picture, worry about time-consuming process, low efficiency measure. 另外骨骼肌在运动过程中肌肉厚度的变化在每帧图像中比较细微,测量容易失真,从而影响测量结果。 Further changes in skeletal muscle during exercise muscle thickness is more subtle in each frame, the distortion measure easily, thus affecting the measurement results.

发明内容 SUMMARY

[0004] 基于此,有必要针对现有技术中测量效率低且测量不准确的问题,提供一种能提高测量准确度和测量效率的基于超声图像的肌肉厚度测量方法。 [0004] Based on this, it is necessary for the low efficiency of the prior art measurement and the measurement is not accurate, there is provided a method of measuring muscle thickness based on the ultrasound image measurement accuracy and measurement efficiency can be improved.

[0005] 此外,还有必要针对现有技术中`测量效率低且测量不准确的问题,提供一种能提高测量准确度和测量效率的基于超声图像的肌肉厚度测量系统。 [0005] In addition, the prior art there is a need for efficient and low `measurement problem of inaccurate measurement, improve muscle thickness to provide a measurement system based on the ultrasound image measurement accuracy and measurement efficiency.

[0006] 一种基于超声图像的肌肉厚度测量方法,包括以下步骤: [0006] A method based on the ultrasonic image measuring muscle thickness, comprising the steps of:

[0007] 从捕捉的超声图像中提取感兴趣图像; [0007] extracted from the ultrasound image of interest in image capture;

[0008] 获取在所述感兴趣图像中选择的多个初始跟踪窗口的位置; [0008] obtaining a plurality of initial tracking window position selected in the image of interest;

[0009] 对多个跟踪窗口进行跟踪,并通过跟踪算法确定后续每帧图像相应的多个跟踪窗口的位置; [0009] The plurality of tracking window for tracking and determining the subsequent frame image corresponding to each of the plurality of tracking position of the window by the tracking algorithm;

[0010] 对每帧图像的多个跟踪窗口中与周围图像模态相似的跟踪窗口采用取对角线交点为中心点进行处理,对每帧图像中其余跟踪窗口采用边缘检测法进行处理; [0010] The image for each frame of the plurality of the tracking window surrounding the image modality similar tracking window for the processing center point of intersection of the diagonals of each frame in the tracking window to rest using an edge detection process;

[0011] 计算每帧图像中经过中心点处理后的跟踪窗口的位置与每个经过边缘检测法处理后的跟踪窗口的位置之间的最大垂直距离,将所述最大垂直距离作为经过中心点处理后的跟踪窗口与经过边缘检测法处理后的跟踪窗口之间的肌肉厚度值。 [0011] After calculating the position of each image in the center point of each window tracking process after the maximum vertical distance between the position of the edge detection process of the tracking window, as the maximum vertical distance through the center point of the processing muscle thickness values ​​between the rear window and the track after the tracking window edge detection process.

[0012] 一种基于超声图像的肌肉厚度测量系统,包括: [0012] Based on the ultrasound image muscle thickness measurement system, comprising:

[0013] 提取模块,用于从捕捉的超声图像中提取感兴趣图像; [0013] extraction means for extracting an image of interest in the captured image from the ultrasound;

[0014] 获取模块,用于获取在所述感兴趣图像中选择的多个初始的跟踪窗口的位置; [0014] obtaining module, configured to obtain a plurality of initial tracking window position selected in the image of interest;

[0015] 跟踪模块,用于对多个跟踪窗口进行跟踪,并通过跟踪算法确定后续每帧图像相应的多个跟踪窗口的位置; [0015] tracking means for tracking a plurality of tracking window and determining each subsequent frame image corresponding to a plurality of position by the tracking window tracking algorithm;

[0016] 处理模块,用于对每帧图像的多个跟踪窗口中与周围图像模态相似的跟踪窗口采用取对角线交点为中心点进行处理,对每帧图像中其余跟踪窗口采用边缘检测法进行处理; [0016] The processing module is configured to use a plurality of track taking the intersection of diagonals of each frame image windows surrounding image modalities similar to the tracking window is processed as a center point, an edge detection using each frame image of the remaining trace window method for processing;

[0017] 计算模块,用于计算每帧图像中经过中心点处理后的跟踪窗口的位置与每个经过边缘检测法处理后的跟踪窗口的位置之间的最大垂直距离,将所述最大垂直距离作为经过中心点处理后的跟踪窗口与经过边缘检测法处理后的跟踪窗口之间的肌肉厚度值。 [0017] calculation means for calculating the position of each frame image through the center point of the trace window treated with the maximum vertical distance between the position of each track through a window after the edge detection process, the maximum vertical distance after muscle thickness values ​​as the center point between the window and the tracking process after the tracking process window edge detection method.

[0018] 上述基于超声图像的肌肉厚度测量方法和系统,通过对选取的多个跟踪窗口进行跟踪,通过跟踪算法确定在后续每帧图像中的多个跟踪窗口的位置,计算每帧图像中的经过中心点处理的跟踪窗口的位置与边缘检测法处理后的跟踪窗口的位置之间的最大垂直距离,作为肌肉厚度值,该测量方法基于超声图像,并采用图像算法进行修正处理得到的肌肉厚度值较为准确,提高了测量的准确度以及测量效率,且能跟踪后续每帧图像的跟踪窗口,并测量每帧图像中的肌肉厚度值,达到了实时测量的目的。 [0018] Based on the above-described thickness measuring method and muscular system of an ultrasound image, by a plurality of trace window selected track, determining the position of a plurality of windows are tracked in the subsequent frame image by each of the tracking algorithm, each frame image is calculated after the maximum vertical distance between the position and the processing position after the edge detection window tracking window center point tracking process, as a muscle thickness values, the measurement method based on the ultrasonic image, and muscular thickness of the image using the correction algorithm of the process to obtain more accurate values, improving the accuracy of measurement and a measurement of the efficiency, and can track follow-up of each image window, and measuring muscle thickness values ​​of each frame to achieve the purpose of real-time measurements.

附图说明 BRIEF DESCRIPTION

[0019] 图1为一个实施例中基于超声图像的肌肉厚度测量方法的流程示意图; [0019] Figure 1 is a flowchart of a method of measuring muscle thickness ultrasound image based on the embodiment of the schematic embodiment;

[0020] 图2为预处理后的超声图像; [0020] FIG. 2 is a pre-processed ultrasound image;

[0021] 图3为图像中跟踪窗口的界定与精确定位的示意图; [0021] FIG. 3 is a schematic view of the precise definition and positioning of the image tracking window;

[0022] 图4为对多个跟踪窗口进行跟踪,并通过跟踪算法确定后续图像帧的跟踪窗口的位置的流程示意图; [0022] FIG. 4 is a plurality of tracking window for tracking, and the process of tracking the position of the window is a schematic view of subsequent image frames is determined by the tracking algorithm;

[0023] 图5为一个实施例中基于超声图像的肌肉厚度测量系统的结构示意图; [0023] FIG. 5 is a schematic structural diagram of a measuring system according to the thickness of the muscle in the ultrasound image based on the embodiment;

[0024] 图6为一个实施例中跟踪模块的内部结构示意图; [0024] FIG. 6 is a schematic view of an internal structure of the tracking module embodiment;

[0025] 图7为另一个实施例中基于超声图像的肌肉厚度测量系统的结构示意图。 [0025] FIG. 7 is a schematic structural embodiment of muscle thickness measurement system based on another embodiment of the ultrasound image.

具体实施方式 detailed description

[0026] 下面结合具体的实施例及附图对基于超声图像的肌肉厚度测量方法和系统的技术方案进行详细的描述,以使其更加清楚。 [0026] The following examples and in conjunction with the accompanying drawings aspect muscle thickness measurement method and system based on the ultrasound image will be described in detail specific embodiments, in order to make it clearer.

[0027] 如图1所示,在一个实施例中,一种基于超声图像的肌肉厚度测量方法,包括以下步骤: [0027] As shown in Figure 1, in one embodiment, a method for measuring muscle thickness based on the ultrasound image, comprising the steps of:

[0028] 步骤S110,从超声图像中提取感兴趣图像。 [0028] step S110, the image of interest extracted from the ultrasound image.

[0029] 本实施例中,通过实时B型超声波扫描仪与一个电子线阵探头获取肌肉的超声图像。 [0029] In the present embodiment, the ultrasound images acquired with an electronic muscle linear array probe through real-time B-mode ultrasound scanner. 具体的,超声波探头的长轴方向垂直地被安排在实验者的大腿上,放置于约40%膝盖的长轴距离处。 Specifically, the major axis direction of the ultrasonic probe is arranged perpendicular to the thigh experimenter, placed at a distance about 40% of the major axis of the knee. 运用大量的超声凝胶确保探头与皮肤在肌肉收缩期间是声耦合的,调整探头以最优化对比度显示超声图像中的肌肉束。 Using a large number of ultrasound gel to ensure that the probe and the skin during muscle contraction is the acoustic coupling of the probe to adjust the contrast of the display to optimize the ultrasound image of muscle bundles. 采用B型超声波扫描仪获取超声图像并传送到视频捕获卡,由其进行数字化处理,并以速度约25帧/秒的采样率采集到计算机内数字化图像采集卡。 B-type ultrasonic scanner and acquiring ultrasound image to the video capture card, by digital processing, the sampling rate and at a rate of about 25 frames / sec is collected into a computer digital image acquisition card.

[0030] 对捕捉的超声图像进行裁剪得到感兴趣图像。 [0030] The ultrasound image to obtain the captured image cropping interest. 感兴趣图像即为包含有所需测量的肌肉厚度信息的图像。 Interest in the image is the thickness of the muscle with the image information necessary for the measurement.

[0031 ] 在Iv实施例中,在从捕捉的超声图像中提取感兴趣图像的步骤之后,还包括步骤:对该感兴趣图像进行预处理,包括:对所述感兴趣图像进行灰度变换及调整图像对比度。 [0031] In the embodiment Iv embodiment, after the step of extracting from the image of interest in the ultrasound image capturing, further comprising the step of: preprocessing the image of interest, comprising: an image of interest and the gradation conversion adjust the image contrast. 如图2所示为预处理后的超声图像。 Ultrasound image is shown in FIG. 2 after the pretreatment. [0032] 步骤S120,获取在该感兴趣图像中选择的多个初始的跟踪窗口的位置。 [0032] step S120, the acquired position of the plurality of initial tracking window selected in this interest in the image.

[0033] 具体的,首先手动在感兴趣图像中选择多个初始的跟踪窗口的位置。 [0033] Specifically, first manually selected position of a plurality of initial tracking windows in the image of interest. 本实施例中,多个为三个,可手动选择三个初始跟踪窗口,分别跟踪股骨、股直肌上部和股直肌下部边界,如图3所示,窗口A、B和C分别表示上述三个初始跟踪窗口。 In this embodiment, three, can manually select the three initial tracking windows, respectively, a lower track of the femur, the upper rectus femoris and rectus femoris boundary, as shown in FIG. 3 for the plurality of windows A, B and C represent the above three initial tracking window.

[0034] 步骤S130,对多个跟踪窗口进行跟踪,并通过跟踪算法确定后续图像帧的跟踪窗口的位置。 [0034] step S130, the tracking window for tracking a plurality of, and determine the position of the tracking window by subsequent image frames tracking algorithm.

[0035] 具体的,跟踪算法为压缩跟踪算法、互相关跟踪算法、形心跟踪算法、质心跟踪算法、波门跟踪算法、边缘跟踪算法、区域平衡跟踪算法等。 [0035] Specifically, the tracking algorithm is tracking algorithm compression, the cross-correlation tracking algorithm, the centroid tracking algorithm, the centroid tracking algorithm gate tracking algorithms, edge tracking algorithm, the tracking area balance algorithms. 互相关跟踪算法是基于图像的相似性度量,在当前图像中寻找最接近基准图像模板区域的一种跟踪算法,它对场景图像质量要求不高,不需分割目标和背景,对与选定的跟踪目标图像不相似的其他一切景物不敏感,能跟踪较小的目标以及目标区域的某一特殊部分或对比度比较差的目标,具有较强的局部抗干扰能力。 Cross-correlation tracking algorithm is based on the similarity measure of the image, looking for a tracking algorithm is closest to the reference image template region of the current image, the image quality of the scene do not ask it, no segmentation target and background, and selected all other tracking target image is not similar to the scene is not sensitive to small target track and can relatively poor contrast of a particular portion of the target area or target, with strong local anti-interference ability. 互相关算法将基准图像在当前图像上以不同的偏移值位置,根据测量两幅图像之间的相关度函数判断跟踪窗口在当前图像中的位置,跟踪窗口是两个图像匹配最好的位置,即相关函数的峰值。 Cross-correlation algorithm on the reference image to the current image position different offset values, determines the position of the tracking window in the current image based on the correlation function between the measured two images, the position of the tracking window is the best match of the two images , i.e., the peak correlation function.

[0036] 步骤S140,对每帧图像的多个跟踪窗口中与周围图像模态相似的跟踪窗口采用取对角线交点为中心点进行处理,对每帧图像中其余跟踪窗口采用边缘检测法进行处理。 [0036] step S140, the tracking of the plurality of windows in each frame and the surrounding image modality adopted similar tracing window intersection of the diagonals taken as a central point for processing, for each frame using the remaining trace window edge detection method deal with.

[0037] 其中,与周围图像模态相似是指窗口内的图像与其附近的图像很相似,通常通过先验知识来确定,在本例超声图像中靠近皮肤的那部分模态是相似的。 [0037] wherein the surrounding image is an image similar modality image within the window and its vicinity is very similar, is typically determined by a priori knowledge, close to the skin in an ultrasound image portion in the present embodiment mode are similar.

[0038] 如图3所示,由于与周围图像模态相似,跟踪窗口A采用取对角线交点为中心点的中心点法进行处理,跟踪窗口B和C采用边缘检测法进行处理,该边缘检测法可为canny算子的边缘检测法。 [0038] 3, due to the similarity with the surrounding image modality, window A tracking processing using the intersection of the diagonals taken as the center point of the center point of the method, the tracking window B and C are processed using edge detection, the edge assay may be a canny edge detection operator. 采用canny算子的边缘检测将窗口图像变换成为二进制图像,参数被调整以确保获得更多的组织细节,再运用最大连通区域搜索技术寻找每个窗口的确切边界。 Using canny operator edge detection window image converted into binary image, the parameters are adjusted to ensure organizations get more details, then use the largest connected area search technology to find the exact boundaries of each window.

[0039] 步骤S150,计算每帧图像中经过中心点处理后的跟踪窗口的位置与每个经过边缘检测法处理后的跟踪窗口的位置之间的最大垂直距离,将所述最大垂直距离作为经过中心点处理后的跟踪窗口与经过边缘检测法处理后的跟踪窗口之间的肌肉厚度值。 [0039] In step S150, the frame image is calculated for each position of the center point through the tracking process window maximum vertical distance between the position of each of the tracking window after the elapse of the edge detection process, as the maximum vertical distance through muscle thickness values ​​between the window and the center point of the trace window tracking process after the elapse of the edge detection process.

[0040] 具体的,以如图3中跟踪窗口A、B和C为例,计算每时刻每帧图像中跟踪窗口A与B之间的最大垂直距离,得到股直肌的厚度(Rectus femorisThickness, RFT),跟踪窗口A与C之间的最大垂直距离,得到股四头肌的厚度(QMT )。 [0040] Specifically, as shown in trace window to A, B and C, for example, calculates the tracking window is the maximum vertical distance between A and B in each frame each time, to obtain a thickness of the rectus muscle (Rectus femorisThickness shares, the maximum vertical distance between the RFT), the trace window a and C, giving the thickness of the quadriceps (QMT).

[0041] 上述基于超声图像的肌肉厚度测量方法,通过对选取的多个跟踪窗口进行跟踪,通过跟踪算法确定在后续每帧图像中的多个跟踪窗口的位置,计算每帧图像中的经过中心点处理的跟踪窗口的位置与边缘检测法处理后的跟踪窗口的位置之间的最大垂直距离,作为肌肉厚度值,该测量方法基于超声图像,并采用图像算法进行修正处理得到的肌肉厚度值较为准确,提高了测量的准确度以及测量效率,且能跟踪后续每帧图像的跟踪窗口,并测量每帧图像中的肌肉厚度值,达到了实时测量的目的。 [0041] The thickness of the muscle measuring method based on an ultrasound image, by selecting a plurality of track trace window, determining a position at each of the subsequent frame image by a plurality of trace window tracking algorithm to calculate the center of each image after the the maximum vertical distance between the position of the point of the position tracking process window edge detection processing window tracking, as muscles thickness value, the measurement method based on the ultrasonic image, and the thickness of the muscle algorithm using the image correction processing obtained more accuracy, improving the accuracy of measurement efficiency and measurement, and can track follow-up of each image window, and measuring muscle thickness values ​​of each frame to achieve the purpose of real-time measurements.

[0042] 进一步的,在一个实施例中,如图4所示,跟踪算法为压缩跟踪算法。 [0042] Further, in one embodiment, shown in Figure 4, the tracking algorithm compression tracking algorithm. 步骤S130具体为: Step S130 is specifically:

[0043] 步骤S131,对跟踪窗口所在帧图像进行采样,得到属于跟踪窗口位置范围内的样本集合。 [0043] Step S131, where the frame image of the tracking window is sampled to obtain samples belonging to the tracking window position set range.

[0044] 具体的,输入第t帧图像,对t帧图像的一系列图像片段进行采样,依据条件为:[0045] [0044] Specifically, the input t th frame image, a series of image segments t frame image is sampled, based on conditions: [0045]

Figure CN103027713AD00071

[0046] 其中,It_i是在第t_l时刻的跟踪位置; [0046] wherein, It_i is t_l tracking position at the time;

[0047] r是设定的度量当前图像与It_i之间差别的参数值,r越小,说明当前图像I(Z)与It-1相差越小; [0047] r is a measure of the difference between image parameter values ​​and the current set It_i, r is smaller, the smaller the difference of It-1 and the current image I (Z);

[0048] Dr是指在所有跟踪的位置中属于跟踪窗口位置范围内的像素点即正样本的集合; [0048] Dr means belonging to the set of pixels i.e., positive samples within the range of the tracking window position at all locations tracked;

[0049] I (Z)表示在t时刻获得的跟踪窗口的位置。 [0049] I (Z) represents a position tracking window obtained at time t.

[0050] 此外,通过采集靠近选择的跟踪窗口的正样本和远离跟踪窗口的负样本对分类器进行更新。 [0050] In addition, the classifier updated by tracking the acquisition window near a selected positive samples and negative samples of the window away from the track.

[0051] 对感兴趣的图像可采用较高分辨率进行滤波,对其它部分采取较低分辨率,以提高处理速度。 [0051] The image of interest can be filtered higher resolution, lower resolution take on other parts, in order to improve the processing speed.

[0052] 步骤S132,采用稀疏矩阵对样本集合中的每个样本进行降维处理,得到压缩特征向量。 [0052] step S132, the sparse matrix of each sample in the sample set dimensionality reduction, feature vector is compressed.

[0053] 具体的,稀疏矩阵为引入的矩阵,如随机投影V=RX,其中,R是一个随机矩阵,Re Rnxm,其中m>n,运用该公式可以将m维的向量X降维到n维的向量,从而达到降维的作用,V即为压缩特征向量。 [0053] Specifically, the sparse matrix is ​​a matrix incorporated, such as random projection V = RX, wherein, R is a random matrix, Re Rnxm, wherein the dimension m> n, the use of this formula may be an m-dimensional vector X down to n dimensional vector, so as to achieve the effect of dimension reduction, V is the feature vector compression. 对于每一个样本z G Rm,它的低维表示为v=( u p. . .,Un)T e Rn,且需满足m >> n。 For each sample z G Rm, its low dimensional representation of v = (u p..., Un) T e Rn, and the need to satisfy m >> n.

[0054] 随机矩阵的选择依据如下: [0054] The random matrix selection according to the following:

[0055] 首先选择稳定的投影矩阵,为了确保信号的线性投影能够保持信号的原始结构,投影矩阵必须满足约束等距性(Restricted isometry property,RIP)条件,然后通过原始信号与测量矩阵的乘积获得原始信号的线性投影测量。 [0055] First selection of stable projection matrix, to ensure that the original structure of the linear projection signals can maintain the signal, the matrix must satisfy the constraints isometric projection of (Restricted isometry property, RIP) condition, then the original signal is obtained by the product of the measurement matrix measuring the linear projection of the original signal. 此处选取的是随机高斯矩阵R,当TijKo, I), RG Rnxm, Chosen here is a Gaussian random matrix R, when TijKo, I), RG Rnxm,

Figure CN103027713AD00072

[0057] 其中,!Tij表示该随机矩阵R中的元素,p表示概率值,s = 2或s = 3,此时满足Johnson-Lindenstrauss 定理。 [0057] wherein,! Tij represents the elements in the random matrix R, p denotes the probability value, s = 2 or s = 3, this time to meet the Johnson-Lindenstrauss theorem.

[0058] 步骤S133,对压缩特征向量采用分类器进行分类。 [0058] step S133, the compressed feature vectors are classified using a classifier.

[0059] 具体的,对压缩特征向量采用朴素贝叶斯分类器分类,且分类器中的条件概率满足高斯正态分布。 [0059] Specifically, the compressed feature vectors, naive Bayes classifier classified, and the conditional probability classifier a Gaussian normal distribution.

[0060] 向量中所有元素都被假定为相互独立。 [0060] All elements of the vector are assumed to be mutually independent. 当p(y = I) = p (y = 0)时,对每一个压缩特征矢量使用朴素贝叶斯分类器分类。 When p (y = I) = p (y = 0), for each feature vector compression using naive Bayes classifier classified. 计算公式如下: Calculated as follows:

W、. ,Ft-,,叫v=1、 W ,., Ft- ,, called v = 1,

[0061 ] [0061]

Figure CN103027713AD00073

[0062] y G {0, 1}是一个二进制随机变量,用来表示样本标签。 [0062] y G {0, 1} is a binary random variable, is used to represent a sample label. 分类器H(V)中的条件概率p (Vi |y = I)和POil |y = 0)被假定为满足参数为的高斯正态分布,且 Classifier H (V) in the conditional probability p (Vi | y = I) and POil | y = 0) is assumed to satisfy the Gaussian normal distribution parameter, and

[0063] [0063]

Figure CN103027713AD00074

[0064] 步骤S134,从样本集合中进行抽样得到两组图像样本。 [0064] step S134, the sample sets sampled from the obtained sets of image samples.

[0065] 具体的,抽样满足且满足a < 4 P,抽样条件分别为: [0065] Specifically, the sampling meet and satisfy a <4 P, sampling conditions were:

[0066] Da = Iz II(Z)-1t I < a },D5'0 = {z| “| I (z)_It | < ,其中IT 和D5'0为两组图像样本; [0066] Da = Iz II (Z) -1t I <a}, D5'0 = {z | "| I (z) _It | <, wherein IT and D5'0 image samples into two groups;

[0067] a,(,@是设定的度量当前图像I (Z)与It之间差别的参数值。 [0067] a, (, @ parameter value setting measure the current image I (Z) and the difference between the It.

[0068] 步骤S135,对所述两组图像样本提取哈尔特征,并采用所述分类器迭代得到相邻的后续图像帧的跟踪窗口的位置。 [0068] step S135, the feature extraction Haar the image sample groups, and the position of the classifier using the iterative tracking window adjacent to obtain a subsequent image frame.

[0069] 具体的,迭代算法公式为: [0069] In particular, the iterative algorithm formula is:

[0070] [0070]

Figure CN103027713AD00081

[0071] 采用公式(2)中的参数对分类器迭代更新,得到第t+1帧图像的跟踪窗口的位置和分类器参数。 [0071] using the equation (2) iteratively updating the parameters of the classifier, to obtain the t + classifier parameter and the position of a tracking window image.

[0072] 如图5所示,在一个实施例中,一种基于超声图像的肌肉厚度测量系统,包括提取模块110、获取模块120、跟踪模块130、处理模块140和计算模块150。 [0072] As shown in FIG. 5, in one embodiment, an ultrasound image based on the muscular thickness measurement system comprising extraction module 110, an obtaining module 120, a tracking module 130, a processing module 140 and a calculation module 150. 其中: among them:

[0073] 提取模块110用于从捕捉的超声图像中提取感兴趣图像。 [0073] The extraction module 110 for extracting an image of interest in the captured image from the ultrasound. 本实施例中,通过实时B型超声波扫描仪与一个电子线阵探头获取肌肉的超声图像。 In this embodiment, the ultrasound images acquired with an electronic muscle linear array probe through real-time B-mode ultrasound scanner. 具体的,超声波探头的长轴方向垂直地被安排在实验者的大腿上,放置于约40%膝盖的长轴距离处。 Specifically, the major axis direction of the ultrasonic probe is arranged perpendicular to the thigh experimenter, placed at a distance about 40% of the major axis of the knee. 运用大量的超声凝胶确保探头与皮肤在肌肉收缩期间是声耦合的,调整探头以最优化对比度显示超声图像中的肌肉神经束。 Using a large number of ultrasound gel to ensure that the probe and the skin during muscle contraction is the acoustic coupling of the probe to adjust the contrast of the display to optimize muscle nerve bundle in the ultrasound image. 采用B型超声波扫描仪获取超声图像并传送到视频捕获卡,由其进行数字化处理,并以速度约25帧/ 秒的采样率采集到计算机内数字化图像采集卡。 B-type ultrasonic scanner and acquiring ultrasound image to the video capture card, by digital processing, the sampling rate and at a rate of about 25 frames / sec is collected into a computer digital image acquisition card. 对捕捉的超声图像进行裁剪得到感兴趣图像。 Ultrasound image capture image cropping get interested. 感兴趣图像即为包含有所需测量的肌肉厚度信息的图像。 Interest in the image is the thickness of the muscle with the image information necessary for the measurement.

[0074] 获取模块120获取在该感兴趣图像中选择的多个初始的跟踪窗口的位置。 [0074] The location acquiring module 120 acquires a plurality of initial tracking window selected in this interest in the image.

[0075] 具体的,首先手动在感兴趣图像中选择多个初始的跟踪窗口的位置。 [0075] Specifically, first manually selected position of a plurality of initial tracking windows in the image of interest.

[0076] 本实施例中,多个为三个,可手动选择三个初始跟踪窗口,分别跟踪股骨、股直肌上部和股直肌下部边界,如图3所示,窗口A、B和C分别表示上述三个初始跟踪窗口。 [0076] In this embodiment, three, can manually select the three initial tracking windows, track the femur, the upper rectus femoris rectus femoris and the lower boundary, as shown in FIG window A, B, and C 3 more of respectively represent the above three initial tracking window.

[0077] 跟踪模块130用于对多个跟踪窗口进行跟踪,并通过跟踪算法确定后续图像帧的跟踪窗口的位置。 [0077] The tracking module 130 for tracking a plurality of tracking window, and subsequent image frames to determine the tracking window by a position tracking algorithm.

[0078] 具体的,跟踪算法为压缩跟踪算法、互相关跟踪算法、形心跟踪算法、质心跟踪算法、波门跟踪算法、边缘跟踪算法、区域平衡跟踪算法等。 [0078] Specifically, the tracking algorithm is tracking algorithm compression, the cross-correlation tracking algorithm, the centroid tracking algorithm, the centroid tracking algorithm gate tracking algorithms, edge tracking algorithm, the tracking area balance algorithms. 互相关跟踪算法是基于图像的相似性度量,在当前图像中寻找最接近基准图像模板区域的一种跟踪算法,它对场景图像质量要求不高,不需分割目标和背景,对与选定的跟踪目标图像不相似的其他一切景物不敏感,能跟踪较小的目标以及目标区域的某一特殊部分或对比度比较差的目标,具有较强的局部抗干扰能力。 Cross-correlation tracking algorithm is based on the similarity measure of the image, looking for a tracking algorithm is closest to the reference image template region of the current image, the image quality of the scene do not ask it, no segmentation target and background, and selected all other tracking target image is not similar to the scene is not sensitive to small target track and can relatively poor contrast of a particular portion of the target area or target, with strong local anti-interference ability. 互相关算法将基准图像在当前图像上以不同的偏移值位置,根据测量两幅图像之间的相关度函数判断跟踪窗口在当前图像中的位置,跟踪窗口是两个图像匹配最好的位置,即相关函数的峰值。 Cross-correlation algorithm on the reference image to the current image position different offset values, determines the position of the tracking window in the current image based on the correlation function between the measured two images, the position of the tracking window is the best match of the two images , i.e., the peak correlation function.

[0079] 处理模块140用于对每帧图像的多个跟踪窗口中与周围图像模态相似的跟踪窗口采用取对角线交点为中心点进行处理,对每帧图像中其余跟踪窗口采用边缘检测法进行处理。 [0079] The processing module 140 is employed for taking a plurality of intersection of the diagonals of each image tracking window surrounding the image modality with similar processing for the center window tracking using edge detection of each image in the remaining trace window method for processing.

[0080] 如图3所示,与周围图像模态相似的跟踪窗口A采用取对角线交点为中心点的中心点法进行处理,跟踪窗口B和C采用边缘检测法进行处理,该边缘检测法可为canny算子的边缘检测法。 [0080] As shown in FIG. 3, the surrounding image modality A similar tracking window is processed using the intersection of the diagonals taken as the center point of the center point of the method, the tracking window B and C are processed using edge detection, the edge detection method may be canny edge detection operator. 采用canny算子的边缘检测将窗口图像变换成为二进制图像,参数被调整以确保获得更多的组织细节,再运用最大连通区域搜索技术寻找每个窗口的确切边界。 Using canny operator edge detection window image converted into binary image, the parameters are adjusted to ensure organizations get more details, then use the largest connected area search technology to find the exact boundaries of each window.

[0081] 计算模块150用于计算经过边缘检测法处理后的后续图像帧的跟踪窗口的位置与经过中心点处理后的初始的跟踪窗口的位置之间的最大垂直距离,将所述最大垂直距离作为肌肉厚度值。 [0081] After calculation module 150 for calculating the position of the tracking window subsequent image frames processed through the edge detection method and the maximum vertical distance between the center point position of the processing of the initial tracking window, the maximum vertical distance as a muscle thickness values.

[0082] 具体的,以如图3中跟踪窗口A、B和C为例,计算每时刻每帧图像中跟踪窗口A与B之间的最大垂直距离,得到股直肌的厚度(Rectus femorisThickness, RFT),跟踪窗口A与C之间的最大垂直距离,得到股四头肌的厚度(QMT )。 [0082] Specifically, as shown in trace window to A, B and C, for example, calculates the tracking window is the maximum vertical distance between A and B in each frame each time, to obtain a thickness of the rectus muscle (Rectus femorisThickness shares, the maximum vertical distance between the RFT), the trace window a and C, giving the thickness of the quadriceps (QMT).

[0083] 上述基于超声图像的肌肉厚度测量系统,通过对选取的多个跟踪窗口进行跟踪,通过跟踪算法确定在后续每帧图像中的多个跟踪窗口的位置,计算每帧图像中的经过中心点处理的跟踪窗口的位置与边缘检测法处理后的跟踪窗口的位置之间的最大垂直距离,作为肌肉厚度值,该测量方法基于超声图像,并采用图像算法进行修正处理得到的肌肉厚度值较为准确,提高了测量的准确度以及测量效率,且能跟踪后续每帧图像的跟踪窗口,并测量每帧图像中的肌肉厚度值,达到了实时测量的目的。 [0083] The muscle thickness measuring system based on an ultrasound image, by passing through the center of each frame image of the plurality of trace window selected track, determining a position at each of the subsequent frame image by a plurality of trace window tracking algorithm to calculate the maximum vertical distance between the position of the point of the position tracking process window edge detection processing window tracking, as muscles thickness value, the measurement method based on the ultrasonic image, and the thickness of the muscle algorithm using the image correction processing obtained more accuracy, improving the accuracy of measurement efficiency and measurement, and can track follow-up of each image window, and measuring muscle thickness values ​​of each frame to achieve the purpose of real-time measurements.

[0084] 在一个实施例中,跟踪算法为压缩跟踪算法时,如图6所示,跟踪模块130包括采样模块131、降维模块132、分类模块133、抽样模块134和迭代模块135。 [0084] In one embodiment, when the tracking algorithm is tracking algorithm compression, as shown in Figure 6, tracking module 130 includes a sampling module 131, a dimension reduction module 132, a classification module 133, a sampling module 134 and module 135 iterations. 其中: among them:

[0085] 采样模块131用于对跟踪窗口所在的帧图像进行采样,得到属于跟踪窗口位置范围内的样本集合。 [0085] The sampling module 131 for the frame image where the tracking window is sampled to obtain samples belonging to the set position of the tracking window range.

[0086] 具体的,输入第t帧图像,对t帧图像的一系列图像片段进行采样,依据条件为: [0086] Specifically, the input t th frame image, a series of image segments t frame image is sampled, based on conditions:

[0087] [0087]

Figure CN103027713AD00091

[0088] [0088]

Figure CN103027713AD00092

其中,It^1是在第t_l时刻的跟踪位置; Wherein, It ^ 1 is t_l tracking position at the time;

[0089] r是设定的度量当前图像与Iw之间差别的参数值,r越小,说明当前图像I(Z)与It-1相差越小 [0089] r is set to measure the difference between the current image and the parameter values ​​Iw, the smaller the r, described current image I (Z) smaller the difference between the It-1

[0090] Dr是指在所有跟踪的位置中属于跟踪窗口位置范围内的像素点即正样本的集合; [0090] Dr means belonging to the set of pixels i.e., positive samples within the range of the tracking window position at all locations tracked;

[0091] I (Z)表不在t时刻犾得的跟踪窗口的位直。 Straight tracking window position [0091] I (Z) table is not time t l is obtained.

[0092] 此外,通过采集靠近选择的跟踪窗口的正样本和远离跟踪窗口的负样本对分类器进行更新。 [0092] In addition, the classifier updated by tracking the acquisition window near a selected positive samples and negative samples of the window away from the track.

[0093] 对感兴趣的图像可采用较高分辨率进行滤波,对其它部分采取较低分辨率,以提高处理速度。 [0093] The image of interest can be filtered higher resolution, lower resolution take on other parts, in order to improve the processing speed.

[0094] 降维模块132用于采用稀疏矩阵对样本集合中的每个样本进行降维处理,得到压缩特征向量。 [0094] The dimension reduction module 132 sparse matrix for each sample in the sample set dimensionality reduction, feature vector is compressed.

[0095] 具体的,稀疏矩阵为引入的矩阵,如随机投影V=RX,其中,R是一个随机矩阵,Re Rnxm,其中m>n,运用该公式可以将m维的向量X降维到n维的向量,从而达到降维的作用,V即为压缩特征向量。 [0095] Specifically, the sparse matrix is ​​a matrix incorporated, such as random projection V = RX, wherein, R is a random matrix, Re Rnxm, wherein the dimension m> n, the use of this formula may be an m-dimensional vector X down to n dimensional vector, so as to achieve the effect of dimension reduction, V is the feature vector compression. 对于每一个样本z G Rm,它的低维表示为v=( u p. . .,Un)T e Rn,且需满足m >> n。 For each sample z G Rm, its low dimensional representation of v = (u p..., Un) T e Rn, and the need to satisfy m >> n.

[0096] 随机矩阵的选择依据如下: [0096] random matrix selection according to the following:

[0097] 首先选择稳定的投影矩阵,为了确保信号的线性投影能够保持信号的原始结构,投影矩阵必须满足约束等距性(Restricted isometry property,RIP)条件,然后通过原始信号与测量矩阵的乘积获得原始信号的线性投影测量。 [0097] The first selection of stable projection matrix, to ensure that the original structure of the linear projection signals can maintain the signal, the matrix must satisfy the constraints isometric projection of (Restricted isometry property, RIP) condition, then the original signal is obtained by the product of the measurement matrix measuring the linear projection of the original signal. 此处选取的是随机高斯矩阵R,当TijKO, I), RG RnXm, Chosen here is a Gaussian random matrix R, when TijKO, I), RG RnXm,

Figure CN103027713AD00101

[0099] 其中,r^-表示该随机矩阵R中的元素,p表示概率值,s = 2或s = 3,此时满足Johnson-Lindenstrauss 定理。 [0099] wherein, r ^ - represents the elements in the random matrix R, p denotes the probability value, s = 2 or s = 3, this time to meet the Johnson-Lindenstrauss theorem.

[0100] 分类模块133用于对所述压缩特征向量采用分类器进行分类。 [0100] The classification module 133 for using the compressed feature vector classifier for classification.

[0101] 具体的,对压缩特征向量采用朴素贝叶斯分类器分类,且分类器中的条件概率满足高斯正态分布。 [0101] Specifically, the compressed feature vectors, naive Bayes classifier classified, and the conditional probability classifier a Gaussian normal distribution.

[0102] 向量中所有元素都被假定为相互独立。 [0102] All elements of the vector are assumed to be mutually independent. when

Figure CN103027713AD00102

时,对每一个压缩特征矢量使用朴素贝叶斯分类器分类。 When, for each feature vector compression using naive Bayes classifier classified. 计算公式如下: Calculated as follows:

Figure CN103027713AD00103

[0104] y G {0,1}是一个二进制随机变量,用来表示样本标签。 [0104] y G {0,1} is a binary random variable, is used to represent a sample label. 分类器H(V)中的条件概率P(vjy = I)和PGilIy = 0)被假定为满足参数为 Conditional probability P (vjy = I) classifier H (V) and the PGilIy = 0) is assumed to meet the parameters

Figure CN103027713AD00104

,的高斯正态分布,且 , Gaussian normal distribution, and

[0105] [0105]

Figure CN103027713AD00105

[0106] 抽样模块134用于从所述样本集合中进行抽样得到两组图像梓本。 [0106] Sampling module 134 is used to obtain two sampled image from the present Zi sample set.

[0107] 具体的,抽样满足且满足a < ( < 0,抽样条件分别为: [0107] Specifically, the sampling meet and satisfy a <(<0, the sampling conditions were:

[0108] [0108]

Figure CN103027713AD00106

,其中Da 和D5'0为两组图像样本;a,40是设定的度量当前图像I (Z)与It之间差别的参数值。 Wherein Da and D5'0 image samples into two groups; a, 40 is a measure of the current image I (Z) and the difference between the parameter value It setting.

[0109] 迭代模块135用于对所述两组图像样本提取哈尔特征,并采用分类器迭代得到相邻的后续图像帧的跟踪窗口的位置。 [0109] Iterative Haar feature extraction module 135 for the two sets of image samples, and using a position adjacent to the classifier to obtain tracking iteration of subsequent image frames of the window.

[0110] 具体的,迭代算法公式为: [0110] In particular, the iterative algorithm formula is:

[0111] [0111]

Figure CN103027713AD00107

[0112] 采用公式(2)中的参数对分类器迭代更新,得到第t+1帧图像的跟踪窗口的位置和分类器参数。 [0112] using the equation (2) iteratively updating the parameters of the classifier, to obtain the t + classifier parameter and the position of a tracking window image.

[0113] 如图7所示,在一个实施例中,上述基于超声图像的肌肉厚度测量系统,除了包括提取模块110、获取模块120、跟踪模块130、处理模块140和计算模块150,还包括预处理模块160。 [0113] As shown in FIG. 7, in one embodiment, the muscle thickness measurement system based on ultrasound images, in addition to including extraction module 110, an obtaining module 120, a tracking module 130, a processing module 140 and a calculation module 150, further comprising a pre- processing module 160. 其中: among them:

[0114] 预处理模块160用于对所述感兴趣图像进行预处理,所述预处理包括对所述感兴趣图像进行灰度变换及调整图像对比度。 [0114] The preprocessing module 160 to preprocess the image of interest, comprising the pretreatment of the image of interest and the gradation conversion adjust the image contrast.

[0115] 以上所述实施例仅表达了本发明的几种实施方式,其描述较为具体和详细,但并不能因此而理解为对本发明专利范围的限制。 [0115] Expression of the above-described embodiments are only several embodiments of the present invention, and detailed description thereof is more specific, but can not therefore be understood as limiting the scope of the present invention. 应当指出的是,对于本领域的普通技术人员来说,在不脱离本发明构思的前提下,还可以做出若干变形和改进,这些都属于本发明的保护范围。 It should be noted that those of ordinary skill in the art, without departing from the spirit of the present invention, can make various changes and modifications, which fall within the protection scope of the present invention. 因此,本发明专利的保护范围应以所附权利要求为准。 Therefore, the protection scope of the present invention should be subject to the appended claims.

Claims (10)

1. 一种基于超声图像的肌肉厚度测量方法,包括以下步骤: 从超声图像中提取感兴趣图像; 获取在所述感兴趣图像中选择的多个初始跟踪窗口的位置; 对多个跟踪窗口进行跟踪,并通过跟踪算法确定后续每帧图像相应的多个跟踪窗口的位置; 对每帧图像的多个跟踪窗口中与周围图像模态相似的跟踪窗口采用取对角线交点为中心点进行处理,对每帧图像中其余跟踪窗口采用边缘检测法进行处理; 计算每帧图像中经过中心点处理后的跟踪窗口的位置与每个经过边缘检测法处理后的跟踪窗口的位置之间的最大垂直距离,将所述最大垂直距离作为经过中心点处理后的跟踪窗口与经过边缘检测法处理后的跟踪窗口之间的肌肉厚度值。 CLAIMS 1. A method of measuring muscle thickness based on an ultrasound image, comprising the steps of: extracting an image of interest from the ultrasound image; an image acquiring position selected in the interest of the plurality of initial tracking windows; a plurality of track windows track, and the subsequent determination of each frame corresponding to a plurality of position by the tracking window tracking algorithm; for each image using a plurality of tracking window surrounding the image modality similar tracking window for the processing center point of intersection of the diagonals , for each frame using the remaining trace window edge detection processing; calculated for each frame of image processing through the center point position of the tracking window between the maximum vertical position of each of the tracking window after the elapse of the edge detection process distance, as the maximum vertical distance through the center point of the trace window muscle thickness value after processing between the processed trace window edge detection method.
2.根据权利要求1所述的基于超声图像的肌肉厚度测量方法,其特征在于,所述跟踪算法为压缩跟踪算法或互相关跟踪算法。 The method of measuring the thickness of the muscle is based on an ultrasound image as claimed in claim, wherein the compression algorithm is a tracking algorithm to track or cross-correlation tracking algorithm.
3.根据权利要求1所述的基于超声图像的肌肉厚度测量方法,其特征在于,所述跟踪算法为压缩跟踪算法; 所述对多个跟踪窗口进行跟踪,并通过跟踪算法确定后续每帧图像相应的多个跟踪窗口的位置的步骤为: 对跟踪窗口所在帧图像进行采样,得到属于所述跟踪窗口位置范围内的样本集合; 采用稀疏矩阵对样本集合中的每个样本进行降维处理,得到压缩特征向量; 对所述压缩特征向量采用分类器进行分类; 从所述样本集合中进行抽样得到两组图像样本; 对所述两组图像样本提取哈尔特征,并采用所述分类器迭代得到后续相邻帧图像相应的跟踪窗口的位置。 The method of measuring the thickness of the muscle is based on an ultrasound image as claimed in claim, wherein the tracking algorithm is tracking a compression algorithm; trace window of the plurality of track, and each frame is determined by the subsequent tracking algorithm a plurality of step positions corresponding trace window is: where the frame image of the tracking window is sampled to obtain samples belonging to the tracking range of positions within the window set; sparse matrix of each sample in the sample set dimensionality reduction, feature vector is compressed; the compressed feature vectors are classified using a classifier; sample sets sampled from the obtained sets of image samples; Hal wherein the extracting sets of image sample, and using the classification iterative subsequent neighboring frame image to obtain a position corresponding to the tracking window.
4.根据权利要求3所述的基于超声图像的肌肉厚度测量方法,其特征在于,所述对所述压缩特征向量采用分类器进行分类的步骤包括: 对所述压缩特征向量采用朴素贝叶斯分类器分类,且分类器中的条件概率满足高斯正态分布。 4. A method of measuring muscle thickness based on an ultrasound image, characterized in that said according to claim 3, the classifier using the feature vector compression step of classifying comprises: the compressed feature vectors, and Naive Bayes classifier classification, conditional probability and the classification of a Gaussian normal distribution.
5.根据权利要求1所述的基于超声图像的肌肉厚度测量方法,其特征在于,在所述从捕捉的超声图像中提取感兴趣图像的步骤之后,还包括步骤: 对所述感兴趣图像进行预处理,包括: 对所述感兴趣图像进行灰度变换及调整图像对比度。 The measuring method of an ultrasound image based on the thickness of the muscle, characterized in that said 1, after the step of extracting from the image of interest in the ultrasound image capturing, further comprising the step of claim: for an image of interest pretreatment, comprising: an image of interest to adjust the gradation conversion and image contrast.
6. 一种基于超声图像的肌肉厚度测量系统,其特征在于,包括: 提取模块,用于从捕捉的超声图像中提取感兴趣图像; 获取模块,用于获取在所述感兴趣图像中选择的多个初始的跟踪窗口的位置; 跟踪模块,用于对多个跟踪窗口进行跟踪,并通过跟踪算法确定后续每帧图像相应的多个跟踪窗口的位置; 处理模块,用于对每帧图像的多个跟踪窗口中与周围图像模态相似的跟踪窗口采用取对角线交点为中心点进行处理,对每帧图像中其余跟踪窗口采用边缘检测法进行处理; 计算模块,用于计算每帧图像中经过中心点处理后的跟踪窗口的位置与每个经过边缘检测法处理后的跟踪窗口的位置之间的最大垂直距离,将所述最大垂直距离作为经过中心点处理后的跟踪窗口与经过边缘检测法处理后的跟踪窗口之间的肌肉厚度值。 A muscular thickness measuring system based on an ultrasound image, wherein, comprising: extracting means for extracting an image of interest in the captured image from the ultrasound; acquiring module, for acquiring an image of interest in selecting the the initial position of the plurality of tracking window; tracking means for tracking a plurality of tracking window and determining the position of a corresponding plurality of tracking subsequent windows each frame by tracking algorithm; processing module for each image a plurality of tracking window surrounding the image modality adopted similar tracing window intersection of the diagonals taken as a central point for processing, for each frame in the tracking window to rest using an edge detection processing; calculating module, for calculating each image after the position of the center point of the trace window treated with the maximum vertical distance between the position of each track through a window after the edge detection process, as the maximum vertical distance through the center point of the trace window edge after treatment with muscle thickness values ​​between trace window treated assay.
7.根据权利要求6所述的基于超声图像的肌肉厚度测量系统,其特征在于,所述跟踪算法为压缩跟踪算法或互相关跟踪算法。 7. muscle thickness measurement system based on the ultrasonic image according to claim 6, characterized in that the tracking algorithm is tracking algorithm compression or cross-correlation tracking algorithm.
8.根据权利要求6所述的基于超声图像的肌肉厚度测量系统,其特征在于,所述跟踪算法为压缩跟踪算法; 所述跟踪模块包括: 采样模块,用于对跟踪窗口所在帧图像进行采样,得到属于跟踪窗口位置范围内的样本集合; 降维模块,用于采用稀疏矩阵对样本集合中的每个样本进行降维处理,得到压缩特征向量; 分类模块,用于对所述压缩特征向量采用分类器进行分类; 抽样模块,用于从所述样本集合中进行抽样得到两组图像样本; 迭代模块,用于对所述两组图像样本提取哈尔特征,并采用分类器迭代得到相邻的后续图像帧的跟踪窗口的位置。 The muscle thickness measuring system based on the ultrasonic image according to claim 6, characterized in that the tracking algorithm is tracking a compression algorithm; the tracking module comprises: a sampling module for tracking window frame image where the sample give samples belonging to a range of positions within the tracking window set; dimension reduction module for sparse matrix sample set for each sample dimensionality reduction, feature vector is compressed; classification module, the compressed feature vector for using the classifier for classification; sampling means for sampling the set of samples obtained from the two sets of image samples; iteration means for extracting the feature sets of image Hal sample, and using a classifier to give an iterative tracking a position adjacent a window of a subsequent image frame.
9.根据权利要求8所述的基于超声图像的肌肉厚度测量系统,其特征在于,所述分类模块还用于对所述压缩特征向量采用朴素贝叶斯分类器分类,且分类器中的条件概率满足高斯正态分布。 9. The measurement system based on the muscular thickness of the ultrasound image of claim 8, wherein said classification module is further configured to use the compressed feature vector classification naive Bayes classifier and classifiers conditions the probability of a Gaussian normal distribution.
10.根据权利要求6所述的基于超声图像的肌肉厚度测量系统,其特征在于,所述系统还包括: 预处理模块,用于对所述感兴趣图像进行预处理,所述预处理包括对所述感兴趣图像进行灰度变换及调整图像对比度。 10. The muscle thickness measuring system based on the ultrasonic image according to claim 6, characterized in that the system further comprises: preprocessing means for preprocessing the image of interest, the pretreatment comprises the image of interest to adjust image contrast and gradation conversion.
CN 201210563056 2012-12-22 2012-12-22 Muscle thickness measuring method and system based on ultrasonic image CN103027713A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201210563056 CN103027713A (en) 2012-12-22 2012-12-22 Muscle thickness measuring method and system based on ultrasonic image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 201210563056 CN103027713A (en) 2012-12-22 2012-12-22 Muscle thickness measuring method and system based on ultrasonic image

Publications (1)

Publication Number Publication Date
CN103027713A true true CN103027713A (en) 2013-04-10

Family

ID=48015481

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201210563056 CN103027713A (en) 2012-12-22 2012-12-22 Muscle thickness measuring method and system based on ultrasonic image

Country Status (1)

Country Link
CN (1) CN103027713A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103679754A (en) * 2013-12-20 2014-03-26 河北汉光重工有限责任公司 Real-time target searching and tracking technique
CN106683115A (en) * 2016-12-21 2017-05-17 中国矿业大学 Video tracking method based on spiral vision-motion model

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5247938A (en) * 1990-01-11 1993-09-28 University Of Washington Method and apparatus for determining the motility of a region in the human body
CN1681439A (en) * 2002-09-12 2005-10-12 株式会社日立医药 Biological tissue motion trace method and image diagnosis device using the trace method
CN101145688A (en) * 2006-05-23 2008-03-19 沙诺夫公司;沙诺夫欧洲公司 Electrostatic discharge protection structures with reduced latch-up risks
CN101464948A (en) * 2009-01-14 2009-06-24 北京航空航天大学 Object identification method for affine constant moment based on key point

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5247938A (en) * 1990-01-11 1993-09-28 University Of Washington Method and apparatus for determining the motility of a region in the human body
CN1681439A (en) * 2002-09-12 2005-10-12 株式会社日立医药 Biological tissue motion trace method and image diagnosis device using the trace method
CN101145688A (en) * 2006-05-23 2008-03-19 沙诺夫公司;沙诺夫欧洲公司 Electrostatic discharge protection structures with reduced latch-up risks
CN101464948A (en) * 2009-01-14 2009-06-24 北京航空航天大学 Object identification method for affine constant moment based on key point

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
JIZHOU LI ETC.: "Real-Time Detection of Muscle Thickness Changes during Isometric Contraction from Ultrasonography:A Feasibility Study", 《COMPUTERIZED HEALTHCARE(ICCH),2012 INTERNATIONAL CONFERENCE ON》, 18 December 2012 (2012-12-18) *
李敏等: "下颌角弧形截骨术后咬肌厚度的变化", 《中华医学美学美容杂志》, vol. 13, no. 2, 30 April 2007 (2007-04-30) *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103679754A (en) * 2013-12-20 2014-03-26 河北汉光重工有限责任公司 Real-time target searching and tracking technique
CN106683115A (en) * 2016-12-21 2017-05-17 中国矿业大学 Video tracking method based on spiral vision-motion model

Similar Documents

Publication Publication Date Title
Zhang et al. VSI: A visual saliency-induced index for perceptual image quality assessment
Islam et al. Multifractal texture estimation for detection and segmentation of brain tumors
Yan et al. Automatic tracing of vocal-fold motion from high-speed digital images
Iakovidis et al. An intelligent system for automatic detection of gastrointestinal adenomas in video endoscopy
US20100284590A1 (en) Systems and Methods for Robust Learning Based Annotation of Medical Radiographs
US20070280530A1 (en) Using Candidates Correlation Information During Computer Aided Diagnosis
Sewell et al. Real-time eye gaze tracking with an unmodified commodity webcam employing a neural network
Wang et al. Retinal vessel segmentation using multiwavelet kernels and multiscale hierarchical decomposition
Li et al. Delving into egocentric actions
Fang et al. Video saliency incorporating spatiotemporal cues and uncertainty weighting
US20090180693A1 (en) Systems and methods for analyzing image data using adaptive neighborhooding
Bharadwaj et al. Biometric quality: a review of fingerprint, iris, and face
WO2011137409A1 (en) Malignant mass detection and classification in radiographic images
Huang et al. Contour tracking in echocardiographic sequences via sparse representation and dictionary learning
Zhu et al. Action recognition in broadcast tennis video using optical flow and support vector machine
US20110082371A1 (en) Medical image processing device and medical image processing method
Hegenbart et al. Scale invariant texture descriptors for classifying celiac disease
Flores-Tapia et al. Semi automatic MRI prostate segmentation based on wavelet multiscale products
Xu et al. Differential evolution with Powell's direction set method in medical image registration
Virmani et al. PCA-SVM based CAD system for focal liver lesions using B-mode ultrasound images
CN103116763A (en) Vivo-face detection method based on HSV (hue, saturation, value) color space statistical characteristics
Rahmatullah et al. Integration of local and global features for anatomical object detection in ultrasound
Wee et al. Nuchal translucency marker detection based on artificial neural network and measurement via bidirectional iteration forward propagation
Darby et al. Automated regional analysis of B-mode ultrasound images of skeletal muscle movement
Fang et al. Towards robust deconvolution of low-dose perfusion CT: Sparse perfusion deconvolution using online dictionary learning

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
C12 Rejection of a patent application after its publication