WO2016123815A1 - 基于pet/ct图像质量客观算法的评价方法及系统 - Google Patents

基于pet/ct图像质量客观算法的评价方法及系统 Download PDF

Info

Publication number
WO2016123815A1
WO2016123815A1 PCT/CN2015/072740 CN2015072740W WO2016123815A1 WO 2016123815 A1 WO2016123815 A1 WO 2016123815A1 CN 2015072740 W CN2015072740 W CN 2015072740W WO 2016123815 A1 WO2016123815 A1 WO 2016123815A1
Authority
WO
WIPO (PCT)
Prior art keywords
objective
algorithm
evaluation
image quality
objective algorithm
Prior art date
Application number
PCT/CN2015/072740
Other languages
English (en)
French (fr)
Inventor
殷夫
纪震
周家锐
李琰
张海婕
Original Assignee
深圳大学
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 深圳大学 filed Critical 深圳大学
Publication of WO2016123815A1 publication Critical patent/WO2016123815A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis

Definitions

  • the invention relates to the field of algorithm evaluation, in particular to an evaluation method and system based on an objective algorithm of PET/CT image quality.
  • PET/CT is a new medical device that combines PET and CT.
  • the combination of PET and CT not only preserves the role of classical anatomical images, but also incorporates advanced molecular imaging functions, the combination of which is far superior to the individual value.
  • the PET/CT examination is characterized by sensitivity, specificity, accuracy, and accurate positioning, and a single image can obtain tomographic images of various aspects of the human body, achieving the purpose of early detection of lesions and diagnosis of diseases.
  • IQA image quality assessment
  • Subjective evaluation method Through psychological test, the subjective evaluation score of image quality of Human Visual System (HVS) is directly obtained. This is a relatively reliable method of evaluation, usually in accordance with the procedures described in the ITU-R_BT.500-11 standard: in a closed, controlled environment, non-expert testers give subjectivity based on their perception of image quality. Evaluation score. This method can produce a more uniform subjective score with higher credibility.
  • HVS Human Visual System
  • the objective evaluation method is also often referred to as an image quality evaluation algorithm or an IQA algorithm.
  • the observation evaluation method uses the algorithm-aware model to calculate the objective evaluation score of image quality, and its purpose is to make the computer imitate the function of HVS.
  • the most basic requirement is to be able to express the differences of images substantially completely and maintain a high degree of consistency with subjective perception.
  • An ideal image quality evaluation algorithm should satisfy the following three aspects: 1) conforming to human visual perception; 2) versatility, stable evaluation performance for different objects and environments; 3) monotonic, consistent, and stable results .
  • the usual objective algorithm evaluation index does not fully reflect the performance of the objective algorithm.
  • the object of the present invention is to provide an evaluation method and system based on an objective algorithm of PET/CT image quality, aiming at solving the problem that the existing objective algorithm evaluation method cannot truly reflect the performance of the algorithm.
  • An evaluation method based on PET/CT medical image quality objective algorithm which comprises the steps of:
  • step B a five-parameter nonlinear logistic model with linear conditional constraints is used as a nonlinear model of objective algorithm evaluation values for nonlinear compensation :
  • ⁇ 1 , ⁇ 2 , ⁇ 3 , ⁇ 4 , ⁇ 5 are 5 partial regression coefficients
  • x represents the objective algorithm evaluation value
  • Quality (x) is the objective algorithm evaluation value after fitting.
  • the indicators of accuracy include a linear correlation coefficient CC, a mean square error root RMSE, and an average absolute error MAE, wherein:
  • X and Y represent image quality subjective measurement data and image quality objective prediction data, respectively, and N is the total number of samples.
  • the indicator of monotonicity includes a rank correlation coefficient SROCC, wherein:
  • X and Y represent image quality subjective measurement data and image quality objective prediction data, respectively, N is the number of samples, and D is the difference between levels.
  • the indicator of consistency includes an outlier rate OR, wherein:
  • N 0 is the number of outliers predicted by the objective algorithm, and N is the total number of samples.
  • the outlier can be determined to be an abnormal value, and discarded; otherwise it should be retained.
  • the evaluation method based on the objective algorithm of PET/CT medical image quality wherein the evaluation result of the objective algorithm is:
  • An evaluation system based on PET/CT medical image quality objective algorithm which includes:
  • An objective algorithm input module is configured to input an objective algorithm to be evaluated, select a test type in a preset subjective evaluation database, and then calculate a prediction result of the objective algorithm according to the test type;
  • a compensation module for nonlinearly compensating the prediction result to eliminate nonlinear factors introduced in the subjective evaluation database
  • the evaluation module is used to evaluate the prediction results of the objective algorithm according to three aspects of algorithm accuracy, monotonicity and consistency.
  • ⁇ 1 , ⁇ 2 , ⁇ 3 , ⁇ 4 , ⁇ 5 are 5 partial regression coefficients
  • x represents the objective algorithm evaluation value
  • Quality (x) is the objective algorithm evaluation value after fitting.
  • the evaluation system based on the objective algorithm of PET/CT medical image quality, wherein, in the evaluation module, the accuracy index includes a linear correlation coefficient CC, a mean square error root RMSE and an average absolute error MAE, and the monotonicity indicators include The rank correlation coefficient SROCC, the indicator of consistency includes the outlier rate OR.
  • the present invention evaluates the performance of objective algorithms from three aspects, including the accuracy, monotony, and consistency of the algorithm, and uses CC, MAE, and RMSE as indicators to evaluate the accuracy of the objective algorithm, and OR as the objective algorithm.
  • the index, SROCC is used as an index to evaluate the monotonicity of the objective algorithm, and the five evaluation indicators are combined to give the final evaluation result, which can comprehensively and objectively reflect the performance of the objective algorithm.
  • the evaluation method of the invention has the characteristics of simple application and strong real-time performance. Through the evaluation method of the present invention, the performance of the objective evaluation algorithm can be tested comprehensively, accurately, and quickly.
  • FIG. 1 is a flow chart of a preferred embodiment of an evaluation method based on an objective algorithm of PET/CT medical image quality according to the present invention.
  • FIG. 2 is a flow chart showing the process of establishing a subjective evaluation database in the method of the present invention.
  • FIG. 3 is an evaluation method of an objective algorithm based on PET/CT medical image quality according to the present invention.
  • the present invention provides an evaluation method and system based on an objective algorithm of PET/CT medical image quality.
  • an objective algorithm of PET/CT medical image quality In order to make the objects, technical solutions and effects of the present invention more clear and clear, the present invention will be further described in detail below. It is understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
  • FIG. 1 is a flowchart of a method for evaluating an objective image based on PET/CT medical image quality according to the present invention. As shown in the figure, the method includes the following steps:
  • S101 input an objective algorithm to be evaluated, select a test type in a preset subjective evaluation database, and then calculate a prediction result of the objective algorithm according to the test type;
  • step S101 as shown in FIG. 2, the process of establishing the subjective evaluation database includes:
  • test pictures collecting test pictures; the test pictures are divided into pictures containing lesions and no lesions, the ratio of the two is about 2:1.
  • the degradation methods selected by the present invention include JPEG compression, JPEG2000 compression, Contrast Change (CC), White Gaussian Noise (WGN), Gaussian Blur (GB), module missing or motion blur (Motion).
  • Blur, MB these types of degradation include actual PET/CT image coding and image transmission possible
  • Various distortions such as blockiness, structure-related noise, false contours, ringing effects, edge blur, particle noise, and flat-area random noise, make it easy to perform versatility testing of objective algorithms and in different distorted images. Functional testing.
  • the subjective test evaluation index is an integer score of 1 to 5, wherein the higher the score, the better the image quality;
  • N is the number of testers
  • U k is the subjective evaluation score of the test picture
  • the subjective evaluation data for which the standard deviation S k is greater than 2.5 is specified as invalid data; the rest is taken as the effective average score.
  • the test type is selected in the subjective evaluation database, including the universality test of all pictures, or the performance test for different degradation methods.
  • the objective algorithm prediction result is calculated after selecting the test type.
  • the video quality expert group VQEG Video Quality Experts Group Phase-I and Phase-II propose to perform nonlinear compensation on the prediction result of the objective algorithm to eliminate the nonlinear factor introduced in the subjective evaluation process.
  • the present invention uses a five-parameter nonlinear logistic model that increases linear conditional constraints as a nonlinear model of objective algorithm evaluation values.
  • the calculation formula is as follows, and the minimum mean square error method is used for fitting:
  • ⁇ 1 , ⁇ 2 , ⁇ 3 , ⁇ 4 , and ⁇ 5 are partial regression coefficients
  • x represents an objective algorithm evaluation value
  • Quality (x) is an objective algorithm evaluation value after fitting.
  • the performance of the objective algorithm is evaluated by the following three aspects: (a) algorithm accuracy: that is, the difference between the objective evaluation value and the subjective evaluation value should be sufficient (b) Monotonicity: The objective evaluation value should increase or decrease with the increase or decrease of the subjective evaluation value; (c) Consistency: The performance of the objective evaluation algorithm on the test set is similar to that on the training set.
  • the present invention tests algorithm performance according to the following five objective indicators:
  • the indicators of accuracy include Linear Correlation Coefficient (CC) and Root Mean Squared Error (RMSE). And Mean Absolute Error (MAE), where:
  • X and Y represent image quality subjective measurement data and image quality objective prediction data, respectively, and N is the total number of samples.
  • Monotonic indicators include the Spearman Rank Order Correlation Coefficient (SROCC), where:
  • X and Y represent image quality subjective measurement data and image quality objective prediction data, respectively, N is the number of samples, and D is the difference between levels, where is the absolute value of the difference between MOS and MOSp.
  • Indicators of consistency include the Outlier Ratio (OR), where:
  • N 0 is the number of outliers predicted by the objective algorithm, and N is the total number of samples.
  • the outlier can be determined to be an abnormal value, and discarded; otherwise it should be retained.
  • the evaluation results of the objective algorithm are:
  • CC, MAE and RMSE are used as objective algorithms to predict the accuracy of subjective quality
  • OR is used as an indicator to predict the stability of objective algorithms
  • SROCC is used as an indicator to predict the monotonicity of objective algorithms. The larger the values of CC and SROCC, the more Ok, the smaller the values of the other three, the better.
  • the comprehensive evaluation method is as follows:
  • the present invention further provides a preferred embodiment of an evaluation system based on an objective algorithm of PET/CT medical image quality, comprising:
  • An objective algorithm input module is configured to input an objective algorithm to be evaluated, select a test type in a preset subjective evaluation database, and then calculate a prediction result of the objective algorithm according to the test type;
  • a compensation module for nonlinearly compensating the prediction result to eliminate nonlinear factors introduced in the subjective evaluation database
  • the evaluation module is used to evaluate the prediction results of the objective algorithm according to three aspects of algorithm accuracy, monotonicity and consistency.
  • a five-parameter nonlinear logistic model with increased linear conditional constraints is used as a nonlinear model of objective algorithm evaluation values for nonlinear compensation:
  • ⁇ 1 , ⁇ 2 , ⁇ 3 , ⁇ 4 , ⁇ 5 are 5 partial regression coefficients
  • x represents the objective algorithm evaluation value
  • Quality(x) is the objective algorithm evaluation value after fitting.
  • the present invention evaluates the performance of the objective algorithm from three aspects, including the accuracy, monotony, and consistency of the algorithm, and uses CC, MAE, and RMSE as indicators to evaluate the accuracy of the objective algorithm, and OR as an objective algorithm for evaluation.
  • the stability index, SROCC is used as an index to evaluate the monotonicity of the objective algorithm, and the five evaluation indicators are combined to give the final evaluation result, which can comprehensively and objectively reflect the performance of the objective algorithm.
  • the evaluation method of the invention has the characteristics of simple application and strong real-time performance. Through the evaluation method of the present invention, the performance of the objective evaluation algorithm can be tested comprehensively, accurately, and quickly.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Apparatus For Radiation Diagnosis (AREA)

Abstract

本发明公开基于PET/CT图像质量客观算法的评价方法及系统，其中，方法包括步骤：输入需评价的客观算法，在预设的主观评价数据库中选择测试类型，然后根据测试类型计算客观算法的预测结果；对预测结果进行非线性补偿，以消除主观评价数据库中引入的非线性因素；按照算法准确性、单调性、一致性三个方面，对客观算法的预测结果进行评价。本发明的评价方法可全面客观反映客观算法的性能，其具有应用简单，实时性强的特点。通过本发明的评价方法，可全面、准确、迅速地测试客观评价算法的性能。

Description

基于PET/CT图像质量客观算法的评价方法及系统 技术领域
本发明涉及算法评价领域，尤其涉及基于PET/CT图像质量客观算法的评价方法及系统。
背景技术
PET/CT是将PET与CT融为一体而形成的新型医学设备。将PET和CT进行结合，不仅保留了经典解剖影像的作用，还加入了先进的分子影像功能，其结合远优于各自单独的价值。PET/CT检查具有灵敏、特异、准确、定位精确等特点，并且一次显像可获得人体各方位的断层图像，达到早期发现病灶和诊断疾病的目的。
医学图像，特别是放射医学成像，是临床医学研究、诊断、和治疗的重要依据。发展医学图像质量评价方法，对于评估和优化成像系统、医学图像的压缩方法、以及成像算法具有重大意义。目前有两类图像质量评价(IQA，Image Quality Assessment)方法：
1.主观评价法：通过心理学测试，直接获得人类视觉系统(Human Visual System,HVS)对图像质量的主观评价分数。这是一类比较可靠地评价方法，通常按照ITU-R_BT.500-11标准所描述的过程进行：在密闭的受控环境中，由非专家的测试者根据自己对图像质量的感受给出主观评价分数。这种方法可以产生较为统一的主观评分，可信度较高。
2.客观评价法，也常被称为图像质量评价算法或IQA算法。客 观评价方法采用算法感知模型计算图像质量客观评价分数，其目的是使计算机模仿HVS的功能。图像质量感知客观模型的研究中，最基本的要求是能够基本完备地表达图像的差异，并与主观感知保持高度一致性。在满足上述要求的基础上，消除模型中的冗余信息，使模型更加高效和简练是研究人员追求的目标。
一个理想的图像质量评价算法应满足以下三个方面：1)符合人类视觉感受；2)具有通用性，针对不同对象和环境，评价性能保持稳定；3)结果具有单调性，一致性，稳定性。而通常的客观算法评价指标并不能完整反映客观算法性能的优劣。
因此，现有技术还有待于改进和发展。
发明内容
鉴于上述现有技术的不足，本发明的目的在于提供基于PET/CT图像质量客观算法的评价方法及系统，旨在解决现有客观算法评价方法不能真实反映算法性能优劣的问题。
本发明的技术方案如下：
基于PET/CT医学图像质量客观算法的评价方法，其中，包括步骤：
A、输入需评价的客观算法，在预设的主观评价数据库中选择测试类型，然后根据测试类型计算客观算法的预测结果；
B、对预测结果进行非线性补偿，以消除主观评价数据库中引入的非线性因素；
C、按照算法准确性、单调性、一致性三个方面，对客观算法的 预测结果进行评价。
所述的基于PET/CT医学图像质量客观算法的评价方法，其中，所述步骤B中，使用增加线性条件限制的五参数非线性logistic模型作为客观算法评价值的非线性模型来进行非线性补偿：
Quality(x)＝β1logisict(β2，(x-β3))+β4x+β5；
其中，
Figure PCTCN2015072740-appb-000001
β1、β2、β3、β4、β5为5个偏回归系数，x代表客观算法评价值,Quality(x)为拟合后的客观算法评价值。
所述的基于PET/CT医学图像质量客观算法的评价方法，其中，所述步骤C中，准确性方面的指标包括线性相关系数CC、均方误差根RMSE及平均绝对误差MAE，其中：
Figure PCTCN2015072740-appb-000002
Figure PCTCN2015072740-appb-000003
X和Y分别代表图像质量主观测量数据和图像质量客观预测数据，N为样本的总数。
所述的基于PET/CT医学图像质量客观算法的评价方法，其中，所述步骤C中，单调性方面的指标包括秩相关系数SROCC，其中：
Figure PCTCN2015072740-appb-000004
X和Y分别代表图像质量主观测量数据和图像质量客观预测数据，N为样本的个数，D为等级间差 异。
所述的基于PET/CT医学图像质量客观算法的评价方法，其中，所述步骤C中，一致性方面的指标包括离群率OR，其中：
Figure PCTCN2015072740-appb-000005
N0为客观算法预测数据为离群值的个数，N为样本的总数。
所述的基于PET/CT医学图像质量客观算法的评价方法，其中，如果离群值Y与客观算法测定平均值之差的绝对值大于3倍的标准偏差(σ)，即：
Figure PCTCN2015072740-appb-000006
则可以判定该离群值为异常值，将其舍弃；否则应予保留。
所述的基于PET/CT医学图像质量客观算法的评价方法，其中，客观算法的评价结果为：
Figure PCTCN2015072740-appb-000007
基于PET/CT医学图像质量客观算法的评价系统，其中，包括：
客观算法输入模块，用于输入需评价的客观算法，在预设的主观评价数据库中选择测试类型，然后根据测试类型计算客观算法的预测结果；
补偿模块，用于对预测结果进行非线性补偿，以消除主观评价数据库中引入的非线性因素；
评价模块，用于按照算法准确性、单调性、一致性三个方面，对客观算法的预测结果进行评价。
所述的基于PET/CT医学图像质量客观算法的评价系统，其中，补偿模块中，使用增加线性条件限制的五参数非线性logistic模型作 为客观算法评价值的非线性模型来进行非线性补偿：
Quality(x)＝β1logisict(β2，(x-β3))+β4x+β5；
其中，
Figure PCTCN2015072740-appb-000008
β1、β2、β3、β4、β5为5个偏回归系数，x代表客观算法评价值,Quality(x)为拟合后的客观算法评价值。
所述的基于PET/CT医学图像质量客观算法的评价系统，其中，评价模块中，准确性方面的指标包括线性相关系数CC、均方误差根RMSE及平均绝对误差MAE，单调性方面的指标包括秩相关系数SROCC，一致性方面的指标包括离群率OR。
有益效果：本发明从三个方面评价客观算法的性能，包括算法的准确性、单调性、一致性，并以CC，MAE和RMSE作为评价客观算法准确性的指标，OR作为评价客观算法稳定性的指标，SROCC作为评价客观算法单调性的指标，综合五个评价指标给出最终评价结果，可全面客观反映客观算法的性能。本发明的评价方法具有应用简单，实时性强的特点。通过本发明的评价方法，可全面、准确、迅速地测试客观评价算法的性能。
附图说明
图1为本发明基于PET/CT医学图像质量客观算法的评价方法较佳实施例的流程图。
图2为本发明的方法中主观评价数据库的建立过程流程图。
图3为本发明基于PET/CT医学图像质量客观算法的评价方法另 一实施例的流程图。
具体实施方式
本发明提供基于PET/CT医学图像质量客观算法的评价方法及系统，为使本发明的目的、技术方案及效果更加清楚、明确，以下对本发明进一步详细说明。应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用于限定本发明。
请参阅图1，图1为本发明基于PET/CT医学图像质量客观算法的评价方法的流程图，如图所示，其包括步骤：
S101、输入需评价的客观算法，在预设的主观评价数据库中选择测试类型，然后根据测试类型计算客观算法的预测结果；
S102、对预测结果进行非线性补偿，以消除主观评价数据库中引入的非线性因素；
S103、按照算法准确性、单调性、一致性三个方面，对客观算法的预测结果进行评价。
在步骤S101中，如图2所示，主观评价数据库的建立过程包括：
S201、采集测试图片；测试图片分为含病灶图片与无病灶图片，二者比例大约为2:1。
S202、构造退化图片，并以此为基础建立主观评价数据库。本发明所选用的退化方法有JPEG压缩、JPEG2000压缩、对比度变化(Contrast Change，CC)、高斯白噪声(White Gaussian Noise，WGN)、高斯模糊(Gaussian Blur，GB)、模块缺失或运动模糊(Motion Blur，MB)；这些退化类型包含了实际PET/CT图像编码和图像传输中可能 导致的各种失真，如块化效应、结构关联噪声、伪轮廓、振铃效应、边缘模糊、颗粒噪声以及平坦区随机噪声等，所以方便进行客观算法的通用性测试以及在不同失真图像中的功能性测试。
S203、进行主观测试，国际电信联盟标准-电视图像质量的主观评价方法ITU-R_BT.500-11协议中要求至少有15名非图像处理专业测试者参加测试，测试方法选择双重刺激失真水平测试法(Double-Stimulus Impairment Scale Method，DSIS)。主观测试评价指标为1到5的整数评分，其中分数越高代表图像质量越佳；
S204、计算测试结果，在所有测试者完成评价后，计算每张图片获得的平均分(Mean-Opinion-Score，MOS)作为最终测试结果。其公式为：
Figure PCTCN2015072740-appb-000009
其中k为测试图片、N为测试者数量,Uk为测试图片主观评价分数；
S205、筛选测试结果，计算每个测试者主观评价数据的标准偏差，公式如下所示：
Figure PCTCN2015072740-appb-000010
对于标准偏差Sk大于2.5的主观评价数据规定为无效数据；其余作为有效平均分。
S206、保存测试结果，将每张图片获得的有效平均分作为最终测试结果存入主观评价数据库，即主观数据库中包含有一定数量的主观 评价测试图片，以及每张图片对应的MOS分数。
如图3所示，在输入测试客观算法之后，在主观评价数据库中选择测试类型，包括全部图片的通用性测试，或针对不同退化方法的性能测试。选择好测试类型后计算客观算法预测结果。
在所述步骤S102中，视频质量专家组VQEG(Video Quality Experts Group)Phase-I和Phase-II中提出需要对客观算法的预测结果进行非线性补偿来消除主观评价过程中引入的非线性因素。本发明使用增加线性条件限制的五参数非线性logistic模型作为客观算法评价值的非线性模型。其计算公式如下，并采用最小均方误差法进行拟合：
Quality(x)＝β1logisict(β2，(x-β3))+β4x+β5；
其中，
Figure PCTCN2015072740-appb-000011
β1、β2、β3、β4、β5为偏回归系数，x代表客观算法评价值,Quality(x)为拟合后的客观算法评价值。
在所述步骤S103中，对客观算法评价值进行预处理之后，通过以下三个方面来评价客观算法的性能：(a)算法准确性：即客观评价值与主观评价值之间的差异应足够小；(b)单调性：客观评价值应随主观评价值的增减而增减；(c)一致性：客观评价算法在测试集上表现出的性能与其在训练集上的性能相似。本发明按照以下五个客观指标来测试算法性能：
其中，准确性方面的指标包括线性相关系数(Linear Correlation Coefficient，CC)、均方误差根(Root Mean Squared Error，RMSE) 及平均绝对误差(Mean Absolute Error，MAE)，其中：
Figure PCTCN2015072740-appb-000012
Figure PCTCN2015072740-appb-000013
X和Y分别代表图像质量主观测量数据和图像质量客观预测数据，N为样本的总数。
单调性方面的指标包括秩相关系数(Spearman Rank Order Correlation Coefficient，SROCC)，其中：
Figure PCTCN2015072740-appb-000014
X和Y分别代表图像质量主观测量数据和图像质量客观预测数据，N为样本的个数，D为等级间差异，在此处为MOS和MOSp之间差值的绝对值。
一致性方面的指标包括离群率(Outlier Ratio，OR)，其中：
Figure PCTCN2015072740-appb-000015
N0为客观算法预测数据为离群值的个数，N为样本的总数。
如果离群值Y与客观算法测定平均值之差的绝对值大于3倍的标准偏差(σ)，即：
Figure PCTCN2015072740-appb-000016
则可以判定该离群值为异常值，将其舍弃；否则应予保留。
客观算法的评价结果为：
Figure PCTCN2015072740-appb-000017
以上五个评价指标中，CC，MAE和RMSE作为客观算法预测主观质量的准确性指标，OR作为预测客观算法稳定性的指标，SROCC作为预测客观算法单调性的指标，CC与SROCC的值越大越好，其余三者的值越小越好。
综合五个评价指标，本发明给出客观算法最终评价结果，其结果数值越大表示客观算法性能越好。综合评价方式如下所示：
Figure PCTCN2015072740-appb-000018
基于上述方法，本发明还提供基于PET/CT医学图像质量客观算法的评价系统较佳实施例，其包括：
客观算法输入模块，用于输入需评价的客观算法，在预设的主观评价数据库中选择测试类型，然后根据测试类型计算客观算法的预测结果；
补偿模块，用于对预测结果进行非线性补偿，以消除主观评价数据库中引入的非线性因素；
评价模块，用于按照算法准确性、单调性、一致性三个方面，对客观算法的预测结果进行评价。
补偿模块中，使用增加线性条件限制的五参数非线性logistic模型作为客观算法评价值的非线性模型来进行非线性补偿：
Quality(x)＝β1logisict(β2，(x-β3))+β4x+β5；
其中，
Figure PCTCN2015072740-appb-000019
β1、β2、β3、β4、β5为5个偏回归系数，x代表客观算法评价值,Quality(x)为拟合后的客观算 法评价值。
综上所述，本发明从三个方面评价客观算法的性能，包括算法的准确性、单调性、一致性，并以CC，MAE和RMSE作为评价客观算法准确性的指标，OR作为评价客观算法稳定性的指标，SROCC作为评价客观算法单调性的指标，综合五个评价指标给出最终评价结果，可全面客观反映客观算法的性能。本发明的评价方法具有应用简单，实时性强的特点。通过本发明的评价方法，可全面、准确、迅速地测试客观评价算法的性能。
应当理解的是，本发明的应用不限于上述的举例，对本领域普通技术人员来说，可以根据上述说明加以改进或变换，所有这些改进和变换都应属于本发明所附权利要求的保护范围。

Claims (10)

  1. 基于PET/CT医学图像质量客观算法的评价方法，其特征在于，包括步骤：
    A、输入需评价的客观算法，在预设的主观评价数据库中选择测试类型，然后根据测试类型计算客观算法的预测结果；
    B、对预测结果进行非线性补偿，以消除主观评价数据库中引入的非线性因素；
    C、按照算法准确性、单调性、一致性三个方面，对客观算法的预测结果进行评价。
  2. 根据权利要求1所述的基于PET/CT医学图像质量客观算法的评价方法，其特征在于，所述步骤B中，使用增加线性条件限制的五参数非线性logistic模型作为客观算法评价值的非线性模型来进行非线性补偿：
    Quality(x)＝β1logistic(β2，(x-β3))+β4x+β5；
    其中，
    Figure PCTCN2015072740-appb-100001
    β1、β2、β3、β4、β5为5个偏回归系数，x代表客观算法评价值,Quality(x)为拟合后的客观算法评价值。
  3. 根据权利要求2所述的基于PET/CT医学图像质量客观算法的评价方法，其特征在于，所述步骤C中，准确性方面的指标包括线性相关系数CC、均方误差根RMSE及平均绝对误差MAE，其中：
    Figure PCTCN2015072740-appb-100002
    Figure PCTCN2015072740-appb-100003
    X和Y分别代表图像质量主观测量数据和图像质量客观预测数据，N为样本的总数。
  4. 根据权利要求3所述的基于PET/CT医学图像质量客观算法的评价方法，其特征在于，所述步骤C中，单调性方面的指标包括秩相关系数SROCC，其中：
    Figure PCTCN2015072740-appb-100004
    X和Y分别代表图像质量主观测量数据和图像质量客观预测数据，N为样本的个数，D为等级间差异。
  5. 根据权利要求4所述的基于PET/CT医学图像质量客观算法的评价方法，其特征在于，所述步骤C中，一致性方面的指标包括离群率OR，其中：
    Figure PCTCN2015072740-appb-100005
    N0为客观算法预测数据为离群值的个数，N为样本的总数。
  6. 根据权利要求5所述的基于PET/CT医学图像质量客观算法的评价方法，其特征在于，如果离群值Y与客观算法测定平均值之差的绝对值大于3倍的标准偏差(σ)，即：
    Figure PCTCN2015072740-appb-100006
    则可以判定该离群值为异常值，将其舍弃；否则应予保留。
  7. 根据权利要求6所述的基于PET/CT医学图像质量客观算法 的评价方法，其特征在于，客观算法的评价结果为：
    Figure PCTCN2015072740-appb-100007
  8. 基于PET/CT医学图像质量客观算法的评价系统，其特征在于，包括：
    客观算法输入模块，用于输入需评价的客观算法，在预设的主观评价数据库中选择测试类型，然后根据测试类型计算客观算法的预测结果；
    补偿模块，用于对预测结果进行非线性补偿，以消除主观评价数据库中引入的非线性因素；
    评价模块，用于按照算法准确性、单调性、一致性三个方面，对客观算法的预测结果进行评价。
  9. 根据权利要求8所述的基于PET/CT医学图像质量客观算法的评价系统，其特征在于，补偿模块中，使用增加线性条件限制的五参数非线性logistic模型作为客观算法评价值的非线性模型来进行非线性补偿：
    Quality(x)＝β1logistic(β2，(x-β3))+β4x+β5；
    其中，
    Figure PCTCN2015072740-appb-100008
    β1、β2、β3、β4、β5为5个偏回归系数，x代表客观算法评价值,Quality(x)为拟合后的客观算法评价值。
  10. 根据权利要求8所述的基于PET/CT医学图像质量客观算法的评价系统，其特征在于，评价模块中，准确性方面的指标包括线性 相关系数CC、均方误差根RMSE及平均绝对误差MAE，单调性方面的指标包括秩相关系数SROCC，一致性方面的指标包括离群率OR。
PCT/CN2015/072740 2015-02-03 2015-02-11 基于pet/ct图像质量客观算法的评价方法及系统 WO2016123815A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201510059345.9 2015-02-03
CN201510059345.9A CN104657987B (zh) 2015-02-03 2015-02-03 基于pet/ct图像质量客观算法的评价方法及系统

Publications (1)

Publication Number Publication Date
WO2016123815A1 true WO2016123815A1 (zh) 2016-08-11

Family

ID=53249063

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2015/072740 WO2016123815A1 (zh) 2015-02-03 2015-02-11 基于pet/ct图像质量客观算法的评价方法及系统

Country Status (2)

Country Link
CN (1) CN104657987B (zh)
WO (1) WO2016123815A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111027589A (zh) * 2019-11-07 2020-04-17 成都傅立叶电子科技有限公司 一种多分制目标检测算法评价系统及方法
CN114820432A (zh) * 2022-03-08 2022-07-29 安徽慧软科技有限公司 一种基于pet和ct弹性配准技术的放射治疗效果评估方法
CN117437211A (zh) * 2023-11-20 2024-01-23 电子科技大学 一种基于双偏置校准学习的低成本图像质量评价方法

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108257132A (zh) * 2018-03-05 2018-07-06 南方医科大学 一种基于机器学习的ct图像质量评估的方法

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102547363A (zh) * 2010-12-07 2012-07-04 江南大学 基于轮廓波变换域图像特征的无参考图像质量评测方法
CN103281556A (zh) * 2013-05-13 2013-09-04 宁波大学 一种基于图像分解的立体图像质量客观评价方法
CN103325113A (zh) * 2013-06-06 2013-09-25 深圳大学 部分参考型图像质量评价方法及装置

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101482973B (zh) * 2009-01-21 2010-11-10 西安交通大学 基于早期视觉的部分参考图像质量评价方法

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102547363A (zh) * 2010-12-07 2012-07-04 江南大学 基于轮廓波变换域图像特征的无参考图像质量评测方法
CN103281556A (zh) * 2013-05-13 2013-09-04 宁波大学 一种基于图像分解的立体图像质量客观评价方法
CN103325113A (zh) * 2013-06-06 2013-09-25 深圳大学 部分参考型图像质量评价方法及装置

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111027589A (zh) * 2019-11-07 2020-04-17 成都傅立叶电子科技有限公司 一种多分制目标检测算法评价系统及方法
CN111027589B (zh) * 2019-11-07 2023-04-18 成都傅立叶电子科技有限公司 一种多分制目标检测算法评价系统及方法
CN114820432A (zh) * 2022-03-08 2022-07-29 安徽慧软科技有限公司 一种基于pet和ct弹性配准技术的放射治疗效果评估方法
CN114820432B (zh) * 2022-03-08 2023-04-11 安徽慧软科技有限公司 一种基于pet和ct弹性配准技术的放射治疗效果评估方法
CN117437211A (zh) * 2023-11-20 2024-01-23 电子科技大学 一种基于双偏置校准学习的低成本图像质量评价方法

Also Published As

Publication number Publication date
CN104657987B (zh) 2017-08-08
CN104657987A (zh) 2015-05-27

Similar Documents

Publication Publication Date Title
Pauchard et al. Quality control for bone quality parameters affected by subject motion in high-resolution peripheral quantitative computed tomography
KR100731358B1 (ko) 화질 평가 방법 및 시스템
Chow et al. Correlation between subjective and objective assessment of magnetic resonance (MR) images
Jain et al. A full-reference image quality metric for objective evaluation in spatial domain
US9418414B2 (en) Image measurement apparatus, image measurement method and image measurement system
TWI825240B (zh) 用於判定體液樣品中之分析物濃度的方法和系統及用於產生軟體實施模組的方法和系統
WO2016123815A1 (zh) 基于pet/ct图像质量客观算法的评价方法及系统
CN108257132A (zh) 一种基于机器学习的ct图像质量评估的方法
Christen et al. Least-detectable and age-related local in vivo bone remodelling assessed by time-lapse HR-pQCT
Elliott et al. Estimating body mass from skeletal material: new predictive equations and methodological insights from analyses of a known-mass sample of humans
CN104572538A (zh) 一种基于k-pls回归模型的中医舌图像颜色校正方法
CN102117329B (zh) 基于小波变换的胶囊内窥镜图像检索方法
Maggio et al. Skeletal age estimation in a contemporary Western Australian population using the Tanner–Whitehouse method
CN113749642B (zh) 量化皮肤潮红反应程度的方法、系统、介质及终端
Lahoulou et al. Full-reference image quality metrics performance evaluation over image quality databases
US11393087B2 (en) Objective analysis of medical image data for white matter abnormality
Morin et al. An observational study investigating inter-observer agreement for variation over time of body condition score in dairy cows
Islam et al. A novel image quality index for image quality assessment
JP2009176023A (ja) 健康診断の検査値換算プログラム
Guo et al. System identification approaches for energy intake estimation: Enhancing interventions for managing gestational weight gain
De Angelis et al. Image quality assessment: an overview and some metrological considerations
Potashnikov et al. Analysis of modern methods used to assess the quality of video sequences during signal streaming
US20240315594A1 (en) A method for x-ray image processing
Nguyen et al. A no-reference measure for uneven illumination assessment on laparoscopic images
Tripathi et al. A novel spatial domain image quality metric

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: 15880771

Country of ref document: EP

Kind code of ref document: A1

DPE1 Request for preliminary examination filed after expiration of 19th month from priority date (pct application filed from 20040101)
NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 15880771

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 15880771

Country of ref document: EP

Kind code of ref document: A1