基于斑块超声图像灰度分布的双峰Gamma模型评估斑块风险

Appraise Risk of Plaque Using a Bimodal Gamma Statistical Model Base on Gray-level Distribution of Carotid Plaque Ultrasound Images

目的:建立基于斑块超声图像灰度分布的双峰Gamma数学模型,识别不同回声特征的斑块,评估斑块的风险等级。方法:研究收集斑块样本137个,采用交叉验证法。首先,对斑块图像进行归一化处理,然后利用Potoshop软件获取斑块的灰度分布,接着利用Matlab内部的Lsqcurvefit非线性最小二乘法拟合函数,将双峰Gamma概率分布曲线拟合斑块原始灰度分布曲线,并建立双峰Gamma斑块灰度分布模型。最后通过曲线误差分析,测试斑块模型分类的准确率。结果:利用斑块模型分类,识别出高回声斑块、混合回声斑块和低回声斑块的准确率分别为100%、65%和75%。结论:双峰Gamma灰度模型有效描述斑块的灰度分布,识别高回声斑块有很高的准确率,对于评估斑块风险等级有很大潜能和良好的前景。

Objective The purpose of this study was to classify plaques between different echogenicity using abimodal Gamma statistical model base on gray-level distribution of carotid plaque ultrasound images. Methods Ultrasound images were obtained from a total of 137 carotid plaque and cross validation was implemented in this study. After images were normalized, gray level distribution of carotid plaque ultrasound images were obtained in Photoshop software. In Matlab, an internal fitting function base on nonlinear least square method,called lsqcurvefit, was used to get the curve of bimodal Gamma distribution base on gray-level distribution of carotid plaque ultrasound images. Lastly, plaques between different echogenicity were classified according to the error between gray level distribution curve of carotid plaque and the statistical model curves. Results The classification accuracy of hypoechoic, intermediate and hyperechoic plaques were 75%, 65% and 100% respectively. Conclusion The bimodal Gamma distribution was reasonable fit to the pixels of carotid plaque ultrasound images, and it had a high accuracy in identifying hyperechoic plaques.It is a promising tool for risk assessment of atherosclerosis.