摘要
目的:探讨超声Nakagami成像纹理特征参数检测微波热消融凝固区的可行性。方法:通过超声仪采集超声射频信号然后检测包络信号,构建超声Nakagami图像。基于超声Nakagami图像提取灰度共生矩阵纹理特征参数,输入支持向量机训练凝固区识别模型后结合多项式拟合得到凝固区最终识别结果。以离体猪肝最大剖面的测量结果为金标准,对比该方法与传统超声Nakagami成像识别凝固区的准确度。结果:该方法检测凝固区面积的准确度为(87.75±7.74)%,传统超声Nakagami成像检测凝固区面积的准确度为(84.38±13.52)%。Bland-Altman回归分析结果表明,该方法识别的凝固区面积与金标准具有很高的一致性。结论:基于Nakagami成像纹理特征可以实现对微波热消融凝固区的识别检测。
Objective To explore the feasibility of ultrasound Nakagami imaging texture feature parameters for detecting the microwave ablation coagulation zone.Methods Ultrasound Nakagami images were constructed by acquiring the ultrasound RF signals and detecting the envelope signals through an ultrasonograph.Based on the images constructed the grey-level co-occurrence matrix texture feature parameters were extracted and input to a support vector machine to train a coagulation zone identification model and then combined with polynomial fitting to obtain the final coagulation zone identification results.The accuracy of the results was compared with that by conventional ultrasound Nakagami imaging in identifying the coagulation zone using the measurements of the largest section of isolated pig liver as the gold standard.Results The accuracy for detecting the coagulation zone was(87.75±7.74)%,while that by conventional ultrasound Nakagami imaging was(84.38±13.52)%.The results of the Bland-Altman regression analysis showed that the area of the coagulation zone identified was in high agreement with the gold standard.Conclusion The texture feature of ultrasound Nakagami imaging can be used as for ultrasound tissue characterization when detecting the microwave ablation coagulation zone.
作者
王赫
周著黄
吴水才
WANG He;ZHOU Zhu-huang;WU Shui-cai(Faculty of Environment and Life,Beijing University of Technology,Beijing 100124,China)
出处
《医疗卫生装备》
CAS
2022年第2期1-6,12,共7页
Chinese Medical Equipment Journal
基金
国家自然科学基金面上项目(61871005)。
关键词
Nakagami成像
微波热消融
定量超声
灰度共生矩阵
支持向量机
纹理特征
Nakagami imaging
microwave ablation
quantitative ultrasound
gray-level co-occurrence matrix
support vector machine
texture feature
