基于自回归模型的超声生物测量分辨力增强方法  

Resolution enhancement method of ultrasonic biological measurement based on auto-regression model

作　　者：张荥娟 郑政[1] 

机构地区：[1]上海理工大学医疗器械与食品学院,上海200093

出　　处：《中国医学物理学杂志》2018年第1期89-95,共7页Chinese Journal of Medical Physics

摘　　要：超声生物测量通常采用包络检测确定渡越时间的方法来测量眼球结构各部分长度,但常规包络法的分辨力受超声频率和仪器的限制。为突破这些限制,达到超声分辨力增强的目的,将一种基于自回归模型的超分辨力算法应用于A型超声。算法利用域变换的思想,将用奇异值分解求得的模型阶数应用于自回归模型,并将其与功率谱估计相结合对超声信号进行处理,最终得到分辨力增强的超声信号。为验证其方法可行性,分别在25、5 d B模拟回波和30 d B层靶回波的情况下进行实验,比较包络法和此算法的分辨力。结果显示,在3种不同情况下,包络法的分辨力分别为0.120、0.135和0.135 mm,而基于自回归模型的超分辨力算法分辨力分别为0.060、0.090和0.075 mm,其较包络法分辨力分别提高1.0、0.5和0.8倍。再将其应用于眼球真实回波,比较可得此算法比包络法的分辨能力强,细节测量更精准,算法真实有效,因此具有A模式生物测量的应用潜力。In the field of ultrasonic biological measurement,the lengths of the eyeball structure of various parts were usually measured by determining the time of flight with envelop detection method,however,the resolution of the conventional envelop method is limited by the ultrasonic frequency and the instruments.Here,super-resolution algorithm based on autoregression（AR） model was applied to A-type ultrasound for overcoming the limitations and achieving the enhancement of ultrasonic resolution.Based on the domain transformation,the model order obtained from singular value decomposition was used in the AR model,which was combined with the power spectrum estimation to process the ultrasonic signals,thus obtaining the ultrasonic signals with enhanced resolution.We carried out experiments under the conditions of 25,5 d B simulated echo and 30 d B layer target echo,respectively,and compared the resolution of envelop method and the proposed method to verify the feasibility of the proposed method.The results showed that,under three different conditions,the resolutions of AR-based method were 0.060,0.090 and 0.075 mm,respectively,which were increased by 1.0,0.5 and 0.8 times as comparison with the resolutions of envelop method（0.120,0.135 and 0.135 mm,respectively）.The comparison in the measurement of eyeball actual echo confirmed that AR-based algorithm has a higher resolution and more accurate measurement than envelop method.The proposed AR-based method is effective and has potential for the application in Amodel biological measurement.

关 键 词：生物测量 包络 分辨力 奇异值分解 自回归 功率谱 

分 类 号：R318.04[医药卫生—生物医学工程]