Speckle Reduction Achievable By Spatial Compounding And Frequency Compounding: Experimental Results And Implications For Target Detectability

Abstract

Speckle Reduction Achievable by Spatial Compounding and Frequency Compounding:Experimental Results and Implications for Target DetectabilityG.E. Trahey,1 J.W. Allison,' S.W. Smith2,3 and O.T. von Ramml'Department of Biomedical Engineering, Duke University, Durham, NC 277062Center for Devices and Radiological Health, Food and Drug Administration, Rockville,MD 208573Department of Radiology, Duke University Medical Center, Durham, NC 27710AbstractThe degree of speckle reduction achievable by spatial and frequency compounding is afunction of the rates of speckle pattern change induced by varying the imaging system'sillumination angle and acoustic frequency, respectively. We have measured these rates undera variety of conditions and derived a method of maximizing speckle reduction using theaverage of partially correlated speckle patterns. Our experimental results agree well withtheoretical predictions of these phenomena and indicate that, under limited conditions,improved target detection is possible using spatial compounding. Frequency compoundingappears to be counterproductive in improving target detectability.IntroductionSpeckle is a coherent interference phenomenon which degrades the inherent targetdetectability of coherent imaging systems. Speckle appears as a random mottle superimposedon an image and, in coherent optical systems, reduces the perceived resolution by a factorof five to seven' and degrades the minimum detectable contrast level2. Although asimilar resolution reduction has not yet been measured in coherent ultrasonic imagingsystems, the underlying physics are similar and it is likely that speckle hinders thedetection of several clinically important targets in medical ultrasound.Acoustical speckle arises from the interference of echoes from targets within a resolu-tion volume of the imaging system. The first- and second -order statistics of speckle can bepredicted by calculating the coherent echo sum of many such targets for Rayleighstatistics3,4. These predictions have been experimentally verified by several groups5.Speckle may be reduced by forming an image which is the incoherent average of images withdiffering speckle patterns after envelope detection. Such images may be acquired by varyingthe angle from which a target is imagedb7 (a technique known as spatial compounding) orby changing the spectrum of the acoustical pulse (frequency compounding). Other methods ofspeckle reduction utilize spatial filtering10 or nonlinear adaptive filteringll.The design of an imaging system employing frequency compounding and its success inspeckle reduction depend on the magnitude of speckle pattern change caused by variations inthe acoustical pulse center frequency. Previous attempts to measure this relationshiphave yielded inconsistent results8,9,12,13.We have measured speckle correlation in ultrasonic images when the illumination angle andthe center frequencies of transmitted pulses of several lengths are varied. The results ofthese measurements are compared to several theoretical predictions and used to quantify therelationship between lateral resolution loss and speckle reduction when spatial compoundingis employed and between axial resolution loss and speckle reduction when frequencycompounding is employed.Experimental TechniquesThe studies described were conducted using the latest configuration of the DukeUniversity Phased Array Scanner14. This machine has 32 channels in transmit and receive,with software -controllable operation via a Digital Equipment Corporation PDP -11/40 computer.The illumination angle was varied by linear translations of the transducer array withrespect to a fixed target position. The acoustic frequency was varied by creating transmitpulses via 32 coherent tone -burst generators with selectable burst lengths and centerfrequencies. The scanner was operated in the linear mode with no signal reject selected.In order to reduce measurement errors due to the sector line format and the imagedigitization process, the ultrasonic images were created at a high line density (6 lines perdegree) and digitized at a high spatial frequency (14 pixels /mm). The transducers utilizedwere 32- element, wideband arrays constructed in our laboratory.