IMPROVED TUMOR DETECTION BY ULTRASONIC SPECKLE REDUCTION

超声斑点是临床上最主要的降噪源

Ultrasonic speckle is themost important source of noise degrading clinical ultrasound scan images. The detectability of tumors in ultrasonic scan images, particularly those of relatively low contrast in the surrounding normal tissues, may be greatly decreased due to speckle. The overall aim of the project proposal is the evaluation of the improvement in tumor detectability which may be achieved in diagnostic ultrasound through the application of techniques for the reduction of speckle in untrasonic images and definition of those forms of waveform processing which will lead to optimum improvement in image quality and tumor detectability. A versatile laboratory system for the acquisition of ultrasonic waveform data from phantoms and human tissues in vitro will be constructed. A variety of ultrasonic phantoms containing low-contrast, hyperechoic and hypechoic targets, similar to low-contrast tumors in the human body, will be scanned using this dedicated laboratory system. Fresh pathologic liver specimens involved with metastatic tumor, and some normal liver specimens will be scanned with the laboratory system as well. A variety of schemes for ultrasonic speckle reduction will be employed: the basic principle will be to use techniques which employ either spatial or frequency diversity techniques to produce differing (possibly uncorrelated) images which may then be averaged to create ultrasonic images with reduced speckle. Spatial diversity will be achieved by both linear translation of transducers perpendicular to scan planes and through transducer angulation techniques. Frequency diversity will be attained by both altering the number of cycles and center frequencies of ultrasound exciting the ultrasound transducer, and via digital filtering of scan data obtained employing broad-band transducers. The relationship between frequency parameters (center frequency, bandwidth) and transducer translation/rotation and speckle decorrelation will be determined experimentally and compared to prior theoretical predictions (of Gehlbach). The effects of the speckle reduction techniques will be evaluated via receiver operating characteristic (ROC) analysis of observers' ability to detect low-contrast targets and tumors in images created via a variety of processing schemes. The framework of knowledge created in these studies will indicate possible hardware and waveform processing improvements in clinical scanners, which will give improved and earlier detection of tumors in parenchymal organs of the body.