IEEE zyxwvutsrqponmlkj TRANSACTIONS ON SONICS AND ULTRASONICS, VOL. SU-31, zyxwvut NO. zyxwvu 4. JULY 1984 251 zy Error Reduction in Through Transmission Tomography Using Large Receiving Arrays with Phase-Insensitive Signal Processing RAINER M. SCHMITT, MEMBER, IEEE, CHARLES R. MEYER, MEMBER, IEEE, PAUL L. CARSON, THOMAS L. CHENEVERT, AND PEYTON H. BLAND, MEMBER, IEEE zyxwv Akrrucr-In order to reduce artifacts due to nonstraight line wave propagation in ultrasonic computed tomography of attenuation, the use of an experimental receiving array with phase-insensitive signal processing was evaluated. A 46-element 2.25 MHz array receiver having a singleelement aperture of 1.2 mm X 1.2 mm and a center-to-center spacing of 1.4 mm was employed and projections were taken of a phantom whose refraction characteristics were controlled by tempera- ture. Array signals were processed incoherently and attenuation pro- jections were calculated according to the energy ratio, frequency shift, and log spectral difference methods. The resulting projections then were compared to those obtained by a 19 mm diameter 35 MHz phase- sensitive single-element receiver. The phase-insensitive array response was superior to the phase-sensitive singleelement receiver, especially when the phantom became refractive. Modest improvements in the reduction of refractive errorswere obtained using the energy ratio method and negligible refractive errors were observed when the fre- quency shift or the log spectral difference method was used. I. INTRODUCTION T has been well documented that quantitative accuracy and resolution in attenuation imaging of the female breast using I ultrasound transmission tomography is limited by effects of refraction and diffraction when a limited-aperture single- element piezoelectric receiver is employed. curved refractive structure, i.e., a cyst or a spherical tumor, the beam is diffracted, refracted, and reflected. Thus in the receiving plane the redirected beam might partially missthe limited aperture receiver, which is aligned opposite the trans- mitter along the line of sight. If the receiver is enlarged in order to receive more completely a beam which is bent or otherwise disturbed by the object, then the detected ampli- tude is reduced due to degraded off-axis sensitivity caused by phase cancellation. As a result, attenuation of the many discontinuous structures in the breast is greatly overestimated. Different approaches have been and are being investigated to solve this problem: Klepper et al. [ 1 ] have suggested the use of large aperture phase-insensitive transducers with radiation When the interrogating ultrasound beam is incident upon a Manuscript received December 28, 1983; revised May zyxwvutsr 25, 1984. This work was supported in part by Grant PHS number R01 CA 31857 awarded by the NationalCancer Institute DHHS and in part by the Deutsche Forschungsgemeinschaft under Grant Schm 565/1-2. R. M. Schmitt, C. R. Meyer, P. L. Carson, T. L.Chenevert,and P. H. Bland are with the Department of Radiology, The University of Michigan Medical Center, Ann Arbor, MI 48109. patterns independent of small off-axis angles and independent of frequency. However, the low sensitivity of their acousto- electric receivers make them inappropriate for breast and most other medical imaging where resolution and sensitivity must be competitive with piezoelectric transducers. Two groups ([2] , [3] )simulated an array by moving a hydro- phone or a single-element receiver [4] across a large receiving aperture for each transmitter position in order to record more of the reflected or refracted beam energy. This indeed reduces artifacts caused by parts of the beam missing the receiver. However, when a medium aperture receiver (diameter 5- 10 mm) is used [4] , the artifacts due to beam bending or phase cancellation are still present. Scanning the receiving aperture using a hydrophone is ideal because it combines both large aperture and sufficient off-axis sensitivity. The scanned re- ceiver approach, however, is time consuming sincethe scan- ning time is multiplied by the number of samples needed to scan the receiving aperture. In this paper a different approach was experimentally in- vestigated. The received beam was sampled by an array and the signals from each array element were processed inco- herently. By doing this, the directivity of the array, under ideal circumstances, is proportional to the directivity of a single element. Thus, both sufficient aperture and off-axis sensitivity are providedby an array if the array is chosen large and the single-element width small enough. No increase in scanning time is needed when the signals are processed in parallel. In a real array, however, where many elements are placed closely together, new problems arise due toelectrical and acoustical cross coupling zyxw [5], [6] . This cross coupling, in addition to several vibrational modes within the array, de- grade off-axis sensitivity, which results in a difference between the directivity calculated from basic diffraction theory and the measured directivity of an element in an array. Therefore, an experimental array was employed as an intermediate step be- fore designing an array for implementation in our clinical scan- ning system. In order to evaluate the array approach we will show attenua- tion projections of a refractive and nonrefractive phantom ob- tained by alarge aperture piezoelectric ceramic array where three different phase insensitive digital signal processingmeth- ods were used: a) energy ratio, b) frequency shift, and c) log spectral differences. We will compare the projections with the ideal projection of the phantom and with those ob- 0018-9537/84/0700-0251$1.00 zyxwvu 0 1984 IEEE