A closer look at contrast source inversion for breast cancer detection

Abstract

Ultrasound is an emerging technology for breast cancer detection. It is an efficient and harmless method that can detect tumors in dense breasts, which may be missed using mammography. Currently, several fully automated ultrasound screening modalities are being developed. For some of those systems, accurate knowledge about the transducer location is available, making the measured data suitable for imaging using non-linear inversion methods. A promising, but costly, inversion method is contrast source inversion, which has been tested successfully on synthetic measurement data. To reduce the computational costs (computing time and memory load) various set-ups were tested. Results obtained with synthetic data show that inversion using a single frequency component only, still yielded excellent imaging results, while significantly reducing memory requirements. In addition, our data indicate that the number of source positions is less important than the number of receiver positions. Thus, while keeping the total number of A-scans identical, the inversion improved when the number of source positions was reduced and the number of receiver positions was increased. This approach efficiently reduced the computational costs associated with the inversion.