Performance evaluation of coherence-based adaptive imaging using clinical breast data.

Sound-velocity inhomogeneities degrade both the spatial resolution and the contrast in diagnostic ultrasound. We previously proposed an adaptive imaging approach based on the coherence of the data received in the channels of a transducer array, and we tested it on phantom data. In this study, the approach was tested on clinical breast data and compared with a correlation-based method that has been widely reported in the literature. The main limitations of the correlation-based method in ultrasonic breast imaging are the use of a near-field, phase-screen model and the integration errors due to the lack of a two-dimensional (2-D) array. In contrast, the proposed coherence-based method adaptively weights each image pixel based on the coherence of the receive-channel data. It does not make any assumption about the source of the focusing errors and has been shown to be effective using 1-D arrays. This study tested its in vivo performance using clinical breast data acquired by a programmable system with a 5 MHz, 128-channel linear array. Twenty-five cases (6 fibroadenomas, 10 carcinomas, 6 cysts, and 3 abscesses) were investigated. Relative to nonweighted imaging, the average improvements in the contrast ratio and contrast-to-noise ratio for the coherence-based method were 8.57 dB and 23.2%, respectively. The corresponding improvements when using the correlation-based method were only 0.42 dB and 3.35%. In an investigated milk-of-calcium case, the improvement in the contrast was 4.47 dB and the axial and lateral dimensions of the object were reduced from 0.39 to 0.32 mm and from 0.51 to 0.43 mm, respectively. These results demonstrate the efficacy of the coherence-based method for clinical ultrasonic breast imaging using 1-D arrays.