Classification of breast masses by ultrasonic Nakagami imaging

Ultrasound is an important clinical tool in noninvasive diagnoses of breast cancer. The Nakagami statistical parameter estimated from the ultrasonic backscattered envelope has been demonstrated to be useful in complementing conventional B-mode scans when classifying breast masses. However, the shadowing effect caused by certain high-attenuation tumors in the B-mode image makes the tumor contour unclear, and thus it is more difficult to choose an appropriate region of interest from which to collect tumor data for estimating the Nakagami parameter. This study explored the feasibility of using the Nakagami parametric image to overcome the shadowing effect for visualizing the properties of breast masses. Experiments were performed on a breast-mimicking phantom and on some typical clinical cases for cysts, fat and tumors (fibroadenoma) (n = 18) in order to explore the performance of the Nakagami image under ideal and practical conditions. The experimental results showed that the Nakagami image pixels (i.e. the local Nakagami parameter) in the cyst, tumor and fat are 0.21 ± 0.01, 0.65 ± 0.05 and 0.98 ± 0.07, respectively, for six independent phantom measurements, and 0.14 ± 0.03, 0.67 ± 0.11 and 0.89 ± 0.08, respectively, for clinical experiments. This suggests that the Nakagami image is able to classify various breast masses (p < 0.005) although the clinical results from tumors of different cases have a larger variance that may be caused by the complexity of real breast tissues. In particular, unlike the B-mode image, the Nakagami image is not subject to significant shadowing effects, making it useful to complement the B-mode image to describe the tumor contour for identifying the tumor-related region when the shadowing effect is stronger or a low system gain is used.