Analysis of backscattered ultrasound rf echoes from adjacent scan lines for surface roughness characterization: a phantom study

Ultrasonography, as a noninvasive and inexpensive routine, can be the modality of choice in roughness characterization of internal body stones for managing treatment methods. In this study, we have investigated the possibility of differentiating roughness levels utilizing backscattered RF data which presumably contain more information than the B-mode images. For this purpose, we modified a conventional medical ultrasound device and recorded RF data from a roughness phantom consisting of 4 standard sandpaper strips. We proposed that the difference of two echoes from adjacent scan lines contains information about the roughness of the imaging surface. Hence, we calculated the Euclidean distance of temporal and spectral features extracted from two adjacent echoes. Then, 3 classifiers of Bayesian, linear, and 1-Nearest Neighbor (NN) were employed for roughness differentiation. The results show that spectral features and 1-NN classifier had the best performance among others. The highest average performance of 99.17%, obtained using all features along with the 1-NN classifier, proves the feasibility of roughness discrimination by acquiring and comparing adjacent echoes.