Title: The spectrum entropy estimation method for ultrasonic backscattering signal in cancel-lous bones

Author(s): XU Feng; XU Yiwen; LIU Chengcheng; TA De'an; GUO Jianzhong

Affiliation(s): Department of Electronic Engineering, Fudan University; et al.

Abstract: In the application of cancellous bone ultrasound diagnosis based on backscattering method, it is of great importance to estimate fast and accurately whether the valid backscattering signal exists in the received signal. We propose a fast estimation method based on spectrum entropy method. With 984 records of adult calcaneus clinical data, we estimate the validity of the backscatter signal using this method. The results of the proposed method and the results of experience-base judgement were compared and analyzed. And two key parameters, the signal range length and the segment number of the spectrum entropy, were analyzed. The results show when the signal range length is 13 μs and the segment number is 15 -- 20, this method can get the best result (accuracy > 95%, sensitivity > 99%, specificity > 87%), while taking little calculation time (1.5 ms). Therefore, this spectrum entropy method can satisfy the accuracy and real-time requirements in the ultrasonic estimation for cancellous bone.