Multichannel Crossed Convolutional Neural Network for Combined Estimation of Cortical Thickness and Bulk Velocities Using Ultrasonic Guided Waves: A Simulation Study

Ultrasonic guided waves (UGW) propagated in the long cortical bone can be measured via axial transmission method. However, ultrasonic identification of long cortical bone is a multi-parameter inverse problem and the optimal solution of the inverse problem often involves complex solving process. Deep neural network is essentially a multi-parameter powerful predictor based on universal approximation theorem. In the study, we investigate the feasibility of applying the multichannel crossed convolutional neural network (MCC-CNN) for simultaneous estimation of cortical thickness and bulk velocities. Finite-difference time-domain method (FDTD) is used to obtain the simulated UGW signals. The results illustrated that the proposed method confirms the feasibility and accuracy, which could be helpful to the UGW based evaluation of long cortical bone.