A Fast Technique of Tissue Biomechanical Analysis for Real-time Prostate Tissue Elasticity Reconstruction
Authors: Seyed Reza Mousavi and Abbas Samani.
Abstract:
Elastography image reconstruction techniques typically involve displacement or stress field calculation of tissue undergoing mechanical stimulation that can be done by Finite Element (FE) analysis. However, traditional FE method is time-consuming, and hence not suitable for real-time or near real-time applications. In this article, we present an alternate accelerated method of stress calculation that can be incorporated in elastography reconstruction algorithms. Shape is an essential input of FE models that is considered in conjunction with material stiffness and loading to yield stress distribution. The essence of the proposed technique is finding a function between shape and stress field. This function takes the shape parameters as input and outputs the stress field very fast. To develop such a function principal component analysis (PCA) is used to obtain the main modes of shape and stress fields. As such, the shape and stress fields can be described by these main modes weighted by a small number of weight factors. Then, an efficient mapping technique is developed to relate the weight factors of shape to those of the stress fields. We used Neural Network (NN) for this mapping, which is the sought function required to input shape and output stress field. Once the mapping function is obtained it can be used for analyzing shapes not included in the NN training database. We employed this technique for prostate tissue stress analysis. For a typical prostate, our results indicate that analysis using our technique takes less than 0.07 seconds on a regular desktop computer irrespective of the model size and complexity. This analysis indicates that stress error of the majority of the samples is less than 5% per node.