Estimation of Linear and Nonlinear Elastic Parameters of Prostate Tumors Using Artificial Neural Networks : Estimation of Linear and Nonlinear Elastic Parameters of Tumors

Due to momentous clinical applications, modeling soft tissues and studying their mechanical properties such as elasticity and hyperelasticity were highly highlighted during the last decade. Because of differences between the mechanical properties of normal and cancerous tissues, precise modeling of mechanical behavior of soft tissues and distinguishing the types of tissues based on their responses to applied stimulations would facilitate the diagnosis of cancerous tissues. The present study sought to noninvasively recognize the mechanical behavior of prostate tissue and its cancerous masses. In this regard, the mechanical parameters of cancerous tissues were accurately estimated using the potent neural network method based on the displacement data. The displacement data related to various tissues and corresponding mechanical properties are required for developing and training neural network models. The finite element modeling using Abaqus software was implemented to simulate prostate tissue behavior and extract the required data for training neural networks. The nonlinear tissue behavior should be considered in soft tissue modeling. For representing the hyperelastic behavior of soft tissues, Ogden and Yeoh models are accurate, which were utilized in the study to prepare the finite element model of prostate tissue containing tumor. In addition, white noise was added into the displacement data obtained by the finite element model for simulating laboratory conditions during extracting tissue data from the model in order to achieve robust neural network models. The results indicate high accuracy and efficiency of the trained neural network models in estimating the mechanical parameters of cancerous prostate tissues based on the displacement data, which is promising outcome for the exact diagnosis of cancerous tissues.