Boundary estimation of soft tissue tumor by using feed forward neural network with application of artificial tactile sensing - Boundary estimation of soft tissue tumor

Abstract

Geometrical feature assessment of a cancerous tumor embedded in biological soft tissue is a necessity in follow-up procedure and making suitable therapeutic decisions. Evidently by having such features in hand, tumor resections will be more curative and beneficial. In this paper a procedure of examining boundaries of a sphere-shaped tumor embedded in the liver tissue was investigated. At first, the main essential was to generate finite element model of the soft tissue including a tumor in ABAQUS. By considering viscoelastic properties, mechanical behavior of the tissue under a specified pattern of loading was studied. In the following, tumor boundary was estimated by using a feed forward neural network (FFNN). Genetic Algorithm (GA) was used for extracting input datasets of the network by extracting mechanical parameters from the tissue surface stress-strain diagrams. Data used for training the FFNN was result of implementing the ABAQUS-based model of the cancerous soft tissue which was tested 120 times with different tumor diameters. Throughout the process, 90 datasets were used for training and the other 30 were used for testing the network. The results affirmed that the produced intelligent procedure of estimating tumor boundaries can be relied on as a trustworthy method.