Thermal Analysis of Malignant Brain Tumors by Employing a Morphological Differentiation-Based Method in Conjunction with Artificial Neural Network

TL;DR

This paper presents a novel method combining morphological analysis of temperature distribution and neural networks to classify brain tumor malignancy with high accuracy.

Contribution

It introduces a new approach using temperature data and neural networks to differentiate tumor malignancy based on shape models.

Findings

Successfully differentiates benign and malignant tumors

Accurately estimates degree of malignancy

Uses finite element analysis with neural network classification

Abstract

In this study, a morphological differentiation-based method has been introduced which employs temperature distribution on the tissue surface to detect brain tumor's malignancy. According to the common tumor CT scans, two different scenarios have been implemented to describe irregular shape of the malignant tumor. In the first scenario, tumor has been considered as a polygon base prism and in the second one, it has been considered as a star-shaped base prism. By increasing the number of sides of the polygon or wings of the star, degree of the malignancy has been increased. Constant heat generation has been considered for the tumor and finite element analysis has been conducted by the ABAQUS software linked with a PYTHON script on both tumor models to study temperature variations on the top tissue surface. This temperature distribution has been characterized by 10 parameters. In each scenario, 98 sets of these parameters has been used as inputs of a radial basis function neural network (RBFNN) and number of sides or wings has been selected to be the output. The RBFNN has been trained to identify malignancy of tumor based on its morphology. According to the RBFNN results, the proposed method has been capable of differentiating between benign and malignant tumors and estimating the degree of malignancy with high accuracy