Novel method of Characterization of dispersive properties of heterogeneous head tissue using Microwave sensing and Machine learning Algorithms

Abstract

A brain tumor is a critical medical condition and early detection is essential for a speedy recovery. Researchers have explored the use of electromagnetic waves in the microwave region for the early detection of brain tumor. However, clinical adoption is not yet realized because of the low resolution of microwave images. This paper provides an innovative approach to improve microwave brain tumor detection intelligently by differentiating normal and malignant tissues using machine learning algorithms. The dataset required for classification is obtained from the antenna measurements. To facilitate the measurement process, an Antipodal Vivaldi antenna with the diamond-shaped parasitic patch (37 mmx21 mm) is designed to operate with a resonance frequency of 3 GHz. The proposed antenna maintains a numerical reflection coefficient (S11) value below -10dB over the entire UWB frequency range. In this paper, Waikato Environment for Knowledge Analysis (WEKA) classification tool with 10 cross-fold validation is used for comparison of various algorithms against the dataset obtained from the proposed antenna.