Brain stroke localization by using microwave-based signal classification

This research presents a new methodology of developing the brain stroke localization algorithm by using the machine-learning based pattern recognition approach for foreign object classification. Due to the new algorithm requires less bandwidth from the system hardware level, the proposed method would be helpful to reduce the challenges of designing a high-performance UWB antenna array required from most microwave stroke detection system nowadays. The presented algorithm employs the interconnection point of two lines that intersect the stroke region inside a brain and evaluates the properties of the stroke by using Support Vector Machines (SVM) classification method; the SVM classifier is trained by the data set acquired from EM simulations as well as the optimized coefficients from PSO algorithm. The proposed methodology of brain stroke classification has been verified by simulated data. The performance and effectiveness demonstrated it's a promising algorithm to be installed in the next-generation microwave brain stroke detection system.