Microwave technique for brain stroke localization and classification using block sparse Bayesian learning with wavelet transform

A microwave imaging technique to classify brain strokes (intra cerebral hemorrhagic and ischemic) is presented. The technique uses a combination of wavelet transform, block sparse Bayesian learning and Born iterative methods to estimate the dielectric profiles (permittivity and conductivity) of the brain tissues. The two types of strokes can be classified based on the difference in their dielectric properties. The presented method is evaluated in a simulation environment that includes a realistic head phantom and 18 antenna elements operating across the band 0.9-1.9 GHz in the multistatic mode. The results indicate that the two types of stroke can be clearly classified under different scenarios of signal to noise ratio.