Development of computer aided detection techniques for brain disorder has been gaining significant importance in the past few years. Out of the various brain diseases, stroke stands first for the reason behind fatality and disability. Significant features extracted from brain MR images, along with machine learning techniques could identify discriminative patterns for automatic detection of ischemic stroke. This research aims at examining the wavelet based statistical features for characterizing such abnormal lesion structures. Five different wavelet functions, namely daubechies, symlet, coiflet, de-meyer and bi-orthogonal wavelets were extensively analyzed for different normal and abnormal datasets. The wavelet co-efficients were calculated for different levels and statistical parameters were extracted from it as features. These features were trained using support vector machines for automatic classification. Experiments indicate that the accuracy of the proposed system was around 98%.