Late progressions in the field of cerebrum imaging innovation have empowered scientists to acquire new viewpoints on mind life systems and capability. Clinical experts depend vig- porously on these imaging components for early identification and treatment. Profound brain organizations (DNNs) have generally been powerful for order and division assignments. In any case, there is a requirement for a highlight decrease to work on the proficiency of DNNs. To address this need, a profound wavelet autoencoder (DWA) has been proposed in this article, which consolidates the element decrease capacities of autoencoders with the picture disintegration capacities of wavelet changes. The subsequent packed highlight set can be used for future DNN- based order errands. The proposed DWA-DNN picture classifier has been tried on a mind picture dataset, and its presentation has been contrasted with that of other existing classifiers, like Autoencoder-DNN and DNN. The proposed approach has shown predominant execution when considered in contrast to the presentation standards of different classifiers