In this paper we present a study to automatically iterate through a database containing hundreds of MRI image slices and classify them as containing a tumor or as healthy. Performing this classification manually is prone to error, costly, time consuming, irreproducible and non-scalable. Our technique uses the midsagittal axis to determine whether an asymmetry is present amongst the two hemispheres. This is done using a deformation process followed by a novel morphological erosion operation to conclude the presence of a tumor. Our dataset contains simulated T1-weighted axial slices derived from 3D volumes. These volumes were obtained from BrainWeb and augmented with tumors by the Utah Center for Neuro Image Analysis. Our results indicate an accuracy rate of roughly 90% for the 277 slices tested. These promising results demonstrate that our technique can be applied to large datasets to detect the presence of tumors. Furthermore, our process can be extended to localize these tumors and apply data mining procedures for quantitative analysis.