The goal of this project is to classify astrocyte morphologies involved in the pathology of Frontotemporal Lobar Degeneration to provide training sets for a deep learning algorithm. Due to the neuropathological and clinical heterogeneity of FTLD, it is difficult to develop reliable therapeutic and diagnostic methods. Therefore, the investigation of FTLD biomarkers is essential to reveal the etiology of the disease. In this study, we identified different astrocyte morphologies in GFAP-stained human brain tissue to provide training data for a supervised deep learning algorithm to allow a machine to automatically detect and quantify astrocytes. Deep learning provides a larger and more efficient dataset which pushes forward the development of clinical trials. Overall, this study provides high-throughput methods to elucidate FTLD neuropathology and reveals astrocyte morphologies implicated in neurodegeneration which can be utilized as biomarkers for FTLD diagnosis.