Deep learning algorithms, in particular long short-term memory (LSTM), have become an increasingly popular choice for natural language processing for a variety of applications such as sentiment analysis and text analysis. In this study, we propose a fully automated deep learning algorithm which learns to classify radiological reports for the presence of intracranial hemorrhage (ICH) diagnosis. The proposed automated deep learning architecture consists of 1D convolution neural networks (CNN), long short-term memory (LSTM) units and a logistic function which was trained and tested on the large dataset of 12,852 head computed tomography (CT) radiological reports. The architecture extracts semantically co-located features using 1D CNNs, the sequential structure of features using LSTM, and finally detects ICH using a logistic function. The receiver operator characteristic (ROC) curve is generated as a metric to test the classification performance of the architecture. The model achieved an area under the curve (AUC) of the ROC curve of 0.94. The promising results suggest that modern deep learning based algorithms are capable of extracting diagnosis information from unstructured medical text. The purpose of this paper is to label 27,148 radiological reports automatically to reduce human error, cost, and time.