Spindles are rhythmic transients present in the sleep electroencephalogram (EEG). Automatic spindle detection techniques are actively sought in order to make sleep staging easier as well as to enable the study of the microstructure of sleep. In the present work an Artificial Neural Network (ANN) based on the Multi-Layer Perceptron (MLP) architecture is used for detecting spindles and quantifying their temporal characteristics. The EEG is band-pass filtered and fed to the ANN without feature extraction. The visual inspection of the original EEG was performed by two experienced polysomnographers, involving three levels of EEG segment characterization. This enabled a thorough assessment of the intra- and inter-scorer variability, to which the performance of the ANN was compared. The overall sensitivity of the network, ranged from 81.3% to 85.7% and false positives rate from 10.3 to 6.3%.