During awake brain surgery for tumour resection, in situ EEG recording (ECoG) is used to identify eloquent areas surrounding the tumour. We used the ECoG setup to record the electrical activity of cortical and subcortical tumours and then performed frequency and connectivity analyses in order to identify ECoG impairments and map tumours. We selected 16 patients with cortical (8) and subcortical (8) tumours undergoing awake brain surgery. For each patient, we computed the spectral content of tumoural and healthy areas in each frequency band. We computed connectivity of each electrode using connectivity markers (linear and non-linear correlations, phase-locking and coherence). We performed comparisons between healthy and tumour electrodes. The ECoG alterations were used to implement automated classification of the electrodes using clustering or neural network algorithms. ECoG alterations were used to image cortical tumours.Cortical tumours were found to profoundly alter all frequency contents (normalized and absolute power), with an increase in the δ activity and a decreases for the other bands (P < 0.05). Cortical tumour electrodes showed high level of connectivity compared to surrounding electrodes (all markers, P < 0.05). For subcortical tumours, a relative decrease in the γ1 band and in the alpha band in absolute amplitude (P < 0.05) were the only abnormalities. The neural network algorithm classification had a good performance: 93.6 % of the electrodes were classified adequately on a test subject. We found significant spectral and connectivity ECoG changes for cortical tumours, which allowed tumour recognition. Artificial neural algorithm pattern recognition seems promising for electrode classification in awake tumour surgery.