Purpose: To create a computational model on the mesoscale of a cortical sheet (5 × 5mm2) to study the mechanisms of microseizure restriction, the recruitment of normal tissue into seizure activity and possibilities for therapeutic intervention. Methods: We develop a model of interacting minicolumns where the dynamics of each minicolumn contains a switch between a low-activity background and high-activity seizure state. The connectivity between minicolumns is based on results from anatomical studies of the neocortex. Results: The model reproduces the clinically observed, spatially restricted microseizures on the cortical mesoscale as transient self-terminating localised events. To model focal seizures, small regions of abnormal, hyperexcitable tissue were included, which spontaneously recruited larger regions of tissue into macroscopic seizure activity. The model predicts that severing cortico-cortical connections between scattered clusters of hyperactive microdomains could be suficient to prevent seizure onset. Significance: Our model provides a framework to test the organization and interactions within epileptic foci on the cortical mesoscale. In addition it allows the development and testing of suitable surgical treatment protocols in silico.