Abstract: A method for segmenting a thrombus and a lesion caused by the thrombus in a three-dimensional brain image, includes supervised training of a primary recurrent artificial neural network to provide a lesion prediction from an image; supervised training of a secondary recurrent artificial neural network to provide a thrombus prediction from an image; using each trained primary recurrent artificial neural network on each image of a set of images obtained from the three-dimensional image, and merging the lesion predictions obtained to obtain a set of lesion segmentations: if the set of lesion segmentations includes a segmentation, selecting the widest segmentation as the lesion segmentation; using each trained secondary recurrent artificial neural network on each image of a set of secondary images obtained from the three-dimensional image, and merging the thrombus predictions obtained to obtain a set of thrombus segmentations.