The present invention relates to an MRA image learning method of a deep learning-based assistant diagnosis system and an assistant diagnosis method for a blood vessel lesion. According to the present invention, the MRA image learning method of a deep learning-based assistant diagnosis system for blood vessel lesion diagnosis comprises: an image input step in which an MRA image with a labeled blood vessel lesion is inputted into the assistant diagnosis system; a blood vessel extraction step in which a blood vessel area is extracted from the MRA image by the assistant diagnosis system; a cube extraction step in which a plurality of three-dimensional cube areas of a preset size are extracted along the blood vessel area by the assistant diagnosis system; a learning data extraction step in which the plurality of three-dimensional cube areas are classified into positive cubes and negative cubes based on the label of the MRA image by the assistant diagnosis system; and a model generation step in which a learning model is generated by learning the positive cubes and the negative cubes as learning data by the assistant diagnosis system. Accordingly, a blood vessel is learned and diagnosed in three-dimensional cube units to perform learning and diagnosis in a manner of viewing a blood vessel inside the blood vessel to simply and accurately diagnose a lesion such as a cerebral aneurysm.;COPYRIGHT KIPO 2020
展开▼