SYSTÈME D'ARCHITECTURE D'APPRENTISSAGE PROFOND POUR LECTURE D'IMAGE DE FOND D'ŒIL AUTOMATIQUE ET PROCÉDÉ DE LECTURE D'IMAGE DE FOND D'ŒIL AUTOMATIQUE FAISANT APPEL À UN SYSTÈME D'ARCHITECTURE D'APPRENTISSAGE PROFOND
The present invention relates to an algorithm for automatic fundus image reading, and to a deep learning architecture for automatic fundus image reading, capable of minimizing the amount of data required for learning by training and reading artificial intelligence in a manner similar to that of an ophthalmologist acquiring medical knowledge. The deep learning architecture system for automatic fundus image reading, according to the present invention, comprises: a trunk module (100) in which common parts are combined into one in a plurality of convolutional neural network (CNN) architectures having at least one serially arranged set of feature extraction layers composed of a plurality of convolution layers that perform fundus image feature extraction and one pooling layer that performs subsampling for reducing computation amount； a plurality of branch modules (200), which generates each architecture in the trunk module (100) so as to receive an output of the trunk module (100), thereby identifying a lesion in the fundus image and diagnosing a corresponding disease name therefor； a section (110), which is an architecture in which any one branch module (200) from among the plurality of branch modules (200) is connected to the trunk module (100)； a root layer (120) for connecting the trunk module (100) and the branch module (200) by transmitting the output of a specific layer of the trunk module (100) to the branch module (200)； and a final decision unit (300), which integrates diagnosis data from the plurality of branch modules (200) so as to determine and output a final disease name.La présente invention concerne un algorithme de lecture d'image de fond d'œil automatique, et une architecture d'apprentissage profond pour lecture d'image de fond d'œil automatique, capable de réduire la quantité de données requises pour l'apprentissage par entraînement et l'intelligence artificielle de lecture d'une manière similaire à celle d'un ophtalmologue acquérant des connaissances
