Apparatus and Method for Generating Medical Image Segmentation Deep-Learning Model, Medical Image Segmentation Deep-Learning Model Generated Therefrom

摘要

The present invention is an apparatus for generating a medical image fractional deep learning model, comprising: a training data generation/allocation unit for generating a training data set based on a fractional result obtained by inputting a given single medical image into an original medical image fractional deep learning model; After obtaining each temporary weight using the output data according to the primary learning by inputting the good task data and bad task data sampled from the training data sets for primary training into the medical image fractional deep learning model, training for secondary training The good task data and bad task data sampled from the data sets are input into the medical image fraction deep learning model, and the weights are updated by adding the obtained gradients using the respective weights obtained using the output data according to the secondary learning performance. Including a learning control unit, the primary learning and the secondary learning are repeated.