DEEP LEARNING-BASED BRAIN TUMOR IMAGE DIVISION METHOD, DEVICE, APPARATUS, AND MEDIUM

摘要

A deep learning-based brain tumor image division method, a device, an apparatus, and a medium, being applied to the technical field of artificial intelligence and relating to the technical field of blockchains. Said method partially comprises: obtaining a multi-modal brain nuclear magnetic resonance image (S1); pre-processing the brain nuclear magnetic resonance image so as to obtain a target image with the skull part removed (S2); inputting the target image into a pre-set brain tumor division model so as to obtain a brain tumor image division result (S3). The pre-set giloma division model is a deep learning model obtained by performing cross-validated training according to an adaptive division framework and the brain nuclear magnetic resonance image with the skull part removed, said adaptive division framework including multiple types of U-Net models and U-Net integrated models. According to the result of cross validation, said method can automatically select from multiple models the optimal network structure for prediction, thereby enhancing the division performance of the pre-set brain tumor division models and enhancing the accuracy of brain tumor image division.