Abstract: Neuronal intranuclear inclusion disease (NIID) is a rare neurodegenerative disease. Its diagnosis is mainly performed through diffusion weighted imaging (DWI). However, it is easy to miss or misdiagnose due to the doctor's vision and experience. This paper proposes a deep learning method based on cross self-supervision, and constructs CO-ResNet50 and CO-ViT models for intelligent auxiliary diagnosis of NIID. This method uses self-supervised learning and effectively combines the characteristics of ResNet50 and ViT networks to improve the model feature extraction ability. The experiment preprocessed 249 DWI data and divided them into 204 training sets and 45 test sets. The test results show that the CO-ResNet50 model performs best, with an accuracy of 95.49%, a precision of 95.51%, a recall of 95.44%, an F1 score of 0.954 7, and an AUC of 0.989 7, which can provide support for doctors in clinical diagnosis of NIID.

Key words: MRI, NIID, self-supervised learning, intelligent diagnosis, deep learning