The Research of Multi-Modality Parkinson's Disease Image Based on Cross-Layer Convolutional Neural Network

摘要

Multi-modal diagnosis technology has become an increasingly important method in the medical field and has been widely used in imaging diagnosis and clinical treatment. This paper fused the T2-MRI and PET images of Parkinson patients. And the Multi-modality fusion images were input to a convolutional neural network for diagnosis of Parkinson's disease. This paper designed a cross-layer convolutional neural network for Parkinson's disease. The cross-layer convolutional neural network was used to learn and test the multi-modality fusion images, PET single-mode fusion images, T2-MRI single-mode fusion images respectively. Experiment result showed that the cross-layer convolutional neural network has higher diagnostic accuracy than the Lenet-5 convolutional neural network. And the diagnostic accuracy of multi-modality fusion images is a higher than that of single-mode fusion images. The diagnostic accuracy of single-mode fusion PET images is higher than that of single-mode fusion T2-MRI images. The result showed that multi-modality fusion images have better auxiliary diagnostic effects than the single modal fusion images. And the auxiliary diagnosis effect of PET single-modal fusion images is better than that of T2-MRI single-modal fusion images. The cross-layer convolutional neural network is much better than the Lenet-5 convolutional neural network. It is more suitable for the diagnosis of Parkinson's disease.