we proposed a novel classification method based on the fusion of multiview 2D and 3D convolutions for MRI-based AD diagnosis. Specifically, we first use multiple sub-networks to extract the local slice-level feature of each slice in different dimensions. Then a 3D convolution network was used to extract the global subject-level information of MRI. Finally, local and global information were fused to acquire more discriminative features. Experiments conducted on the ADNI-1 and ADNI-2 dataset demonstrated the superiority of this proposed model over other state-of-the-art methods for their ability to discriminate AD and Normal Controls (NC). Our model achieves 90.2% and 85.2% of accuracy on ADNI-2 and ADNI-1 respectively, thus it can be effective in AD diagnosis.
The dataset used in this study was obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) that is avaiable at http://adni.loni.usc.edu/ .
All MRIs were prepocessed by a standard pipeline in CAT12 toolbox which is avaiable at http://dbm,neuro.uni-jena.de/cat/.
Linux python 3.7 Pytorch version 1.2.0 NVIDIA GPU + CUDA CuDNN (CPU mode, untested) Cuda version 10.0.61
Please cite our paper if you use this code in your own work. H. Qiao, L. Chen and F. Zhu, "A Fusion of Multi-view 2D and 3D Convolution Neural Network based MRI for Alzheimer’s Disease Diagnosis," 2021 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), 2021, pp. 3317-3321, doi: 10.1109/EMBC46164.2021.9629923.