The input data in this paper consisted of sMRI image data from the ADNI dataset (https://adni.loni.usc.edu/), including 98 scans from 30 AD subjects, 100 scans from 24 EMCI subjects, 105 scans from 25 LMCI subjects, and 107 scans from 26 NC subjects, for a total of 433 scans from 105 subjects, with roughly equal numbers of AD, EMCI, LMCI, and NC data. The basic information of the subject data showed in Table 4.It is important to note that the data we downloaded from the ADNI public dataset was in DICOM format, and we needed to convert it to NII format using SPM12.
This dataset contains AD, EMCI, LMCI, and NC data related to AD aided diagnosis. The folder names correspond to the labels in the classification task. In actual deep learning experiments, it is easy to write code to divide this dataset into training, validation, and testing sets.