Table 1.
MRI and FDG PET fusion classification accuracies (%).
Fig 1.
Pre-processing pipeline for a single subject.
A subject has N MRI scanning sessions and M PET scanning sessions; therefore, the pipeline yields N MRI images and M PET images. The pipeline is repeated for each subject in the dataset.
Fig 2.
Convolutional neural network for one modality.
A single MRI or PET volume is taken as input, and the output is a binary diagnosis label of either “Healthy” or “AD”.
Fig 3.
Convolutional neural network for fusing MRI and PET modalities.
An MRI and PET scan from a single patient is taken as input, and the output is again a binary diagnosis label.
Table 2.
MRI and amyloid PET fusion classification accuracies (%).
Table 3.
Classification subject age and gender breakdown.