We study two different methods to predict the age of patients based on the brain MRIs.
In the first case we assume there is a brain segmentation model which can differentiate grey matter (GM), white matter (WM), and cerebrospinal fluid (CSF). We first train the model (a simple U-net) and then use the predicted volumes for each tissue type as regressors for the age of the patient.
In the second case, we use the images themselves as input to a network which regresses the age. This model outperforms the 2-step case above which attests to the flexibility and power of neural networks as end-to-end models. The case above though shows that given a segmentation model we do not need to train a new model necessarily, but instead build on top of it.