You could take a look at this GitHub repo Brain-MRI-Age-Classification-using-Deep-Learning. The function prepare2DImageDataset.m might be relevant for you.
I think this example might be some help: 3-D Brain Tumor Segmentation Using Deep Learning, or this video tutorial.
For a 2-D semantic seg example (not involving brains), take a look at Semantic Segmentation Using Deep Learning. You'd need to get your data from the 3-D MRI scans into 2-D slices. You could look at writing a custom datastore to read 2-D slices from the 3-D data.
