Alzheimer's disease pathology follows a characteristic pattern of spread through the human brain. Understanding this pattern has been critical to the development of imaging biomarkers and interpretation of in vivo imaging studies in Alzheimer's and related dementias. Yet this understanding is based on conventional 2D histology studies that only sample the brain sparsely and, more recently, on molecular imaging with PET, which has low resolution and limited specificity. Focusing on the medial temporal lobe, the region of the brain associated with memory function and the earliest site of Alzheimer's disease related neurodegeneration, we have developed pipelines for acquisition of serial histology, co-registration with high-resolution ex vivo MRI, and groupwise registration of ex vivo MRI for population-level analysis. These pipelines use 3D printing to simplify the underlying registration problems and combine deep learning with conventional image translation to facilitate multi-modality image registration. Using these pipelines, we are working to create 3D maps of neurodegenerative pathologies, to link pathology with macroscopic loss of brain gray matter and, ultimately, to improve biomarkers available to Alzheimer's disease clinical trials.