AssemblyNet-AD-FTD is an artificial intelligence tool available on the volBrain platform that analyzes T1-weighted MRI to identify structural brain patterns associated with neurodegenerative diseases. The method builds on the AssemblyNet whole-brain segmentation framework, which provides detailed anatomical parcellation of brain structures. Using this segmentation, a deep learning strategy called deep grading evaluates how similar each brain structure is to patterns observed in healthy individuals, Alzheimer’s disease, or Frontotemporal dementia populations. The framework relies on an ensemble of 3D convolutional neural networks trained on large MRI datasets to learn disease-related anatomical signatures. These networks generate a 3D grading map highlighting brain regions whose structural patterns are closer to specific disease profiles. Deep grading features are then combined with volumetric biomarkers derived from brain segmentation to improve robustness and interpretability. The pipeline finally produces quantitative outputs summarizing disease-related structural patterns across the brain. By combining deep learning with detailed brain parcellation, AssemblyNet-AD-FTD enables automated and reproducible analysis of neurodegenerative patterns from structural MRI.