Saved in:
Bibliographic Details
Main Authors: Shah, Uzair, Agus, Marco, Gamal, Mahmoud, Alzubaidi, Mahmood, Cali, Corrado, Magistretti, Pierre J., Bouzerdoum, Abdesselam, Househ, Mowafa
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.20970
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917356493602816
author Shah, Uzair
Agus, Marco
Gamal, Mahmoud
Alzubaidi, Mahmood
Cali, Corrado
Magistretti, Pierre J.
Bouzerdoum, Abdesselam
Househ, Mowafa
author_facet Shah, Uzair
Agus, Marco
Gamal, Mahmoud
Alzubaidi, Mahmood
Cali, Corrado
Magistretti, Pierre J.
Bouzerdoum, Abdesselam
Househ, Mowafa
contents Neuronal morphology encodes critical information about circuit function, development, and disease, yet current methods analyze topology or graph structure in isolation. We introduce GraPHFormer, a multimodal architecture that unifies these complementary views through CLIP-style contrastive learning. Our vision branch processes a novel three-channel persistence image encoding unweighted, persistence-weighted, and radius-weighted topological densities via DINOv2-ViT-S. In parallel, a TreeLSTM encoder captures geometric and radial attributes from skeleton graphs. Both project to a shared embedding space trained with symmetric InfoNCE loss, augmented by persistence-space transformations that preserve topological semantics. Evaluated on six benchmarks (BIL-6, ACT-4, JML-4, N7, M1-Cell, M1-REG) spanning self-supervised and supervised settings, GraPHFormer achieves state-of-the-art performance on five benchmarks, significantly outperforming topology-only, graph-only, and morphometrics baselines. We demonstrate practical utility by discriminating glial morphologies across cortical regions and species, and detecting signatures of developmental and degenerative processes. Code: https://github.com/Uzshah/GraPHFormer
format Preprint
id arxiv_https___arxiv_org_abs_2603_20970
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle GraPHFormer: A Multimodal Graph Persistent Homology Transformer for the Analysis of Neuroscience Morphologies
Shah, Uzair
Agus, Marco
Gamal, Mahmoud
Alzubaidi, Mahmood
Cali, Corrado
Magistretti, Pierre J.
Bouzerdoum, Abdesselam
Househ, Mowafa
Computer Vision and Pattern Recognition
Neuronal morphology encodes critical information about circuit function, development, and disease, yet current methods analyze topology or graph structure in isolation. We introduce GraPHFormer, a multimodal architecture that unifies these complementary views through CLIP-style contrastive learning. Our vision branch processes a novel three-channel persistence image encoding unweighted, persistence-weighted, and radius-weighted topological densities via DINOv2-ViT-S. In parallel, a TreeLSTM encoder captures geometric and radial attributes from skeleton graphs. Both project to a shared embedding space trained with symmetric InfoNCE loss, augmented by persistence-space transformations that preserve topological semantics. Evaluated on six benchmarks (BIL-6, ACT-4, JML-4, N7, M1-Cell, M1-REG) spanning self-supervised and supervised settings, GraPHFormer achieves state-of-the-art performance on five benchmarks, significantly outperforming topology-only, graph-only, and morphometrics baselines. We demonstrate practical utility by discriminating glial morphologies across cortical regions and species, and detecting signatures of developmental and degenerative processes. Code: https://github.com/Uzshah/GraPHFormer
title GraPHFormer: A Multimodal Graph Persistent Homology Transformer for the Analysis of Neuroscience Morphologies
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.20970