Abstract
Longitudinal multimodal data characterize Alzheimer’s disease (AD) from complementary neuroanatomical, molecular, and genetic perspectives, but joint analysis is limited by sparse irregular visits, entangled common and disease-related changes, and modality-specific temporal availability. We propose Biologically Guided Multi-Axis Delta Learning (BioMAD), which assigns asymmetric roles to structural magnetic resonance imaging (sMRI), plasma biomarkers, and single-nucleotide polymorphisms (SNPs). BioMAD decomposes inter-visit sMRI deltas into a weakly supervised reference component and learned residual axes; plasma changes dynamically modulate residual deltas, whereas SNP embeddings provide static priors for axis assignment and latent regional modeling. Masked time-aware attention aggregates valid visit pairs across irregular intervals while retaining sMRI as the primary phenotype. Across four Alzheimer’s Disease Neuroimaging Initiative (ADNI) tasks, BioMAD achieved the highest area under the receiver operating characteristic curve (AUC) and accuracy among the compared methods. Removing multi-axis delta disentanglement (MAD), time-aware aggregation (TAA), or ROI interaction reduced accuracy; progressive mild cognitive impairment (pMCI) versus stable MCI (sMCI) was evaluated retrospectively, not as prospective conversion prediction.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/ELAMI_020.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: Not Available
BibTex
@InProceedings{ZhuYan_BioMAD_MICCAISAT2026,
author = { Zhu, Yanrui AND Li, Jiaqiang AND Cheng, Nina AND Wang, Tianfu AND Lei, Baiying AND Lei, Haijun AND Yang, Peng},
title = { { BioMAD: Biologically Guided Multi-Axis Delta Learning for Longitudinal Alzheimer’s Disease Classification } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17262},
month = {pending},
page = {pending}
}