Explainable AI for Multimodal Biomarker Integration in Neurodegenerative Diseases: Bridging Neuroimaging and Biofluid Profiling

Neurodegenerative diseases vary widely, both clinically and biologically. This group includes Alzheimer’s disease, Parkinson’s disease, atypical parkinsonian syndromes, frontotemporal dementia, and amyotrophic lateral sclerosis. Pathology often begins years before symptoms appear, and the clinical features of these diseases often overlap. Structural, functional and molecular neuroimaging have advanced, as have ultrasensitive assays for cerebrospinal fluid and blood biomarkers, including amyloid-β, phosphorylated tau, neurofilament light chain, GFAP and α-synuclein seed amplification. Together, these tools provide complementary views of disease processes. Artificial intelligence can combine these large, varied data streams. However, many models cannot show how they reach their outputs, which limits clinical trust, regulatory acceptance and biological insight.This Research Topic aims to advance explainable AI (XAI) methods that combine neuroimaging and biofluid data. The goal is to support earlier and differential diagnosis, staging, prognosis and patient stratification in neurodegenerative diseases. We seek work that makes model reasoning transparent and clinically interpretable, and links the features a model learns to the underlying pathophysiology.We welcome Original Research, Reviews, Systematic Reviews, Methods and Perspective articles addressing, but not limited to:• Interpretable multimodal fusion architectures combining MRI, PET, SPECT or EEG with fluid biomarkers• Post hoc and inherently interpretable methods, such as SHAP, attention mapping, concept-based models and causal models, applied to biomarker integration• Data-driven subtyping and disease progression modelling• Approaches to handling missing modalities, site effects, harmonisation and data imbalance• External validation, generalisability, fairness and uncertainty quantification• Integration of genetic, digital and clinical data with imaging–biofluid profiles• Translational frameworks, clinician-in-the-loop evaluation and regulatory considerationsSubmissions should prioritise rigorous validation, reproducibility and open-science practices. By linking computational innovation with clinical neurology, this Topic aims to move multimodal AI beyond proof of concept towards trustworthy tools for precision medicine in neurodegeneration. Topic editors: Anastasia Bougea, Allison B Reiss, Alberto Ouro Source: https://www.frontiersin.org/research-topics/86370/explainable-ai-for-multimodal-biomarker-integration-in-neurodegenerative-diseases-bridging-neuroimaging-and-biofluid-profiling