Benchmarking Multimodal Deep Fusion Strategies for Heterogeneous Neuroimaging and Cognitive Data Using a Controlled Sex Classification Task.

Overview

1. Department of Medical and Surgical Sciences, Magna Graecia University, 88100 Catanzaro, Italy; (C.C.); (A.P.)
2. Brain Health Imaging Centre, Centre for Addiction and Mental Health, Toronto, ON M6J 1H4, Canada
3. Neuroscience Research Center, Magna Graecia University, 88100 Catanzaro, Italy
4. Institute of Neurology, Department of Medical and Surgical Sciences, Magna Graecia University, 88100 Catanzaro, Italy

Abstract

Highlights: What are the main findings? Early feature-level fusion consistently outperforms intermediate and late fusion strategies in multimodal sex classification. Standard feature scaling significantly enhances multimodal deep learning performance across architectures.

What are the implications of the main findings? Architectural complexity does not guarantee superior performance in heterogeneous multimodal integration. Fusion strategy and preprocessing must be jointly optimized for reliable and reproducible multimodal modeling in neuroscience.

Abstract: Background/Objectives: Multimodal data fusion is increasingly applied in neuroinformatics to integrate heterogeneous sources of information. However, the optimal strategies for combining modalities with markedly different dimensionality, scale, and noise characteristics remain unclear. To our knowledge, this is among the first systematic and controlled benchmarks explicitly disentangling the effects of fusion strategy and feature scaling within a unified deep learning framework. Methods: Using data from 747 healthy participants from the Human Connectome Project, we evaluated multiple fusion paradigms—including early fusion, attention-based fusion, subspace-based fusion, and graph-based fusion—within a unified and reproducible framework. Importantly, we assessed how different feature scaling techniques (Standard, Min–Max, and Robust scaling) interact with fusion strategies and influence model performance. Biological sex was used as a controlled benchmark task to focus on methodological insights rather than task-specific optimization. Results: Early feature-level fusion consistently achieved the highest classification performance across all evaluated configurations. In particular, direct concatenation of cognitive and neuroimaging features combined with Standard Scaling yielded the best results (AUC–ROC = 0.96 (0.95–0.96)), outperforming unimodal baselines as well as intermediate and late fusion strategies. Conclusions: This systematic benchmark demonstrates that multimodal deep learning performance in neuroscience is driven primarily by the interaction between fusion strategy and feature scaling rather than by architectural complexity alone. By explicitly disentangling the effects of fusion level and preprocessing within a unified framework, this study provides practical methodological guidance for the design, evaluation, and reproducible deployment of multimodal deep learning models in neuroscience.

Reproduced under the paper's license (CC BY), from the paper cited above.

Code

The paper says that its authors' code is available on request: it was not published with the paper, so there is nothing to verify.

The paper's code and data availability statement is in the Data section.

Tracing map

A tracing map links a paper to the code its authors published: this paper has none (its code is available on request), so it has no map.

Data

Datasets cited

• humanconnectome.org/study/hcp-young-adult, at Human Connectome Project; found in the text, “2.1. Data Preparation”

Data Availability Statement

The data analyzed in this study are publicly available from the Human Connectome Project (HCP) Young Adult dataset (WU–Minn Consortium) and can be accessed through ConnectomeDB (https://db.humanconnectome.org) upon registration and agreement to the data use terms. No new data were generated during the study. The code used to implement the multimodal fusion models is available from the corresponding author upon reasonable request.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 29 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 4 authors, 7 keywords, 2 funders, 42 references.

Cite

This paper

Camastra, C., Pelagi, A., Quattrone, A., & Sarica, A. (2026). Benchmarking Multimodal Deep Fusion Strategies for Heterogeneous Neuroimaging and Cognitive Data Using a Controlled Sex Classification Task. Brain sciences, 16(4), 405. https://doi.org/10.3390/brainsci16040405

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1038/s41467-026-73262-2 [code]

Robust but independent sex differences in human brain function, structure, and behavior.

Journal: Nature communications

In common: humanconnectome.org/study/hcp-young-adult, 2 references

[2] doi:10.1002/mds.70334

A Brainstem Radiomics Framework to Distinguish Progressive Supranuclear Palsy from Parkinson's Disease.

Journal: Movement disorders : official journal of the Movement Disorder Society

In common: author Andrea Quattrone

[3] doi:10.3390/ijms27093934 [code]

Serum Neurofilament Light Chain and GFAP Levels Are Associated with Structural Brain Connectivity in Parkinson's Disease.

Journal: International journal of molecular sciences

In common: author Andrea Quattrone

[4] doi:10.3389/fnins.2026.1817743

Estimation of head motion in structural MRI and its impact on cortical morphometry.

Journal: Frontiers in neuroscience

In common: humanconnectome.org/study/hcp-young-adult, methods / tools, 1 reference

[5] doi:10.1186/s13293-026-00891-z [code]

Sex differences in dynamic and static measures of brain integration derived from resting-state functional magnetic resonance imaging.

Journal: Biology of sex differences

In common: humanconnectome.org/study/hcp-young-adult, 1 reference

[6] doi:10.3389/fneur.2026.1796739 [code]

Network localization of regional intrinsic neural activity alterations in migraine and their neurochemical correlates.

Journal: Frontiers in neurology

In common: humanconnectome.org/study/hcp-young-adult, 1 reference

[7] doi:10.1126/sciadv.aef2894 [code]

Human cortical networks trade communication efficiency for computational reliability.

Journal: Science advances

In common: humanconnectome.org/study/hcp-young-adult, 1 reference

[8] doi:10.1002/sim.70712

A Bayesian Spatiotemporal Model for Joint Estimation of Brain Activation and Connectivity in fMRI Studies.

Journal: Statistics in medicine

In common: humanconnectome.org/study/hcp-young-adult, 1 reference

[9] doi:10.5812/ijem-168867

Large-Scale Brain Network Connectivity Mediates the Association Between Metabolic Risk Factors and Cognition: An fMRI Study of the Human Connectome Project.

Journal: International journal of endocrinology and metabolism

In common: humanconnectome.org/study/hcp-young-adult, 1 reference

[10] doi:10.1002/nbm.70277 [code]

Hierarchical Bayesian Modelling Improves Microstructural Parameter Mapping in Diffusion and Exchange MRI Data.

Journal: NMR in biomedicine

In common: humanconnectome.org/study/hcp-young-adult, 1 reference

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.