How Reliable Is Individualized Functional-Network Attribution in Resting-State fMRI Cognitive Prediction in ADNI?

Sheethal Bhat; Mahfuzur Rahman Chowdhury; Imam Ahmed Shaik; Rose Dawn Bharath; Siming Bayer; Andreas Maier

cs.CVOct 27, 2026

How Reliable Is Individualized Functional-Network Attribution in Resting-State fMRI Cognitive Prediction in ADNI?

Authors: Sheethal Bhat, Mahfuzur Rahman Chowdhury, Imam Ahmed Shaik, Rose Dawn Bharath, Siming Bayer, Andreas Maier

Organizations: Pattern Recognition Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Martensstr. 3, 91058 Erlangen, Germany · Siemens Healthineers, Henkestr. 127, 91052 Erlangen, Germany

Abstract

Individualized attribution estimates which functional networks drive a model prediction for a given subject and is often proposed as an interpretable resting-state fMRI (rs-fMRI) biomarker, but its test-retest reliability is rarely quantified. We evaluate per-subject network-ablation attribution for a frozen resting-state functional connectivity (FC) predictor (Schaefer-200 FC predicting MMSE/ADAS-Cog13) trained on N=875 subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI), using, and (iii) a subject-level permutation null. Split-half reliability is near-ceiling (Spearman p=0.948/0.950) but drops on scan-rescan (p=0.609/0.616). Under the mismatched-subject null, correlations remain substantial (p~~0.46-0.47), indicating strong population-shared structure, yet matched pairs exceed this null (z~~6.6-6.9, p<=10-4). Thus, in ADNI, individualized network attribution contains a modest but non-zero person-specific component atop a strong population-shared pattern. We recommend reporting scan-rescan reliability together with a permutation-null control, rather than within-session proxies alone, to avoid overstating individual specificity.

Figures & tables

Figure 1: Top 6 of the 7 Yeo networks by group-level importance, visualized on anatomical MRI. Group-level importance is quantified by the drop in test-set explained variance when ablating each network, $\Delta R^{2}=R^{2}-R^{2}_{\setminus n}$ , computed on the held-out test set and averaged over three independently trained runs (shown separately for MMSE and ADAS-Cog13). $\Delta R^{2}$ can exceed the total $R^{2}$ when $R^{2}_{\setminus n}<0$ . The displayed “combined MMSE+ADAS-Cog13 rank” orders networks by the summed $\Delta R^{2}$ across the two targets. The 7th (lowest-ranked, Somatomotor) network is omitted for layout only; all 7 are used in every analysis.

Figure 2: Overview of the three progressively stricter reliability checks. (a) Within-session split-half: odd and even volumes of one session are each attributed and compared by Spearman

\rho

. (b) Scan–rescan: each subject’s baseline is compared with all repeat visits (median 24.4 months apart). (c) Permutation null: 10,000 subject-level derangements match each baseline with another subject’s repeat visits (kept as a block, no self-match); the observed mean

T_{\mathrm{obs}}

is compared with this null.

	Split-half		Scan–rescan
Statistic	MMSE	ADAS-Cog13	MMSE	ADAS-Cog13
Mean	0.948	0.950	0.609	0.616
95 % CI	0.929–0.964	0.934–0.964	0.553–0.663	0.571–0.660
Median	0.988	0.988	0.625	0.661
Std	0.101	0.085	0.242	0.198
10–90%	0.87–1.00	0.85–1.00	0.34–0.90	0.33–0.85

Table 1: Attribution reliability (Spearman

\rho

of the per-subject 7-network importance signature). The subject is the unit of analysis (split-half:

n{=}132

; scan–rescan:

n{=}74

, with pairs averaged within subject). CIs are subject-level bootstrap intervals; “run SD” is the SD across 3 runs; “–” indicates not computed. Permutation-null rows use 10,000 subject-level derangements.

Feature Attribution

Feature attribution estimates which input features or internal components contribute to a model’s predictions.

See all 234 papers

Interpretable Cross-Network Attention for Resting-State fMRI Representation Learning
Feb 28, 2026Karanpartap Singh, Adam Turnbull, Mohammad Abbasi +3Brain Network Analysis Functional Connectivity
Stanford University, Stanford, CA 94305
Benchmarking External Generalization of SPD Matrix Learning for Resting-State fMRI Connectome Prediction
Aug 31, 2026Ce Ju, Antoine Collas, Florent Bouchard +1Functional Connectivity Brain Age Prediction
Inria, CEA, Universit´e Paris-Saclay, Palaiseau, France · Universit´e Paris-Saclay, CNRS, CentraleSup´elec, L2S, France
An Attention-Based Framework for Alzheimers Disease Classification Using Resting-State fMRI
Jul 29, 2026Harshiddhi Pathak, Gowtham Reddy N, Mrinal Acharya +1Alzheimer's Disease Functional Connectivity
School of Medical Science and Technology · Department of Computer Science and Engineering at the Indian Institute of Technology Gandhinagar · Dr B C Roy Multi Speciality Medical Research Centre, Indian Institute of Technology Kharagpur, India

Feb 28, 2026J/K move · Enter open · S save

Interpretable Cross-Network Attention for Resting-State fMRI Representation Learning

Karanpartap Singh, Adam Turnbull, Mohammad Abbasi, Kilian Pohl+2

Stanford University, Stanford, CA 94305

Brain Network Analysis Functional Connectivity Self-Supervised Time Series Representation Learning Neural Network Interpretability Alzheimer's Disease Classification Cross-Attention

Open paper AskPDF arXiv

Understanding how large-scale functional brain networks reorganize during cognitive decline remains a central challenge in neuroimaging. While recent self-supervised models have shown promise for learning representations from resting-state fMRI, their internal mechanisms are difficult to interpret, limiting mechanistic insight. We propose BrainInterNet, a network-aware self-supervised framework based on masked reconstruction with cross-attention that explicitly models inter-network dependencies in rs-fMRI. By selectively masking predefined functional networks and reconstructing them from remaining context, our approach enables direct quantification of network predictability and interpretable analysis of cross-network interactions. We train BrainInterNet on multi-cohort fMRI data (from the ABCD, HCP Development, HCP Young Adults, and HCP Aging datasets) and evaluate on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, in total comprising 5,582 recordings. Our method reveals systematic alterations in the brain's network interactions under AD, including in the default mode, limbic, and attention networks. In parallel, the learned representations support accurate Alzheimer's-spectrum classification and yield a compact summary marker that tracks disease severity longitudinally. Together, these results demonstrate that network-guided masked modeling with cross-attention provides an interpretable and effective framework for characterizing functional reorganization in neurodegeneration.

Abstract

Feature Attribution

Explore similar work

Interpretable Cross-Network Attention for Resting-State fMRI Representation Learning

Benchmarking External Generalization of SPD Matrix Learning for Resting-State fMRI Connectome Prediction

An Attention-Based Framework for Alzheimers Disease Classification Using Resting-State fMRI