The reviewed record of science sign in
Pith

arxiv: 2606.17867 · v1 · pith:W4OMFMFZ · submitted 2026-06-16 · cs.CV · cs.AI

A Quantitative Analysis of Multimodal Biomarkers in Alzheimer's Disease

Antonio Scardace , Daniele Rav\`i This is my paper

Reviewed by Pith2026-06-27 01:45 UTCgrok-4.3pith:W4OMFMFZopen to challenge →

classification cs.CV cs.AI
keywords Alzheimer's diseasemultimodal biomarkerstau imagingbrain structuremutual informationcognitive assessmentneurodegenerative trajectorygenetic markers
0
0 comments X

The pith

Quantitative analysis of multimodal data quantifies redundancies and decomposes tau-cognition links in Alzheimer's disease.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper integrates tau imaging, brain structure scans, thinking test scores, and genetic markers from hundreds of subjects to measure how these different data types relate. It calculates shared information and explained variance to spot overlaps, examines which brain areas show the strongest tau and shrinkage connections, splits the link between tau and cognition into parts tied to or separate from shrinkage, and extracts one main disease progression path that matches worsening test scores. A reader would care because the results could help choose which measures add unique value, drop redundant ones to ease patient load, and clarify distinct ways brain changes lead to symptoms. The work treats the modalities as an interconnected system rather than isolated signals.

Core claim

By applying mutual information measures, variance calculations, and statistical decomposition to combined tau-PET, structural MRI, cognitive scores, and genetic data, the study reveals substantial cross-modal overlaps, identifies brain regions where tau topologies align with atrophy, partitions the tau-cognition association into atrophy-related and atrophy-independent components, and extracts a dominant neurodegenerative trajectory aligned with cognitive decline.

What carries the argument

Cross-modal mutual information quantification together with statistical decomposition of the tau-cognition association into atrophy-related and atrophy-independent components.

If this is right

  • Modalities sharing high mutual information can be deprioritized to reduce redundant testing.
  • Brain regions with the strongest tau-atrophy associations become the focus for targeted region selection.
  • The atrophy-independent tau component isolates pathways to cognitive change not explained by structure alone.
  • The dominant trajectory offers a simplified model for tracking how neurodegeneration aligns with symptom progression.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same mutual information and decomposition approach could be applied to other diseases measured with multiple data types to check for similar redundancies.
  • Panels of biomarkers could be designed around the unique information each modality adds once overlaps are quantified.
  • The atrophy-independent component might suggest separate research directions into non-structural contributors to cognitive loss.

Load-bearing premise

The mutual information values and statistical decompositions capture genuine biological relationships rather than dataset-specific artifacts or measurement noise.

What would settle it

Repeating the mutual information calculations and decompositions on an independent cohort of subjects and obtaining substantially different overlap values or component sizes.

Figures

Figures reproduced from arXiv: 2606.17867 by Antonio Scardace, Daniele Rav\`i.

Figure 1
Figure 1. Figure 1: Overview of the multimodal preprocessing pipeline that operates through parallel branches to extract structural volumes [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: Cross-Modal Directional Explained Variance. Cells in the matrix represent the percentage of variance (R2 ) in a target modality (columns) explained by a predictor modality (rows) using cross-validated Ridge regression. This asymmetry demon￾strates a clear directional relationship: structural and molecular changes are strong predictors of clinical outcomes, whereas clinical assessments reflect aggregate fun… view at source ↗
Figure 5
Figure 5. Figure 5: Cross-Regional Tau–Atrophy Associations. Cells in the matrix quantify how strongly tau accumulation in a specific region (rows) is coupled with structural volume loss in another region (columns) on average across the entire cohort. Only interaction weights with 95% CI (derived from 7,500 bootstrap iterations) are retained. was significant (0.284, 95% CI: [0.197, 0.384]), suggesting that structural neurodeg… view at source ↗
Figure 6
Figure 6. Figure 6: Temporal Cascade of Biomarker Abnormalities. The vertical axis lists the multimodal biomarkers, while the horizontal axis represents the SuStAIn stages. Colors indicate the transition to mild (blue), moderate (gray), and severe (yellow) abnormality thresholds. Block opacity reflects the MCMC sampling probability of that specific event occurring at a given stage: solid, vibrant colors denote high certainty … view at source ↗
read the original abstract

Despite increasing adoption of multimodal approaches in Alzheimer's Disease (AD) research -- aimed at integrating molecular, structural, clinical, and genetic biomarkers to enhance disease characterization -- the relationships among these modalities remain poorly understood. A systematic analysis of their dynamic interaction is essential for improving disease modeling, identifying redundant assessments, and reducing patient burden and acquisition costs. In this paper, we present a quantitative analysis of multimodal AD biomarkers by integrating tau-PET, structural MRI, cognitive scores (MMSE and CDR), and APOE4 data from 789 subjects drawn from the ADNI dataset. In our analyses, we (A) quantify cross-modal mutual information and explained variance to assess redundancy and predictive dependencies; (B) examine associations between tau topologies and structural atrophy across brain regions to select informative ROIs; (C) perform a statistical decomposition of the tau-cognition association into atrophy-related and atrophy-independent components; (D) and identify a dominant neurodegenerative trajectory that aligns with cognitive decline. This study provides a systematic characterization of cross-modal relationships, improving the interpretability and selection of biomarkers in AD. Code is publicly available at: https://github.com/antonioscardace/Multimodal-AD.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript presents a quantitative analysis of multimodal biomarkers in Alzheimer's Disease using tau-PET, structural MRI, cognitive scores (MMSE and CDR), and APOE4 data from 789 ADNI subjects. It quantifies cross-modal mutual information and explained variance to assess redundancy and predictive dependencies, examines associations between tau topologies and structural atrophy to select informative ROIs, performs a statistical decomposition of the tau-cognition association into atrophy-related and atrophy-independent components, and identifies a dominant neurodegenerative trajectory aligned with cognitive decline. Code is made publicly available.

Significance. If the reported cross-modal MI values, explained-variance decompositions, and tau-atrophy-cognition trajectory accurately reflect stable biological dependencies, the study offers a systematic characterization that could improve biomarker interpretability, reduce redundant assessments, and guide more efficient multimodal protocols in AD research. The public code is a clear strength for reproducibility.

major comments (2)
  1. [Methods and Results] The central claims (abstract) that the analyses improve interpretability and selection of biomarkers in AD rest on results from the single ADNI cohort (n=789) with no external replication cohort, no explicit modeling of scanner/site effects, and no sensitivity checks for known ADNI selection criteria (APOE4 enrichment, MCI focus). This is load-bearing because the reported MI values, variance decompositions, and dominant trajectory could be driven by cohort-specific artifacts rather than general biological relationships.
  2. [Results (parts C and D)] The decomposition of the tau-cognition association into atrophy-related and atrophy-independent components (analysis C) and the identification of the dominant trajectory (analysis D) lack reported checks for unmodeled confounders (e.g., age, education, or head motion) or post-hoc ROI selection effects; if these alter the components materially, the interpretability conclusions do not follow.
minor comments (2)
  1. [Abstract] The abstract states the sample size and analyses (A-D) but does not preview the magnitude of key findings (e.g., peak MI values or explained-variance percentages); adding one or two quantitative highlights would improve clarity.
  2. [Figures] Figure legends and axis labels for mutual-information heatmaps and variance-decomposition plots should explicitly state units and whether values are normalized.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below and note planned revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Methods and Results] The central claims (abstract) that the analyses improve interpretability and selection of biomarkers in AD rest on results from the single ADNI cohort (n=789) with no external replication cohort, no explicit modeling of scanner/site effects, and no sensitivity checks for known ADNI selection criteria (APOE4 enrichment, MCI focus). This is load-bearing because the reported MI values, variance decompositions, and dominant trajectory could be driven by cohort-specific artifacts rather than general biological relationships.

    Authors: We agree that single-cohort analyses limit generalizability and that ADNI's multi-site nature, APOE4 enrichment, and MCI focus warrant explicit discussion. We will add a dedicated limitations section addressing these issues and potential cohort artifacts. ADNI data undergo standard harmonization, but we will include sensitivity analyses stratified by site or excluding enriched subgroups where feasible. An external replication cohort was outside the original study scope, so we will frame the results as ADNI-specific and recommend future validation studies. revision: partial

  2. Referee: [Results (parts C and D)] The decomposition of the tau-cognition association into atrophy-related and atrophy-independent components (analysis C) and the identification of the dominant trajectory (analysis D) lack reported checks for unmodeled confounders (e.g., age, education, or head motion) or post-hoc ROI selection effects; if these alter the components materially, the interpretability conclusions do not follow.

    Authors: Analysis C regressions already incorporated age and education as covariates; we will explicitly report this and the associated coefficients in the revision. Head-motion metrics are not uniformly available across the ADNI tau-PET subset, which we will note as a limitation. For post-hoc ROI effects from analysis B, we will add sensitivity checks using the full ROI set and alternative selection thresholds, reporting whether the atrophy-related and independent components or the dominant trajectory change materially. These additions will appear in the revised Methods and Results. revision: yes

Circularity Check

0 steps flagged

No circularity in statistical analysis chain

full rationale

The paper conducts direct empirical computations—mutual information, explained variance decompositions, ROI associations, and trajectory identification—on the ADNI cohort data. No equations, fitted parameters, or claims are shown to reduce by construction to their own inputs, self-citations, or ansatzes; all reported quantities are computed outputs from the dataset rather than renamed inputs or self-referential definitions. The derivation chain is therefore self-contained as standard statistical processing without circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No free parameters, axioms, or invented entities are described in the abstract; the work relies on standard statistical tools applied to an existing public dataset.

pith-pipeline@v0.9.1-grok · 5733 in / 995 out tokens · 30450 ms · 2026-06-27T01:45:54.278208+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

28 extracted references

  1. [1]

    Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the global burden of disease study 2019,

    E. Nichols, J. Steinmetz, S. V ollset, K. Fukutaki, J. Chalek, F. Abd-Allah, A. Abdoli, A. Abualhasan, E. Abu-Gharbieh, T. Akramet al., “Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the global burden of disease study 2019,”The Lancet Public Health, vol. 7, 2022

    2019

  2. [2]

    The clinical use of structural mri in alzheimer’s disease,

    G. B. Frisoni, N. C. Fox, C. R. Jack, P. Scheltens, and P. M. Thompson, “The clinical use of structural mri in alzheimer’s disease,”Nature Reviews Neurology, vol. 6, no. 2, pp. 67–77, 2010

    2010

  3. [3]

    18f-av-1451 and csf t-tau and p-tau as biomarkers in alzheimer’s disease,

    N. Mattssonet al., “18f-av-1451 and csf t-tau and p-tau as biomarkers in alzheimer’s disease,”EMBO molecular medicine, vol. 11, no. 2, p. e9105, 2019

    2019

  4. [4]

    Apolipoprotein e and alzheimer’s disease: risk, mechanisms and therapy,

    C.-C. Liu, T. Kanekiyo, H. Xu, and G. Bu, “Apolipoprotein e and alzheimer’s disease: risk, mechanisms and therapy,”Nature Reviews Neurology, vol. 9, no. 2, pp. 106–118, 2013

    2013

  5. [5]

    Alzheimer’s disease neuroimaging initiative (adni),

    R. C. Petersen, P. S. Aisen, L. A. Beckett, M. C. Donohue, A. C. Gamst, D. J. Harvey, C. R. Jack, W. J. Jagust, L. M. Shaw, A. W. Toga, J. Q. Trojanowski, and M. W. Weiner, “Alzheimer’s disease neuroimaging initiative (adni),”Neurology, vol. 74, no. 3, pp. 201–209, 2010

    2010

  6. [6]

    Synthesizing images of tau pathology from cross-modal neuroimaging using deep learning,

    J. Lee, B. J. Burkett, H.-K. Min, M. L. Senjem, E. Dicks, N. Corriveau- Lecavalier, C. T. Mester, H. J. Wiste, E. S. Lundtet al., “Synthesizing images of tau pathology from cross-modal neuroimaging using deep learning,”Brain, vol. 147, no. 3, pp. 980–995, 2023

    2023

  7. [7]

    Anatomy- Guided Multimodal Graph Networks for Alzheimer’s Disease: Integrative Analysis of Cross-Modal Brain Connectivity Signatures ,

    W. Hu, Z. Guan, P. Yang, J. Li, Y . Liu, S. Gan, T. Cai, A. Zhang, T. Zhang, J. Qu, S. Wang, G. Cai, X. Dong, T. Wang, and B. Lei, “ Anatomy- Guided Multimodal Graph Networks for Alzheimer’s Disease: Integrative Analysis of Cross-Modal Brain Connectivity Signatures ,” inMedical Image Computing and Computer Assisted Intervention – MICCAI 2025, vol. LNCS 15...

    2025

  8. [8]

    Multi-modal multi-task learning for joint pre- diction of multiple regression and classification variables in alzheimer’s disease,

    D. Zhang and D. Shen, “Multi-modal multi-task learning for joint pre- diction of multiple regression and classification variables in alzheimer’s disease,”NeuroImage, vol. 59, no. 2, pp. 895–907, 2012

    2012

  9. [9]

    Holodx: Knowledge- and data- driven multimodal diagnosis of alzheimer’s disease,

    Q. Chen, J. Wang, G. Wang, and Y . Hong, “Holodx: Knowledge- and data- driven multimodal diagnosis of alzheimer’s disease,”IEEE Transactions on Medical Imaging, vol. 45, no. 1, pp. 206–219, 2026

    2026

  10. [10]

    Heterogeneous multimodal biomarkers analysis for alzheimer’s disease via bayesian network,

    Y . Jin, Y . Su, X.-H. Zhou, S. Huang, and T. A. D. N. Initiative, “Heterogeneous multimodal biomarkers analysis for alzheimer’s disease via bayesian network,”EURASIP Journal on Bioinformatics and Systems Biology, vol. 2016, no. 1, p. 12, 2016

    2016

  11. [11]

    Relationships between flortaucipir pet tau binding and amyloid burden, clinical diagnosis, age and cognition,

    M. J. Pontecorvo, M. D. Devous Sr, M. Navitsky, M. Lu, S. Salloway, F. W. Schaerf, D. Jennings, A. K. Aroraet al., “Relationships between flortaucipir pet tau binding and amyloid burden, clinical diagnosis, age and cognition,”Brain, vol. 140, no. 3, pp. 748–763, 01 2017

    2017

  12. [12]

    Uncovering the heterogeneity and temporal complexity of neurodegenerative diseases with subtype and stage inference,

    A. L. Young, R. V . Marinescu, N. P. Oxtoby, M. Bocchetta, K. Yong, N. C. Firth, D. M. Cash, D. L. Thomaset al., “Uncovering the heterogeneity and temporal complexity of neurodegenerative diseases with subtype and stage inference,”Nature Communications, vol. 9, no. 1, p. 4273, 2018

    2018

  13. [13]

    pysustain: A python implementation of the subtype and stage inference algorithm,

    L. M. Aksman, P. A. Wijeratne, N. P. Oxtoby, A. Eshaghi, C. Shand, A. Altmann, D. C. Alexander, and A. L. Young, “pysustain: A python implementation of the subtype and stage inference algorithm,”SoftwareX, vol. 16, p. 100811, 2021

    2021

  14. [14]

    Associations between quantitative[18f]flortaucipir tau pet and atrophy across the alzheimer’s disease spectrum,

    T. Timmers, R. Ossenkoppele, E. E. Wolters, S. C. J. Verfaillie, D. Visser, S. S. V . Golla, F. Barkhof, P. Scheltenset al., “Associations between quantitative[18f]flortaucipir tau pet and atrophy across the alzheimer’s disease spectrum,”Alzheimer’s Research & Therapy, vol. 11, no. 1, p. 60, 2019

    2019

  15. [15]

    Evaluation of tau imaging in staging alzheimer disease and revealing interactions betweenβ-amyloid and tauopathy,

    L. Wang, T. L. Benzinger, Y . Su, J. Christensen, K. Friedrichsen, P. Aldea, J. McConathy, N. J. Cairns, A. M. Fagan, J. C. Morris, and B. M. Ances, “Evaluation of tau imaging in staging alzheimer disease and revealing interactions betweenβ-amyloid and tauopathy,”JAMA Neurology, vol. 73, no. 9, pp. 1070–1077, 2016

    2016

  16. [16]

    The ANTsX ecosystem for quantitative biological and medical imaging,

    N. J. Tustison, P. A. Cook, A. J. Holbrook, H. J. Johnson, J. Muschelli, G. A. Devenyi, J. T. Duda, S. R. Daset al., “The ANTsX ecosystem for quantitative biological and medical imaging,”Scientific Reports, vol. 11, no. 1, p. 9068, 2021

    2021

  17. [17]

    Automated brain extraction of multisequence mri using artificial neural networks,

    F. Isensee, M. Schell, I. Pflueger, G. Brugnara, D. Bonekamp, U. Neu- berger, A. Wick, H.-P. Schlemmer, S. Heiland, W. Wick, M. Bendszus, K. H. Maier-Hein, and P. Kickingereder, “Automated brain extraction of multisequence mri using artificial neural networks,”Human Brain Mapping, vol. 40, no. 17, pp. 4952–4964, 2019

    2019

  18. [18]

    Statistical normalization techniques for magnetic resonance imaging,

    R. T. Shinohara, E. M. Sweeney, J. Goldsmith, N. Shiee, F. J. Mateen, P. A. Calabresi, S. Jarso, D. L. Pham, D. S. Reich, and C. M. Crainiceanu, “Statistical normalization techniques for magnetic resonance imaging,” NeuroImage: Clinical, vol. 6, pp. 9–19, 2014

    2014

  19. [19]

    Synthseg: Segmentation of brain mri scans of any contrast and resolution without retraining,

    B. Billot, D. N. Greve, O. Puonti, A. Thielscher, K. Van Leemput, B. Fischl, A. V . Dalca, and J. E. Iglesias, “Synthseg: Segmentation of brain mri scans of any contrast and resolution without retraining,”Medical Image Analysis, vol. 86, p. 102789, 2023

    2023

  20. [20]

    An automated labeling system for subdividing the human cerebral cortex on mri scans into gyral based regions of interest,

    R. S. Desikan, F. S ´egonne, B. Fischl, B. T. Quinn, B. C. Dickerson, D. Blacker, R. L. Buckner, A. M. Dale, R. P. Maguire, B. T. Hyman, M. S. Albert, and R. J. Killiany, “An automated labeling system for subdividing the human cerebral cortex on mri scans into gyral based regions of interest,”NeuroImage, vol. 31, no. 3, pp. 968–980, 2006

    2006

  21. [21]

    Petpvc: a toolbox for performing partial volume correction techniques in positron emission tomography,

    B. A. Thomas, V . Cuplov, A. Bousse, A. Mendes, K. Thielemans, B. F. Hutton, and K. Erlandsson, “Petpvc: a toolbox for performing partial volume correction techniques in positron emission tomography,”Physics in Medicine & Biology, vol. 61, no. 22, p. 7975, 2016

    2016

  22. [22]

    Comparison of multiple tau-pet measures as biomarkers in aging and alzheimer’s disease,

    A. Maass, S. Landau, S. L. Baker, A. Horng, S. N. Lockhart, R. La Joie, G. D. Rabinovici, and W. J. Jagust, “Comparison of multiple tau-pet measures as biomarkers in aging and alzheimer’s disease,”NeuroImage, vol. 157, pp. 448–463, 2017

    2017

  23. [23]

    Estimating mutual infor- mation,

    A. Kraskov, H. St ¨ogbauer, and P. Grassberger, “Estimating mutual infor- mation,”Phys. Rev. E, vol. 69, p. 066138, 2004

    2004

  24. [24]

    Normalized mutual information feature selection,

    P. A. Estevez, M. Tesmer, C. A. Perez, and J. M. Zurada, “Normalized mutual information feature selection,”IEEE Transactions on Neural Networks, vol. 20, no. 2, pp. 189–201, 2009

    2009

  25. [25]

    Controlling the false discovery rate: A practical and powerful approach to multiple testing,

    Y . Benjamini and Y . Hochberg, “Controlling the false discovery rate: A practical and powerful approach to multiple testing,”Journal of the Royal Statistical Society: Series B (Methodological), vol. 57, no. 1, pp. 289–300, 1995

    1995

  26. [26]

    Partial least squares (pls) methods for neuroimaging: A tutorial and review,

    A. Krishnan, L. J. Williams, A. R. McIntosh, and H. Abdi, “Partial least squares (pls) methods for neuroimaging: A tutorial and review,” NeuroImage, vol. 56, no. 2, pp. 455–475, 2011

    2011

  27. [27]

    A general approach to causal mediation analysis,

    K. Imai, L. Keele, and D. Tingley, “A general approach to causal mediation analysis,”Psychological methods, vol. 15, no. 4, pp. 309–334, 2010

    2010

  28. [28]

    Un- derstanding heterogeneity in clinical cohorts using normative models: Beyond case-control studies,

    A. F. Marquand, I. Rezek, J. K. Buitelaar, and C. F. Beckmann, “Un- derstanding heterogeneity in clinical cohorts using normative models: Beyond case-control studies,”Biological Psychiatry, vol. 80, no. 7, pp. 552–561, 2016

    2016