The DOI matches its Crossref record; 174 of the 176 references it cites were found in a registry or an index; 36 of the 125 sources its claims cite were checked for support.
Multiclass characterization of frontotemporal dementia variants via multimodal brain network computational inference
Open accessAbstract
Characterizing a particular neurodegenerative condition against others possible diseases remains a challenge along clinical, biomarker, and neuroscientific levels. This is the particular case of frontotemporal dementia (FTD) variants, where their specific characterization requires high levels of expertise and multidisciplinary teams to subtly distinguish among similar physiopathological processes. Here, we used a computational approach of multimodal brain networks to address simultaneous multiclass classification of 298 subjects (one group against all others), including five FTD variants: behavioral variant FTD, corticobasal syndrome, nonfluent variant primary progressive aphasia, progressive supranuclear palsy, and semantic variant primary progressive aphasia, with healthy controls. Fourteen machine learning classifiers were trained with functional and structural connectivity metrics calculated through different methods. Due to the large number of variables, dimensionality was reduced, employing statistical comparisons and progressive elimination to assess feature stability under nested cross-validation. The machine learning performance was measured through the area under the receiver operating characteristic curves, reaching 0.81 on average, with a standard deviation of 0.09. Furthermore, the contributions of demographic and cognitive data were also assessed via multifeatured classifiers. An accurate simultaneous multiclass classification of each FTD variant against other variants and controls was obtained based on the selection of an optimum set of features. The classifiers incorporating the brain’s network and cognitive assessment increased performance metrics. Multimodal classifiers evidenced specific variants’ compromise, across modalities and methods through feature importance analysis. If replicated and validated, this approach may help to support clinical decision tools aimed to detect specific affectations in the context of overlapping diseases.
Overview
CoChat Summary
TL;DR
This study used multimodal functional and structural brain-network features plus demographic/cognitive variables to perform simultaneous multiclass classification of five frontotemporal dementia variants and healthy controls. XGBoost models achieved up to AUC = 0.95, with cognitive features generally improving performance and feature-importance analyses highlighting variant-specific network alterations.
Key findings
- In 298 subjects (HC n = 99; bvFTD n = 47; CBS n = 38; nfvPPA n = 34; PSP n = 42; svPPA n = 38), multiclass classification reached an average AUC of 0.81 (SD = 0.09) across models, with the best model—raw functional connectivity plus cognitive features—reaching AUC = 0.95.
- Adding cognitive/demographic features increased performance across modalities: raw functional connectivity rose from AUC = 0.92 to 0.95; ROI-to-ROI functional from 0.75 to 0.80; functional graph from 0.63 to 0.89; raw structural from 0.84 to 0.85; ROI-to-ROI structural from 0.70 to 0.80; structural graph from 0.73 to 0.83; and all combined connectivity from 0.80 to 0.89.
- The top three models were the raw functional multifeatured model, the raw functional connectivity model, and the multimodal multifeatured model; differences among these top models were not statistically significant (p > 0.05).
- Feature-importance analyses showed variant-specific patterns, including frontal/insular/temporal alterations in bvFTD, speech-language and left-hemisphere regions in nfvPPA, temporal disconnection in svPPA, motor/parietal areas in CBS, and thalamic connections in PSP.
Methods
- Data came from LONI’s NIFD and 4RTNI databases; clinical diagnosis followed current criteria, and healthy controls were split into two matched subsamples (HCsub1 and HCsub2, each n = 50) for age, sex, and education matching.
- Brain-network features included two modalities (functional and structural connectivity) and three methods (raw/voxel-level, ROI-to-ROI, and graph connectivity), derived from rs-fMRI and DWI; structural DWI data were available for 264 subjects.
- Fourteen XGBoost models were trained for one-vs-all multiclass classification, with dimensionality reduction via group-level statistical filtering and progressive feature elimination under nested cross-validation; performance was evaluated with microaverage AUC, using 80% training/validation and 20% independent testing.
Limitations
- The paper states that the sample came from a single database, lacked histopathological diagnosis confirmation, did not compare against standard PET/plasma biomarkers, used only the MMSE rather than FTD-specific cognitive tests, relied mainly on the AAL atlas (with a representative HCP-atlas comparison), and included mostly early-to-moderate-stage patients.
- The text also makes evident that the study is cross-sectional, so it cannot assess disease progression or longitudinal monitoring, and that despite scanner-variability checks, MRI data were acquired from two scanners.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures










Verification
Claims the paper makes
18 Valid · 1 Not valid · 41 Not checked · 98 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“In particular, the clinical diagnosis of each frontotemporal dementia (FTD) variant requires high levels of expertise and multidisciplinary teams to distinguish among subtle phenotypes with similar underlying physiopathological processes (Bejanin et al., 2020; Staffaroni et al., 2019; Zetterberg et al., 2019).”
§ Multiclass characterization of frontotemporal dementia variants via multimodal brain network computational inference > INTRODUCTION
What the sources say
- Partly supportedcites Bejanin et al., 2020 sourceExperimental — accuracy not yet measured
The Participants section (page unavailable) says this study established FTD-variant diagnoses by multidisciplinary consensus using diagnostic criteria. The paper does not say that such teams or high levels of expertise are required, or that they are needed to distinguish subtle phenotypes with similar underlying processes; its introductory passage instead describes distinct early patterns of neurodegeneration. (LLM, conf=0.86)
- Partly supportedcites Staffaroni et al., 2019 sourceExperimental — accuracy not yet measured
The Introduction describes distinct FTD clinical syndromes that can share tau or TDP-43 pathology. The Participants section says patients were diagnosed at a multidisciplinary consensus conference. The paper does not establish that diagnosing every variant requires such a team or a high level of expertise, or that the phenotypic differences are subtle. (LLM, conf=0.86)
- Not checked yetcites Zetterberg et al., 2019 source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Clinical diagnosis of the FTD variants was based on current criteria (Armstrong et al., 2013; Gorno-Tempini et al., 2011; Rascovsky & Grossman, 2013; Rascovsky et al., 2011).”
§ Multiclass characterization of frontotemporal dementia variants via multimodal brain network computational inference > MATERIALS AND METHODS > Subjects
What the sources say
- Partly supportedcites Armstrong et al., 2013 sourceExperimental — accuracy not yet measured
The CBD diagnostic criteria section defines frontal behavioral-spatial syndrome and nonfluent/agrammatic primary progressive aphasia as clinical phenotypes used in diagnosing CBD. Armstrong et al. therefore supports criteria relevant to some presentations, but it does not provide diagnostic criteria for FTD variants generally. No page number is available in our copy. (LLM, conf=0.84)
- Supportedcites Gorno-Tempini et al., 2011 source
The paper sets out clinical diagnostic criteria for the nonfluent/agrammatic and semantic variants of primary progressive aphasia (Tables 2 and 3). Its terminology section identifies PPA as a clinical presentation within the frontotemporal spectrum, so this citation supports the criteria for those variants, rather than for every FTD variant. (LLM, conf=0.91)
- Supportedcites Rascovsky & Grossman, 2013 source
The Introduction identifies revised diagnostic criteria for the behavioural variant of FTD and for primary progressive aphasia, and says the review summarizes those guidelines. The Conclusions likewise describes revised criteria for bvFTD and PPA. This supports citing the review for the current clinical criteria used to diagnose those variants. (LLM, conf=0.90)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Although two different scanners were used (Siemens model Trio; scanner 1 = 285 subjects, 95.6 %; and scanner 2 = 13 subjects, 4.4 %), previous studies with this dataset have pointed out that this variability is negligible (Dickerson et al., 2008; Fox et al., 2012; Melzer et al., 2020; Noble et al., 2017a; Noble et al., 2017b; Zhou et al., 2018).”
§ Multiclass characterization of frontotemporal dementia variants via multimodal brain network computational inference > MATERIALS AND METHODS > MRI: Analysis
What the sources say
- Partly supportedcites Dickerson et al., 2008 sourceExperimental — accuracy not yet measured
The Discussion reports highly reliable cortical-thickness correlates of cognitive performance across scan sessions and scanner platforms, supporting a limited claim that scanner differences did not substantially affect those findings. But the Materials and methods describe 15 participants scanned on Siemens Sonata and Trio and GE Signa scanners—not the citing study’s 298 subjects on two Trio scanners. The paper does not establish that scanner variability was negligible in the citing study’s dataset. (LLM, conf=0.88)
- Partly supportedcites Fox et al., 2012 sourceExperimental — accuracy not yet measured
The Discussion reports strong concordance for several DTI measures across scanners, which supports limited scanner comparability. But the MR Imaging Scanners section describes a study of two volunteers on five scanners, not the citing paper’s 298-subject dataset; the Longitudinal Scan and Software Upgrade Comparison section also reports cross-scanner differences in diffusivity. This paper therefore does not establish that scanner variability in the cited dataset was negligible. (LLM, conf=0.87)
- Partly supportedcites Melzer et al., 2020 sourceExperimental — accuracy not yet measured
The Discussion reports negligible systematic bias after a scanner was relocated, and the Results found no systematic difference in variance between locations. But the Introduction and Methods describe the same scanner before and after relocation, tested in 20 volunteers—not variability between two different scanners in the citing study’s dataset. The source therefore supports the narrower relocation finding, not the dataset-specific claim. (LLM, conf=0.88)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Unlike other methods for dimensionality reduction, such as principal component analysis, this approach allows a more direct interpretation of the results (Huys et al., 2016; Pereira et al., 2009).”
§ Multiclass characterization of frontotemporal dementia variants via multimodal brain network computational inference > MATERIALS AND METHODS > Machine Learning
What the sources say
- Could not verifycites Huys et al., 2016 sourceExperimental — accuracy not yet measured
Box 1, “Unsupervised dimensionality reduction” (page unavailable), identifies principal component analysis as a dimensionality-reduction method and discusses selecting variables through regularization. Neither that passage nor the passages found by searching for filter methods and interpretability compare how directly their results can be interpreted. (LLM, conf=0.72)
- Partly supportedcites Pereira et al., 2009 sourceExperimental — accuracy not yet measured
Section 3.3 describes filter methods as ranking and selecting individual features, while Section 3.1 says PCA transforms original voxel features into a new feature space. That distinction is consistent with more direct interpretation of selected voxels, but the paper does not explicitly make the claimed comparison about interpretability. Page numbers are unavailable in our copy. (LLM, conf=0.78)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Could not verify
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“The few studies with multiclass comparisons of FTD variants were conducted only with atrophy metrics (Kim et al., 2019) and tractography (Torso et al., 2020, 2021).”
§ Multiclass characterization of frontotemporal dementia variants via multimodal brain network computational inference > DISCUSSION
What the sources say
- Supportedcites Kim et al., 2019 source
The Abstract Methods describe classifiers comparing bvFTD with PPA and svPPA with nfvPPA using cortical-thickness data from MRI. The Conclusion states that cortical-thickness data alone classified FTD clinical subtypes. This supports the atrophy-metrics portion attributed to Kim et al.; the tractography portion is cited to other papers. (LLM, conf=0.92)
- Not matchedcites Torso et al., 2020, 2021
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
93 more claims · 8 with their cited sources checked
Create a free account to see all 158 claims and to verify the 86 references still open.
- Probably not a substantial claim. 36
- It may state two separate claims as one. 33
- The quote only partly states this claim. 3
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“The current clinical diagnostic criteria, however, is the current gold standard for clinical diagnosis within the literature and has been used globally to diagnose AD and bvFTD.53,56,101,102”
From Altered spatiotemporal brain dynamics of interoception in behavioural-variant frontotemporal dementia (2025) · cites this paper as [101]
Experimental — accuracy not yet measuredIn the citing paper
“Although the AAL3 has been systematically used in neurodegenerative research101,107, 108, 109 and yields similar results when compared to other parcellation methods,101 it is purely anatomical (i.e., does not provide functional information).”
From Altered spatiotemporal brain dynamics of interoception in behavioural-variant frontotemporal dementia (2025) · cites this paper as [101]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Multiclass characterization of frontotemporal dementia variants via multimodal brain network computational inference > MATERIALS AND METHODS > MRI: Analysis
“seven graph connectivity metrics were analyzed for each ROI:”
§ Multiclass characterization of frontotemporal dementia variants via multimodal brain network computational inference > MATERIALS AND METHODS > MRI: Analysis
“(5) clustering coefficient: fraction of node’s neighbors that are neighbors of each other (Bassett & Bullmore, 2006); (6) betweenness centrality: the fraction of all shortest paths in the network that contain a given node (Rubinov & Sporns, 2010).”
Citations
Cites
Statistical Power Analysis for the Behavioral Sciences
Cohen, J. (2013) · doi:10.4324/9780203771587 · Cited by 25,373
An introduction to ROC analysis
Fawcett, T. (2006) · Pattern Recognition Letters · doi:10.1016/j.patrec.2005.10.010 · Cited by 21,879
Tzourio-Mazoyer, N., Landeau, B., Papathanassiou, D., Crivello, F., Etard, O., Delcroix, N., Mazoyer, B., & Joliot, M. (2002) · NeuroImage · doi:10.1006/nimg.2001.0978 · Cited by 16,319
Complex brain networks: graph theoretical analysis of structural and functional systems
Bullmore, E., & Sporns, O. (2009) · Nature Reviews Neuroscience · doi:10.1038/nrn2575 · Cited by 11,820
Complex network measures of brain connectivity: Uses and interpretations
Rubinov, M., & Sporns, O. (2010) · NeuroImage · doi:10.1016/j.neuroimage.2009.10.003 · Cited by 11,522
Cohen, J. (1992) · Psychological Bulletin · doi:10.1037//0033-2909.112.1.155 · Cited by 8,771
Spontaneous fluctuations in brain activity observed with functional magnetic resonance imaging
Fox, M. D., & Raichle, M. E. (2007) · Nature Reviews Neuroscience · doi:10.1038/nrn2201 · Cited by 6,700
MR diffusion tensor spectroscopy and imaging
Basser, P. J., Mattiello, J., & LeBihan, D. (1994) · Biophysical Journal · doi:10.1016/s0006-3495(94)80775-1 · Cited by 5,970
SMITH, S., & NICHOLS, T. (2009) · NeuroImage · doi:10.1016/j.neuroimage.2008.03.061 · Cited by 5,250 · cited 3×
“The multiple comparison problem was accounted for by using a threshold-free cluster enhancement (TFCE) method (Chen et al., 2018; Smith & Nichols, 2009), except for the graph measures, where we used the false discovery rate (FDR) method with p < 0.05.”
InfluentialSensitivity of revised diagnostic criteria for the behavioural variant of frontotemporal dementia
Rascovsky, K., Hodges, J. R., Knopman, D., Mendez, M. F., Kramer, J. H., Neuhaus, J., van Swieten, J. C., Seelaar, H., Dopper, E. G. P., Onyike, C. U., et al. (2011) · Brain · doi:10.1093/brain/awr179 · Cited by 5,221
Research works printed but not linked: 2
- Koyejo, 2015, Advances in neural information processing systems
- Wade, 2020, Hands-on gradient boosting with XGBoost and scikit-learn: Perform accessible machine learning and extreme gradient boosting with Python
Cited by
A synergetic turn in cognitive neuroscience of brain diseases
Ibanez, A., Kringelbach, M. L., & Deco, G. (2024) · Trends in Cognitive Sciences · doi:10.1016/j.tics.2023.12.006 · Cited by 65
Educational disparities in brain health and dementia across Latin America and the United States
Gonzalez‐Gomez, R., Legaz, A., Moguilner, S., Cruzat, J., Hernández, H., Baez, S., Cocchi, R., Coronel‐Olivero, C., Medel, V., Tagliazuchi, E., et al. (2024) · Alzheimer's & Dementia · doi:10.1002/alz.14085 · Cited by 55
Self-citationModel-based whole-brain perturbational landscape of neurodegenerative diseases
Perl, Y. S., Fittipaldi, S., Campo, C. G., Moguilner, S., Cruzat, J., Fraile-Vazquez, M. E., Herzog, R. A., Kringelbach, M. L., Deco, G., Prado, P., et al. (2023) · eLife · doi:10.7554/elife.83970 · Cited by 52
Self-citationMoguilner, S., Whelan, R., Adams, H., Valcour, V., Tagliazucchi, E., & Ibáñez, A. (2023) · eBioMedicine · doi:10.1016/j.ebiom.2023.104540 · Cited by 39
Self-citationHazelton, J. L., Della Bella, G., Barttfeld, P., Dottori, M., Gonzalez-Gomez, R., Migeot, J., Moguilner, S., Legaz, A., Hernandez, H., Prado, P., et al. (2025) · eBioMedicine · doi:10.1016/j.ebiom.2025.105614 · Cited by 14
Self-citation
Metadata
- Authors
- Gonzalez-Gomez, R., Ibañez, A., & Moguilner, S.
- Affiliations
- Latin American Brain Health Institute (BrainLat), Universidad Adolfo Ibañez, Santiago de Chile, Chile
- Center for Social and Cognitive Neuroscience, School of Psychology, Universidad Adolfo Ibañez, Santiago de Chile, Chile
- ORCID
- Moguilner, Sebastian
- Venue
- Network Neuroscience
- Publisher
- MIT Press
- Published
- 2023-01-01
- Volume
- 7
- Issue
- 1
- Pages
- 322-350
- ISSN
- 2472-1751 (ISSN-L)
- PMC
- PMC10270711
- OpenAlex
- W4304956754
- Cited by
- 10
- References
- 176
- Citations by year
- 2026: 1 · 2025: 2 · 2024: 5 · 2023: 2
- FWCI
- 1.16
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Network Neuroscience (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- Takeda Pharmaceutical Company (CW2680521)
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › Functional Brain Connectivity Studies
- Health Sciences › Medicine › Psychiatry and Mental health › Dementia and Cognitive Impairment Research
- Health Sciences › Medicine › Clinical Neurology › Traumatic Brain Injury Research
- Keywords
- frontotemporal dementia, multi-class classification, multimodal brain networks, functional connectivity, structural connectivity, behavioral variant FTD, corticobasal syndrome, primary progressive aphasia, progressive supranuclear palsy, semantic dementia, feature selection, feature importance, neurodegenerative disease diagnosis
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-05, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-05 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Multiclass characterization of frontotemporal dementia variants via multimodal brain network computational inference. https://app.cochat.ai/papers/10.1162/netn_a_00285
- The paper
- Gonzalez-Gomez, R., Ibañez, A., & Moguilner, S. (2023). Multiclass characterization of frontotemporal dementia variants via multimodal brain network computational inference. Network Neuroscience. https://doi.org/10.1162/netn_a_00285
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.