Academic Journal
View in EDS

TASNet: A tri-modal attentive scale robust adaptive fusion framework for progressive mild cognitive impairment classification.

Bibliographic Details
Title: TASNet: A tri-modal attentive scale robust adaptive fusion framework for progressive mild cognitive impairment classification.
Authors: Zhu H; School of Computer and Artificial Intelligence, Shandong Jianzhu University, Jinan, Shandong, China., Zhang L; School of Computer and Artificial Intelligence, Shandong Jianzhu University, Jinan, Shandong, China., Yang L; School of Computer and Artificial Intelligence, Shandong Jianzhu University, Jinan, Shandong, China., Hao F; School of Computer and Artificial Intelligence, Shandong Jianzhu University, Jinan, Shandong, China. Electronic address: 13731@sdjzu.edu.cn.
Source: Journal of neuroscience methods [J Neurosci Methods] 2026 Jun; Vol. 430, pp. 110726. Date of Electronic Publication: 2026 Feb 23.
Publication Type: Journal Article
Language: English
Journal Info: Publisher: Elsevier/North-Holland Biomedical Press Country of Publication: Netherlands NLM ID: 7905558 Publication Model: Print-Electronic Cited Medium: Internet ISSN: 1872-678X (Electronic) Linking ISSN: 01650270 NLM ISO Abbreviation: J Neurosci Methods Subsets: MEDLINE
Imprint Name(s): Original Publication: Amsterdam, Elsevier/North-Holland Biomedical Press.
MeSH Terms: Cognitive Dysfunction*/diagnostic imaging , Cognitive Dysfunction*/classification , Cognitive Dysfunction*/pathology , Magnetic Resonance Imaging*/methods , Positron-Emission Tomography*/methods , Brain*/diagnostic imaging , Brain*/pathology , Image Interpretation, Computer-Assisted*/methods , Neuroimaging*/methods , Neural Networks, Computer*, Alzheimer Disease/diagnostic imaging ; Humans ; Disease Progression
Abstract: Background: Accurate prediction of progressive Mild Cognitive Impairment (pMCI) versus stable MCI (sMCI) is crucial for early Alzheimer's disease (AD) intervention and delaying disease progression. The combination of Positron Emission Tomography (PET) and structural Magnetic Resonance Imaging (sMRI) is currently the most recognized diagnostic basis. However, PET and sMRI reveal lesions with heterogeneous scales, irregular spatial distributions, and inter-modal temporal asynchrony-factors causing inadequate feature extraction and poor fusion in existing methods.
New Method: To address these challenges, this paper introduces Tri-modal Attention Scale-robust Network (TASNet), an innovative feature-scale robust framework integrating brain imaging (PET, sMRI) with Cerebrospinal Fluid (CSF) data. TASNet adaptively fuses these modalities via symmetric dynamic channel attention in an end-to-end manner. A 3D Parallel Axial Fusion Attention (PAFA) mechanism and a Distribution-Variable Atrous Spatial Pyramid Pooling (DV-ASPP) module are designed to efficiently capture multi-scale irregular pathological features. Additionally, a Hardness-Aware Hybrid Loss (HAHL) function is introduced to enhance the discriminability of challenging cases.
Results: Experiments on the ADNI dataset resulted in an accuracy of 85.00% and an AUC of 91.67% in pMCI/sMCI classification. Furthermore, explainability analysis confirms the model focuses on AD pathology-consistent regions (e.g., temporal and parietal lobes).
Comparison With Existing Methods: The proposed TASNet demonstrates superior performance compared to state-of-the-art methods. By mitigating information loss typical of conventional serial models and effectively handling multi-scale features, the framework achieves higher classification metrics than existing approaches.
Conclusions: This work provides a technical solution for early AD prediction with high accuracy, reliability, and clinical interpretability.
(Copyright © 2026 Elsevier B.V. All rights reserved.)
Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Contributed Indexing: Keywords: Alzheimer’s disease; Attention mechanism; Deep learning; Mild cognitive impairment; Multimodal fusion
Entry Date(s): Date Created: 20260225 Date Completed: 20260709 Latest Revision: 20260709
Update Code: 20260711
DOI: 10.1016/j.jneumeth.2026.110726
PMID: 41740679
Database: MEDLINE
FullText Text:
  Availability: 0
CustomLinks:
  – Url: https://www.doi.org/10.1016/j.jneumeth.2026.110726?
    Name: ScienceDirect (all content) (s7799221)
    Category: fullText
    Text: View record from ScienceDirect
    MouseOverText: View record from ScienceDirect
  – Url: https://resolver.ebsco.com/c/fiv2js/result?sid=EBSCO:cmedm&genre=article&issn=1872678X&ISBN=&volume=430&issue=&date=20260601&spage=110726&pages=110726&title=Journal of neuroscience methods&atitle=TASNet%3A%20A%20tri-modal%20attentive%20scale%20robust%20adaptive%20fusion%20framework%20for%20progressive%20mild%20cognitive%20impairment%20classification.&aulast=Zhu%20H&id=DOI:10.1016/j.jneumeth.2026.110726
    Name: Full Text Finder (for New FTF UI) (ns324271)
    Category: fullText
    Text: Full Text Finder
    MouseOverText: Full Text Finder
Header DbId: cmedm
DbLabel: MEDLINE
An: 41740679
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: TASNet: A tri-modal attentive scale robust adaptive fusion framework for progressive mild cognitive impairment classification.
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AU" term="%22Zhu+H%22">Zhu H</searchLink>; School of Computer and Artificial Intelligence, Shandong Jianzhu University, Jinan, Shandong, China.<br /><searchLink fieldCode="AU" term="%22Zhang+L%22">Zhang L</searchLink>; School of Computer and Artificial Intelligence, Shandong Jianzhu University, Jinan, Shandong, China.<br /><searchLink fieldCode="AU" term="%22Yang+L%22">Yang L</searchLink>; School of Computer and Artificial Intelligence, Shandong Jianzhu University, Jinan, Shandong, China.<br /><searchLink fieldCode="AU" term="%22Hao+F%22">Hao F</searchLink>; School of Computer and Artificial Intelligence, Shandong Jianzhu University, Jinan, Shandong, China. Electronic address: 13731@sdjzu.edu.cn.
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="JN" term="%227905558%22">Journal of neuroscience methods</searchLink> [J Neurosci Methods] 2026 Jun; Vol. 430, pp. 110726. <i>Date of Electronic Publication: </i>2026 Feb 23.
– Name: TypePub
  Label: Publication Type
  Group: TypPub
  Data: Journal Article
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: TitleSource
  Label: Journal Info
  Group: Src
  Data: <i>Publisher: </i><searchLink fieldCode="PB" term="%22Elsevier%2FNorth-Holland+Biomedical+Press%22">Elsevier/North-Holland Biomedical Press </searchLink><i>Country of Publication: </i>Netherlands <i>NLM ID: </i>7905558 <i>Publication Model: </i>Print-Electronic <i>Cited Medium: </i>Internet <i>ISSN: </i>1872-678X (Electronic) <i>Linking ISSN: </i><searchLink fieldCode="IS" term="%2201650270%22">01650270 </searchLink><i>NLM ISO Abbreviation: </i>J Neurosci Methods <i>Subsets: </i>MEDLINE
– Name: PublisherInfo
  Label: Imprint Name(s)
  Group: PubInfo
  Data: <i>Original Publication</i>: Amsterdam, Elsevier/North-Holland Biomedical Press.
– Name: SubjectMESH
  Label: MeSH Terms
  Group: Su
  Data: <searchLink fieldCode="MM" term="%22Cognitive+Dysfunction%22">Cognitive Dysfunction*</searchLink>/<searchLink fieldCode="MM" term="%22Cognitive+Dysfunction+diagnostic+imaging%22">diagnostic imaging</searchLink> <br /><searchLink fieldCode="MM" term="%22Cognitive+Dysfunction%22">Cognitive Dysfunction*</searchLink>/<searchLink fieldCode="MM" term="%22Cognitive+Dysfunction+classification%22">classification</searchLink> <br /><searchLink fieldCode="MM" term="%22Cognitive+Dysfunction%22">Cognitive Dysfunction*</searchLink>/<searchLink fieldCode="MM" term="%22Cognitive+Dysfunction+pathology%22">pathology</searchLink> <br /><searchLink fieldCode="MM" term="%22Magnetic+Resonance+Imaging%22">Magnetic Resonance Imaging*</searchLink>/<searchLink fieldCode="MM" term="%22Magnetic+Resonance+Imaging+methods%22">methods</searchLink> <br /><searchLink fieldCode="MM" term="%22Positron-Emission+Tomography%22">Positron-Emission Tomography*</searchLink>/<searchLink fieldCode="MM" term="%22Positron-Emission+Tomography+methods%22">methods</searchLink> <br /><searchLink fieldCode="MM" term="%22Brain%22">Brain*</searchLink>/<searchLink fieldCode="MM" term="%22Brain+diagnostic+imaging%22">diagnostic imaging</searchLink> <br /><searchLink fieldCode="MM" term="%22Brain%22">Brain*</searchLink>/<searchLink fieldCode="MM" term="%22Brain+pathology%22">pathology</searchLink> <br /><searchLink fieldCode="MM" term="%22Image+Interpretation%2C+Computer-Assisted%22">Image Interpretation, Computer-Assisted*</searchLink>/<searchLink fieldCode="MM" term="%22Image+Interpretation%2C+Computer-Assisted+methods%22">methods</searchLink> <br /><searchLink fieldCode="MM" term="%22Neuroimaging%22">Neuroimaging*</searchLink>/<searchLink fieldCode="MM" term="%22Neuroimaging+methods%22">methods</searchLink> <br /><searchLink fieldCode="MM" term="%22Neural+Networks%2C+Computer%22">Neural Networks, Computer*</searchLink><br /><searchLink fieldCode="MH" term="%22Alzheimer+Disease%22">Alzheimer Disease</searchLink>/<searchLink fieldCode="MH" term="%22Alzheimer+Disease+diagnostic+imaging%22">diagnostic imaging</searchLink> ; <searchLink fieldCode="MH" term="%22Humans%22">Humans</searchLink> ; <searchLink fieldCode="MH" term="%22Disease+Progression%22">Disease Progression</searchLink>
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Background: Accurate prediction of progressive Mild Cognitive Impairment (pMCI) versus stable MCI (sMCI) is crucial for early Alzheimer's disease (AD) intervention and delaying disease progression. The combination of Positron Emission Tomography (PET) and structural Magnetic Resonance Imaging (sMRI) is currently the most recognized diagnostic basis. However, PET and sMRI reveal lesions with heterogeneous scales, irregular spatial distributions, and inter-modal temporal asynchrony-factors causing inadequate feature extraction and poor fusion in existing methods.<br />New Method: To address these challenges, this paper introduces Tri-modal Attention Scale-robust Network (TASNet), an innovative feature-scale robust framework integrating brain imaging (PET, sMRI) with Cerebrospinal Fluid (CSF) data. TASNet adaptively fuses these modalities via symmetric dynamic channel attention in an end-to-end manner. A 3D Parallel Axial Fusion Attention (PAFA) mechanism and a Distribution-Variable Atrous Spatial Pyramid Pooling (DV-ASPP) module are designed to efficiently capture multi-scale irregular pathological features. Additionally, a Hardness-Aware Hybrid Loss (HAHL) function is introduced to enhance the discriminability of challenging cases.<br />Results: Experiments on the ADNI dataset resulted in an accuracy of 85.00% and an AUC of 91.67% in pMCI/sMCI classification. Furthermore, explainability analysis confirms the model focuses on AD pathology-consistent regions (e.g., temporal and parietal lobes).<br />Comparison With Existing Methods: The proposed TASNet demonstrates superior performance compared to state-of-the-art methods. By mitigating information loss typical of conventional serial models and effectively handling multi-scale features, the framework achieves higher classification metrics than existing approaches.<br />Conclusions: This work provides a technical solution for early AD prediction with high accuracy, reliability, and clinical interpretability.<br /> (Copyright © 2026 Elsevier B.V. All rights reserved.)
– Name: Abstract
  Label: Competing Interests
  Group: Ab
  Data: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
– Name: SubjectMinor
  Label: Contributed Indexing
  Group:
  Data: <i>Keywords: </i>Alzheimer’s disease; Attention mechanism; Deep learning; Mild cognitive impairment; Multimodal fusion
– Name: DateEntry
  Label: Entry Date(s)
  Group: Date
  Data: <i>Date Created: </i>20260225 <i>Date Completed: </i>20260709 <i>Latest Revision: </i>20260709
– Name: DateUpdate
  Label: Update Code
  Group: Date
  Data: 20260711
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1016/j.jneumeth.2026.110726
– Name: AN
  Label: PMID
  Group: ID
  Data: 41740679
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=cmedm&AN=41740679
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1016/j.jneumeth.2026.110726
    Languages:
      – Code: eng
        Text: English
    PhysicalDescription:
      Pagination:
        StartPage: 110726
    Subjects:
      – SubjectFull: Alzheimer Disease diagnostic imaging
        Type: general
      – SubjectFull: Humans
        Type: general
      – SubjectFull: Disease Progression
        Type: general
      – SubjectFull: Cognitive Dysfunction diagnostic imaging
        Type: general
      – SubjectFull: Cognitive Dysfunction classification
        Type: general
      – SubjectFull: Cognitive Dysfunction pathology
        Type: general
      – SubjectFull: Magnetic Resonance Imaging methods
        Type: general
      – SubjectFull: Positron-Emission Tomography methods
        Type: general
      – SubjectFull: Brain diagnostic imaging
        Type: general
      – SubjectFull: Brain pathology
        Type: general
      – SubjectFull: Image Interpretation, Computer-Assisted methods
        Type: general
      – SubjectFull: Neuroimaging methods
        Type: general
      – SubjectFull: Neural Networks, Computer
        Type: general
    Titles:
      – TitleFull: TASNet: A tri-modal attentive scale robust adaptive fusion framework for progressive mild cognitive impairment classification.
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Zhu H
      – PersonEntity:
          Name:
            NameFull: Zhang L
      – PersonEntity:
          Name:
            NameFull: Yang L
      – PersonEntity:
          Name:
            NameFull: Hao F
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 06
              Text: 2026 Jun
              Type: published
              Y: 2026
          Identifiers:
            – Type: issn-electronic
              Value: 1872-678X
          Numbering:
            – Type: volume
              Value: 430
          Titles:
            – TitleFull: Journal of neuroscience methods
              Type: main
ResultId 1