D-GATNet: Interpretable Temporal Graph Attention Learning for ADHD Identification Using Dynamic Functional Connectivity

Dimensions overall score 7.0

• D-GATNet is a temporal graph-based deep learning framework for ADHD classification using dynamic functional connectivity (dFC).method 95%
  Evidencepartial

  In this work, we propose D-GATNet, an interpretable temporal graph-based framework for automated ADHD classification using dynamic functional connectivity (dFC).

  Implicationpartial

  This is the core proposal of the paper, explicitly stated in the abstract and introduction.

  Verificationpartial

  partial

• Dynamic functional connectivity (dFC) is extracted using a sliding-window Pearson correlation approach with a window size of 40 and a step size of 20.method 95%
  Evidencepartial

  Specifically, a temporal window of sizew= 40with step size s= 20is applied to the ROI time series.

  Implicationpartial

  The abstract and methodology section clearly describe the method for dFC extraction.

  Verificationpartial

  partial

• The framework utilizes a multi-layer Graph Attention Network (GAT) to model spatial dependencies among brain regions.method 95%
  Evidencepartial

  Spatial dependencies are learned via a multi-layer Graph Attention Network, while temporal dynamics are modeled using 1D convolution followed by temporal attention.

  Implicationpartial

  The abstract and methodology section explicitly mention the use of GAT for spatial modeling.

  Verificationpartial

  partial

• Temporal dynamics of connectivity are captured using 1D convolution followed by temporal attention.method 95%
  Evidencepartial

  while temporal dynamics are modeled using 1D convolution followed by temporal attention.

  Implicationpartial

  The abstract and methodology section clearly state the approach for temporal modeling.

  Verificationpartial

  partial

• The D-GATNet framework achieved 85.18% +_5.64 balanced accuracy and 0.881 AUC on the Peking University site of the ADHD-200 dataset.result 95%
  Evidencepartial

  achieved 85.18% +_5.64 balanced accuracy and 0.881 AUC, outperforming state-of-the-art methods.

  Implicationpartial

  These performance metrics are explicitly stated in the abstract with confidence intervals.

  Verificationpartial

  partial

• The interpretability of D-GATNet is achieved through graph attention weights, ROI importance scores, and temporal attention.technical 95%
  Evidencepartial

  Interpretability is achieved through graph attention weights revealing dominant ROI interactions, ROI importance scores identifying influential regions, and temporal attention emphasizing informative connectivity segments.

  Implicationpartial

  The abstract explicitly lists the mechanisms for interpretability.

  Verificationpartial

  partial

• The study uses the Peking University site of the ADHD-200 dataset, comprising 194 subjects (116 Healthy Controls and 78 ADHD patients).method 95%
  Evidencepartial

  Resting-state fMRI (rs-fMRI) scans have been obtained from the Peking University (PU) site of the publicly available ADHD-200 dataset, which includes 194 subjects (116 HCs and 78 ADHD patients).

  Implicationpartial

  The dataset details are clearly provided in the experimental evaluation section.

  Verificationpartial

  partial

• The brain is parcellated into 116 regions using the Automated Anatomical Labeling (AAL) atlas for ROI-based analysis.method 95%
  Evidencepartial

  An ROI-based analysis is adopted, where the brain is parcellated using the Automated Anatomical Labeling (AAL) atlas [20] into 116 regions, enabling extraction of regional time series for subsequent functional connectivity modeling.

  Implicationpartial

  The methodology section specifies the atlas and number of ROIs used.

  Verificationpartial

  partial

Competitive landscape

An interpretable temporal graph attention network for ADHD identification using dynamic brain connectivity, outperforming existing methods.