Clinical Report: APLG-Net: A Hybrid Network for Classifying NC, MCI, and AD
Overview
APLG-Net classifies normal controls, mild cognitive impairment, and Alzheimer's disease using structural MRI. The model achieves accuracy and balanced metrics, outperforming existing classification methods.
Background
Alzheimer's disease (AD) is a leading cause of dementia, and accurate classification among normal cognition (NC), mild cognitive impairment (MCI), and AD is crucial for early diagnosis and intervention. Structural MRI provides valuable insights into brain morphology, yet challenges remain in distinguishing subtle anatomical variations associated with these conditions. The development of advanced models like APLG-Net aims to enhance diagnostic accuracy and patient stratification.
Data Highlights
Metric
Value
Accuracy
87.1%
Balanced Accuracy
86.4%
Macro-F1
86.8%
MCI F1
85.6%
Key Findings
APLG-Net integrates a global whole-brain encoder and a local ROI-based encoder.
The model employs cross-attention fusion and vector-gated integration for enhanced feature interaction.
Ordinal supervision is introduced to model disease progression effectively.
APLG-Net outperforms CNN-based, Transformer-based, and hybrid baselines on the ADNI dataset.
