Table 4 Comparison between the proposed Model and State of the Art models using ADNI Dataset.
From: Revolutionizing Alzheimer’s disease detection with a cutting-edge CAPCBAM deep learning framework
Model | Architecture | Accuracy | Strengths | Limitations |
|---|---|---|---|---|
CAPCBAM (Our study) | Capsule Networks + CBAM | 99.95% | Preserves spatial hierarchies, dynamic routing, enhanced feature selection | Computationally expensive |
Ensemble CNN (Fathi et al., 2024)23 | CNN-based Ensemble | 99.83% | Strong generalization, robust classification | High model complexity |
Conv-Swinformer (Hu et al., 2023)24 | CNN + Transformer | N/A | Shift window attention enhances spatial feature extraction | Requires extensive training data |
3D DCGAN (Kang et al., 2023)25 | Generative Adversarial Network (GAN) | 92.8% | Improves learning on small datasets | Prone to mode collapse |
Feature Fusion Network (Illakiya & Karthik, 2024)26 | GSDW + GCN + EfficientNet-B0 | 77.2% | Multi-focus attention improves MCI detection | Lower accuracy compared to CNN-based models |
Thayumanasamy & Ramamurthy (2022)27 | Machine Learning & Deep Learning Hybrid | 96.1% | Combines ML and DL for robust classification | Computational cost and data dependency |
Illakiya & Karthik (2023)28 | Swin Transformer + Attention Network | 94.5% | Captures long-range dependencies, enhances MCI classification | Requires significant computational resources |
