MPS-FFA: A multiplane and multiscale feature fusion attention network for Alzheimer's disease prediction with
Fei Liu1, Huabin Wang1, Shiuan-Ni Liang2
1Anhui Provincial International Joint Research Center for Advanced Technology in Medical Imaging, Anhui University, Hefei, China; School of Computer Science and Technology, Anhui University, Hefei, China.
Abstract:
Structural magnetic resonance imaging (sMRI) is a popular technique that is widely applied in Alzheimer's disease (AD) diagnosis. However, only a few structural atrophy areas in sMRI scans are highly associated with AD. The degree of atrophy in patients' brain tissues and the distribution of lesion areas differ among patients. Therefore, a key challenge in sMRI-based AD diagnosis is identifying discriminating atrophy features. Hence, we propose a multiplane and multiscale feature-level fusion attention (MPS-FFA) model. The model has three components, (1) A feature encoder uses a multiscale feature extractor with hybrid attention layers to simultaneously capture and fuse multiple pathological features in the sagittal, coronal, and axial planes. (2) A global attention classifier combines clinical scores and two global attention layers to evaluate the feature impact scores and balance the relative contributions of different feature blocks. (3) A feature similarity discriminator minimizes the feature similarities among heterogeneous labels to enhance the ability of the network to discriminate atrophy features. The MPS-FFA model provides improved interpretability for identifying discriminating features using feature visualization. The experimental results on the baseline sMRI scans from two databases confirm the effectiveness (e.g., accuracy and generalizability) of our method in locating pathological locations. The source code is available at https://github.com/LiuFei-AHU/MPSFFA.
Insights
This study introduces a novel AI model for Alzheimer's disease (AD) diagnosis using structural magnetic resonance imaging (sMRI). The multiplane and multiscale feature-level fusion attention (MPS-FFA) model effectively identifies subtle brain atrophy patterns for improved diagnostic accuracy.
Area of Science:
- Neuroimaging
- Artificial Intelligence
- Medical Diagnostics
Background:
- Structural magnetic resonance imaging (sMRI) is crucial for Alzheimer's disease (AD) diagnosis, but identifying specific atrophy patterns remains challenging due to inter-patient variability.
- Existing methods struggle to pinpoint the most discriminating atrophy features in sMRI scans, hindering accurate AD diagnosis.
Purpose of the Study:
- To develop and validate a novel deep learning model, the multiplane and multiscale feature-level fusion attention (MPS-FFA) model, for enhanced AD diagnosis using sMRI.
- To improve the identification and localization of discriminating brain atrophy features in sMRI scans for Alzheimer's disease.
Main Methods:
- The proposed MPS-FFA model employs a feature encoder with multiscale feature extraction and hybrid attention to fuse pathological features across sagittal, coronal, and axial planes.
- A global attention classifier integrates clinical scores and attention layers to weigh feature importance, while a feature similarity discriminator enhances discrimination of atrophy features.
- Feature visualization techniques are utilized within the MPS-FFA model to improve interpretability in identifying key diagnostic features.
Main Results:
- Experimental results on two independent sMRI databases demonstrate the effectiveness of the MPS-FFA model in terms of accuracy and generalizability.
- The model successfully identified pathological locations associated with Alzheimer's disease, showcasing its capability in precise localization.
- Improved interpretability was achieved through feature visualization, aiding in the understanding of the model's diagnostic reasoning.
Conclusions:
- The MPS-FFA model represents a significant advancement in sMRI-based Alzheimer's disease diagnosis by effectively identifying and discriminating subtle atrophy features.
- The model's multiplane and multiscale approach, combined with attention mechanisms, enhances diagnostic accuracy and generalizability.
- The developed model offers improved interpretability, facilitating a better understanding of the neuroimaging markers crucial for AD diagnosis.
