Alzheimer's disease (AD) is a fatal disorder that destroys memory and cognitive skills in the elderly population, according to the research paper 'ProMUSE: Progressive Multi-modal Uncertainty-guided Staged Evidential Alzheimer Disease Classification' on arXiv. Most treatments for AD are effective in the early stage, creating an increasing demand for early AD diagnosis. However, diagnosis increasingly relies on multimodal data such as clinical assessments, structural Magnetic Resonance Imaging (MRI), and Positron Emission Tomography (PET) imaging. MRI and PET acquisition remain costly and not universally accessible, making full-modality inference impractical in real-world clinical workflows.
The ProMUSE Approach
The researchers propose ProMUSE, a Progressive Multi-modal Uncertainty Guided Staged Evidential Network that adaptively determines when additional modalities are necessary, helping reduce the overall cost of data acquisition while maintaining accuracy. ProMUSE first performs evidential classification using low-cost clinical data and quantifies uncertainty via a Dirichlet-based subjective logic model. When uncertainty exceeds a learned threshold, ProMUSE progressively incorporates MRI or PET features, fusing modality-wise belief and uncertainty through Dempster-Shafer theory to obtain a calibrated multimodal prediction. This staged acquisition strategy enables accurate diagnosis while minimizing reliance on expensive imaging.
Experimental Results
Experiments on three datasets — ADNI, AIBL, and OASIS — across three classification tasks demonstrate ProMUSE's effectiveness. The tasks are:
- CN-AD (cognitive normal vs. Alzheimer's disease)
- CN-MCI (cognitive normal vs. mild cognitive impairment)
- MCI-AD (mild cognitive impairment vs. Alzheimer's disease)
The results show that ProMUSE achieves competitive or superior accuracy compared to full-modality baselines while reducing MRI/PET usage by 50-90%, yielding substantial cost savings. The following table summarizes the key outcomes:
| Task | Datasets | MRI/PET Usage Reduction | Accuracy vs. Full-Modality |
|---|---|---|---|
| CN-AD | ADNI, AIBL, OASIS | 50-90% | Competitive or superior |
| CN-MCI | ADNI, AIBL, OASIS | 50-90% | Competitive or superior |
| MCI-AD | ADNI, AIBL, OASIS | 50-90% | Competitive or superior |
Implications for Enterprise Technology
While ProMUSE is developed for clinical Alzheimer screening, its uncertainty-guided staged acquisition framework has broader applicability. Enterprise technology leaders evaluating AI for diagnostic or resource-intensive processes can learn from ProMUSE's design: using low-cost data first, quantifying prediction uncertainty, and progressively incorporating expensive data only when needed. This approach reduces operational costs and improves efficiency. ProMUSE uses a Dirichlet-based subjective logic model for uncertainty quantification and Dempster-Shafer theory for belief fusion — techniques that could be adapted for supply chain anomaly detection or quality control where sensor data acquisition is costly. The paper highlights ProMUSE as a practical, uncertainty-aware, and resource-efficient solution for real-world AD screening.
The authors of the paper are Doan, Long, Chen, Branden, Litton, Ethan, Huang, Jiajing, Xie, Yixin, Zhou, Weihua, Narayanan, Nandakumar, and Zhao. The code and data are associated with the article, but no specific companies or products are mentioned beyond the university affiliations implied by the authorship. The research is licensed under a Creative Commons Attribution 4.0 International License.