Dual-Modal Model Detects Mild Cognitive Impairment
Researchers at National Taiwan University publish a 2026 JMIR Med Inform paper introducing a dual-modal longitudinal system that uses autobiographical memory speech and text to detect mild cognitive impairment (MCI). The model incorporates an aging trajectory module to align local and global temporal features across visits. Experiments report AUROC of 0.85 and 0.89 on two Chinese datasets and validation accuracy above 0.88 on ADReSSo.
Key Points
- 1Introduces aging trajectory module aligning multimodal speech-text features across visits
- 2Demonstrates improved longitudinal detection with AUROC 0.85 and 0.89 on two datasets
- 3Enables noninvasive, scalable MCI monitoring using autobiographical speech for long-term clinical screening
Scoring Rationale
Strong peer-reviewed longitudinal multimodal results driving practical MCI screening, limited by dataset scope and clinical validation.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
