Speech Features for Alzheimer's Screening Across 2,320 Individuals

2,320 individuals across the cognitive spectrum provided speech samples for this study. The dataset includes extracted multidomain features from three spoken tasks to support machine learning models for cognitive classification and biomarker prediction. Josep Blazquez-Folch authored this research, which was last updated on May 4, 2026.

Use Cases

• Classifying cognitive stages (SCD, MCI, ADD) based on multidomain speech features.
• Predicting amyloid, tau, and neurodegeneration (ATN) biomarker positivity from speech patterns.
• Estimating scores across six neuropsychological domains using speech-based models.

Strengths

• Dataset includes speech from 2,320 individuals spanning subjective cognitive decline, mild cognitive impairment, and Alzheimer's disease dementia.
• Models achieved AUC values up to 0.94 for differentiating cognitive stages and up to 0.83 correlation with cognitive function scores.
• Features cover acoustic, lexical, syntactic, and semantic domains extracted from three spoken tasks.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is small, at 118.8 KB, suggesting it contains processed feature summaries rather than raw audio.

Provenance

Source

figshare

Collection Method

Speech analysis from three spoken tasks per participant.

Freshness

Last updated 2026-05-04 05:34:11; freshness should be verified.