Related Experiment Video
Updated: May 15, 2026
05:48
Memorization-Based Training and Testing Paradigm for Robust Vocal Identity Recognition in Expressive Speech Using Event-Related Potentials Analysis
Published on: August 9, 2024
Mandarin speech-based early detection of SCD: a feature-fusion residual network method.
Summary
A new speech-based model shows promise for quickly screening subjective cognitive decline (SCD), a key stage for early Alzheimer's disease intervention. This non-invasive method identifies acoustic differences, potentially aiding early diagnosis.
Area of Science:
- Neurology
- Speech analysis
- Biomarkers
Background:
- Alzheimer's disease (AD) presents a significant global health burden.
- Early intervention in subjective cognitive decline (SCD) is critical for potentially delaying AD progression.
- SCD represents a crucial, early stage for therapeutic strategies.
Purpose of the Study:
- To evaluate an exploratory speech-based model for the rapid screening of SCD.
- To assess the efficacy of acoustic analysis in differentiating SCD from other cognitive states.
Main Methods:
- The study involved 459 participants across four groups: AD, mild cognitive impairment (MCI), SCD, and normal controls.
- Mandarin speech data, specifically Pic-Talk clips, were analyzed using residual network features.
- A cross-sectional design was employed to assess the speech model's performance.
Main Results:
- The speech-based model demonstrated high performance in SCD screening.
- Key performance metrics included accuracy (81.77%), recall (80.53%), precision (82.27%), F1-score (81.39%), and AUC (82.85%).
- The model outperformed other existing speech analysis models for SCD detection.
Conclusions:
- The non-invasive, speech-based approach shows potential for SCD assessment.
- Acoustic differences were identified at a group level between individuals with SCD and other diagnostic categories.
- Future integration of biomarkers is expected to enhance model accuracy and broaden its clinical applicability.
