WATCH-SS: Developing a Trustworthy and Explainable Modular Framework for Detecting Cognitive Impairment from...

Area of Science:

• Neurology
• Computational Linguistics
• Artificial Intelligence

Background:

• Early detection of cognitive impairment (CI) is vital for managing Alzheimer's disease and related dementias.
• Current screening methods may lack accessibility or efficiency for widespread primary care use.

Purpose of the Study:

• To introduce the Warning Assessment and Alerting Tool for Cognitive Health from Spontaneous Speech (WATCH-SS).
• To develop a modular, explainable framework for detecting CI using speech analysis.

Main Methods:

• The WATCH-SS framework employs a three-stage process with detectors for five linguistic and acoustic CI indicators.
• Outputs are aggregated into clinically interpretable features for a predictive CI classification model.
• Both efficient real-time and high-accuracy offline analysis methods were explored.

Main Results:

• The WATCH-SS framework demonstrated strong predictive performance on the DementiaBank ADReSS dataset.
• Achieved an Area Under the Curve (AUC) of 80% on the test set.
• The modular, feature-based approach provided a transparent diagnostic profile.

Conclusions:

• The WATCH-SS tool shows significant potential as a trustworthy and clinically usable screening tool for primary care.
• This approach offers a transparent diagnostic profile, enhancing clinical trust.
• Further development could facilitate earlier intervention for cognitive impairment.