To develop and evaluate SpeechCARE, a multimodal speech-processing pipeline for detecting cognitive impairment, specifically focusing on Alzheimer's Disease and Related Dementias (ADRD) and Mild Cognitive Impairment (MCI), across diverse populations and languages.
Key Findings:
Achieved an average F1-score of 72.11% on the held-out test set.
Demonstrated strong multilingual generalizability despite moderate disparities for Spanish speakers.
Incorporating age as a demographic factor significantly improved predictive accuracy.
Earned a special recognition award from NIA for its innovative approach.
Interpretation:
SpeechCARE shows promise as an effective tool for early detection of cognitive impairment, complementing existing biomarkers by capturing functional speech deficits, particularly in the context of ADRD and MCI.
Limitations:
Moderate disparities in performance for Spanish speakers.
Potential challenges in generalizing findings across all demographic groups, particularly in terms of cultural and linguistic diversity.
Conclusion:
SpeechCARE represents a significant advancement in cognitive impairment detection, offering a scalable solution for early identification across diverse populations.
