Spontaneous speech feature analysis for alzheimer's disease screening using a random forest classifier.

Area of Science:

• Neurology
• Computational Linguistics
• Biomedical Engineering

Background:

• Alzheimer's disease (AD) diagnosis and monitoring are crucial for patient care.
• Current methods can be invasive, costly, or time-consuming.
• Speech analysis offers a potential non-invasive, cost-effective alternative.

Purpose of the Study:

• To predict early Alzheimer's disease (AD) diagnosis using speech data.
• To evaluate the stages of AD progression through acoustic feature analysis.
• To explore the utility of machine learning (ML) for AD detection via speech.

Main Methods:

• Collected spontaneous speech signals from individuals with AD and cognitively normal (CN) subjects.
• Performed exploratory analysis of acoustic features, non-stationarity, and non-linearity.
• Applied data augmentation techniques to enhance speech signal datasets.
• Utilized a Random Forest classifier for AD prediction and staging.

Main Results:

• Achieved an 82.2% accuracy rate for early AD prediction.
• Attained a 71.5% accuracy rate for classifying AD stages.
• Demonstrated the potential of acoustic and linguistic features in speech for AD detection.

Conclusions:

• Speech analysis combined with ML provides a promising avenue for non-invasive AD detection and monitoring.
• Further research into non-stationarity and non-linearity in audio features can refine AD classification models.
• This approach can significantly aid early diagnosis and repetitive monitoring of Alzheimer's disease progression.