Topic-Based Measures of Conversation for Detecting Mild Cognitive Impairment

Area of Science:

• Computational linguistics
• Neuroscience
• Gerontology

Background:

• Conversation is a complex cognitive process involving memory, language, and emotion recognition.
• Cognitive impairments, such as mild cognitive impairment (MCI), can affect these conversational abilities.
• Existing methods may not fully capture subtle conversational deficits in early-stage cognitive decline.

Purpose of the Study:

• To propose a novel computational method for quantifying topical variations in conversations.
• To develop new conversational measures indicative of cognitive deficits in older adults.
• To assess the utility of these measures in distinguishing individuals with MCI from healthy controls.

Main Methods:

• Developed a computational approach analyzing lexical coherence between consecutive utterances in semi-structured conversations.
• Extracted lexical knowledge from conversational data.
• Generated novel conversational measures to quantify topical shifts.

Main Results:

• The proposed method generates unique conversational measures.
• These measures effectively indicate underlying cognitive deficits.
• Preliminary results show the method can distinguish individuals with MCI from healthy controls.

Conclusions:

• Lexical coherence analysis offers a promising avenue for detecting cognitive impairments through conversation.
• The developed conversational measures show potential for early identification of mild cognitive impairment (MCI).
• This approach could aid in the non-invasive assessment of cognitive health in older adults.