Enhancing the Classification of Aphasia: A Statistical Analysis Using Connected Speech.

Area of Science:

• Computational linguistics
• Neuroscience
• Speech-language pathology

Background:

• Large databases and automated language analysis offer novel methods for studying connected speech in people with aphasia (PWA).
• These techniques can provide new insights into the linguistic characteristics of PWA.

Purpose of the Study:

• To group people with aphasia (PWA) into coherent clusters based on their language output using unsupervised statistical methods.
• To identify specific linguistic features most strongly correlated with these identified clusters.

Main Methods:

• Applied K-means clustering and random forests classification algorithms to language production data from 168 PWA.
• Utilized language samples from a standard discourse protocol covering four genres: free speech, personal narratives, picture descriptions, Cinderella storytelling, and procedural discourse.

Main Results:

• Identified seven distinct clusters of PWA using the K-means algorithm.
• Developed and validated a random forests classification tree with 91% agreement with cluster assignments.
• The most discriminative features were total words from free speech tasks and total closed-class words from the Cinderella storytelling task.

Conclusions:

• Connected speech data effectively categorizes PWA into distinct, coherent groups.
• Findings offer insights into traditional aphasia classifications and can inform future discourse research and clinical practice.