Measuring speech and language characteristics of effects of medications on cognition.

Abstract

Many of the drugs prescribed for various diseases and syndromes, including chronic pain and epilepsy, significantly impair cognitive functioning with large individual variation. Language is a highly individualized and directly observable product of human cognition that is central to our everyday functioning. A system for automated language and speech analysis (SALSA) was developed that relies on using an automatic speech recognition engine (HTK 3.4) [Young et al. (2008)] to perform forced-alignment between the audio of spontaneous speech samples and their transcripts to measure a number of speech and language characteristics including fluency, speaking rate, change in fundametal frequency, and information content of spontaneous narratives. The system was piloted on a population of 14 healthy volunteers who participated in a randomized, placebo-controlled study of cognitive effects of an anti-epileptic medication (topiramate). Our preliminary results suggest that SALSA captures a number of fluency and lexical characteristics sensitive to the effects of topiramate, thus providing an objective mechanism to quantify the degree of cognitive impairment in individulas affected by medications. Our current results are consistent with prior work investigating speech and language correlates of mild cognitive impairment [Roark et al. (2007)] and fronto-temporal dementia [Pakhomov et al. (2009)]. [Work supported by a grant from the Univ. of Minnesota Academic Health Crt.]