Dial-a-doctor Print this article

Researchers at the Technology Research for Independent Living (TRIL) Centre, based in Dublin, are exploring the development of an automated system to analyse voice recordings in an effort to assess cognitive function and enable early identification of cognitive decline, in a project called ‘Dear Diary’

Through the Dear Diary project, TRIL Centre researchers are attempting to identify characteristics of older people’s speech that predict their level of cognitive functioning—speech biomarkers that could signal the onset of problems. “Speech—in particular, reading and describing things—provides a window into cognitive function,” says Professor Richard Reilly, Professor of Neural Engineering in Trinity College Dublin (TCD) and co-leader of the project. “It targets several areas of the brain, which control the visual system; the articulators that allow us to form words; and semantic ability, which is the ability to understand meaning and relationships. By measuring speech characteristics, we’re tapping into a wide range of brain functions.”

Cognitive decline can occur gradually, making it difficult to detect. Typically, problems are diagnosed during visits to a doctor’s office—in many cases, only after the decline becomes serious enough to jeopardise the patient’s independence.

A remote monitoring system that leverages telephone technology would be a less costly and more convenient alternative that could detect problems earlier, when interventions can be more effective. “We want to intercept declines in cognitive function before they become manifested in larger problems,” says Professor Ian Robertson, Professor of Psychology in TCD, who co-leads the research effort along with Reilly. “Our goal is to enable mass cognitive screening and assessment in a way that’s not feasible with current technologies, and in a manner that’s acceptable to older people.”

Studies have demonstrated that speech can be used to diagnose mild cognitive impairment and neurodegenerative diseases such as depression and schizophrenia. The TRIL Centre researchers believe that speech could be leveraged more broadly to detect cognitive decline related to ageing.

Cognitive testing in the TRIL clinic
As part of the TRIL Centre initiative, 600 older adults were given comprehensive health exams at the TRIL clinic in St James’s Hospital in Dublin. Half of the group also completed a battery of cognitive assessments, including the Mini Mental State Exam (MMSE)—a gold standard test of overall cognitive function, which was used to establish a baseline measurement for each of the volunteers.

A subset of this group agreed to participate in several experiments for the Dear Diary project. Each of these mini-studies involved at least 80 older adults. In addition to undergoing standard tests of cognitive function, Dear Diary volunteers were recorded as they performed a series of speech-related tasks. The researchers developed algorithms that extract specific features of speech from the recordings made in the clinic, then evaluated the results to determine if certain speech characteristics were correlated with performance on standard cognitive tests. Pauses (both the duration and placement of pauses) proved to be the most promising biomarker of cognitive decline, able to differentiate between older adults with lower and higher levels of cognitive functioning, as measured by their MMSE scores.

These results suggest that it’s possible to assess cognitive functioning and detect signs of decline by analysing recorded speech. The next step for the Dear Diary team was to determine if features of speech which correlate with cognitive functioning could be extracted reliably from telephone recordings. To do this, the team recruited several older adults who had volunteered for Dear Diary studies in the TRIL clinic to participate in telephone testing in their homes. For the home studies, roughly 60 volunteers went through a one-time cognitive assessment, which validated that the speech biomarkers identified in the clinic also could be extracted from a quality telephone recording. Another 30 volunteers participated in monthly telephone testing for nine months.

Automating the tests
With the viability of telephone testing confirmed, the researchers set out to develop an interactive voice response (IVR) system that would eliminate the need for a trained psychologist to participate in the assessment. A fully automated system is essential to making routine home monitoring of cognitive functioning scalable and affordable. Using feedback from the focus group, engineers built a prototype IVR system. A few older adults volunteered to test the system, with positive results.

Piloting the IVR system
In the latter half of 2010, researchers conducted a larger pilot programme to validate the IVR system with 80 volunteers.
Each volunteer had a face-to-face cognitive assessment in the TRIL Clinic and two telephone assessments using the IVR system. Half of the group completed the face-to-interviews first, and the other half underwent telephone assessments initially. Researchers are currently reviewing the results, to assess the stability of speech characteristics between the face-to-face and automated interviews, determine if the order of testing (phone vs. in-clinic interviews) makes a difference, and to weed out tests that prove to be unstable or unreliable. This study is a test bed for deploying a fully automated cognitive assessment system.

This article is an abridged version of a longer article, Dear Diary: analysing speech to detect cognitive decline.
The full article can be read on the TRIL website, www.trilcentre.org