Digital Biomarkers

Speech and Language

Background

One of the hallmarks of cognitive impairment is deterioration of speech. To develop and validate tools that can score each conversation we have, a reference dataset is required. Emphasis is given to free speech generation, such as in a phone conversation.

Approach

Circadic is helping ADDF design and run a large (N=3000) multicenter longitudinal (3-year) data collection study with academic and research centers around the world. The objective is to build systems that can operate on a phone, computer, or digital assistant and score conversations longitudinally to early detect signs of cognitive decline.

Under the hood

Circadic has co-designed numerous speech-eliciting tests that are delivered quarterly to a patient’s home through a tablet. Speech audio clips from each participant are manually spliced to remove personal identifying information, and clips are then harmonized together with clinical information obtained for each participant. The end result is a complete training (speech samples) and ground truth (clinical evaluation) stored at ADDI servers, which can be readily used by researchers to create algorithms that can assess the existence of cognitive decline. ADDF presents their vision of a consortium to accelerate research into speech and language biomarkers for Alzheimer’s Disease.