Early Dementia Detection Using Multiple Spontaneous Speech Prompts: The PROCESS Challenge

TL;DR

The paper introduces the PROCESS Challenge to promote early-stage dementia detection using spontaneous speech. It provides a new Cognospeak corpus with three neurologist-designed prompts—Semantic Fluency, Phonemic Fluency, and Cookie Theft—used to collect speech for two tasks: classification of cognitive decline/dementia status and regression to MMSE scores. Baselines combine acoustic features (OpenSmile eGeMAPS) with traditional classifiers and regressors, and a Whisper ASR + RoBERTa pipeline for transcripts; results show best F1 of 55.0% for the classification task and RMSE around 2.98–3.31 for regression, with text baselines lagging behind the acoustic ones. The findings highlight the potential of speech-based signals for early detection while underscoring the need for improved models, positioning the dataset as a benchmark to accelerate progress in this domain.

Abstract

Dementia is associated with various cognitive impairments and typically manifests only after significant progression, making intervention at this stage often ineffective. To address this issue, the Prediction and Recognition of Cognitive Decline through Spontaneous Speech (PROCESS) Signal Processing Grand Challenge invites participants to focus on early-stage dementia detection. We provide a new spontaneous speech corpus for this challenge. This corpus includes answers from three prompts designed by neurologists to better capture the cognition of speakers. Our baseline models achieved an F1-score of 55.0% on the classification task and an RMSE of 2.98 on the regression task.