Call for participation

The Prediction and Recognition Of Cognitive declinE through Spontaneous Speech (PROCESS) Signal Processing Grand Challenge, sets out signal processing and prediction tasks to detect dementia via speech. This challenge aims to bring together researchers from academia and industry to collaborate on innovative solutions. Participants will have the opportunity to contribute to the area of dementia detection, network with like-minded researchers, and gain recognition for their contributions to this important cause.

This time, we aim to detect not only cognitive decline that has already developed into diseases, e.g., Alzheimer’s but also to identify earlier stages, such as mild cognitive impairment (MCI), which is crucial for timely clinical intervention. The PROCESS challenge follows that of ADReSS [1], ADReSSo [2] and ADReSS-M [3]. Unlike these previous challenges, PROCESS will be based on data, which is not released yet to the public. In addition, in this corpus, we have provided two additional prompts, alongside the commonly used “Cookie Theft” picture description prompt. These two prompts are designed in line with clinical diagnosis to test the semantic and phonetic fluency of speakers (e.g., “Name as many animals as you can in a minute”).

We invite participants to take part in two tasks in this challenge: the classification task, which aims to predict the speaker’s diagnosis (healthy, MCI, dementia), and the regression task, which aims to predict a test score that reflects cognitive decline. The top five teams in the ranking will be invited to detail their approach in a two-page paper and present their results at ICASSP 2025. Accepted papers will be included in the proceedings of ICASSP. Additionally, teams that present their work will be invited to submit an extended paper to the IEEE Open Journal of Signal Processing (OJ-SP). We will encourage all participants, regardless of their final rankings, to make their papers available on a pre-print platform like arXiv or medRxiv and to share their code through a publicly accessible repository.

This challenge aims to create a platform for contributions and discussions on early-stage dementia detection using speech signal processing and Artificial Intelligence (AI) models. We invite submissions of papers presenting novel AI models, novel speech signal processing techniques, and novel feature selection and extraction approaches for the PROCESS. This is crucial for advancing early-stage dementia detection and the development of applications used in clinical dementia diagnosis.

If you are interested in taking part in the challenge, please fill out the online registration form. There are no limitations on the number of members per team, but each member can only join one team.

References

[1] Saturnino Luz, Fasih Haider, Sofia de la Fuente Garcia, Davida Fromm, and Brian MacWhinney, “Alzheimer’s dementia recognition through spontaneous speech,” 2021.

[2] Saturnino Luz, Fasih Haider, Sofia de la Fuente, Davida Fromm, and Brian MacWhinney, “Detecting cognitive decline using speech only: The ADReSSo challenge,” arXiv preprint arXiv:2104.09356, 2021.

[3] Saturnino Luz, Fasih Haider, Davida Fromm, Ioulietta Lazarou, Ioannis Kompatsiaris, and Brian MacWhinney, “An overview of the ADReSS-M signal processing grand challenge on multilingual Alzheimer’s dementia recognition through spontaneous speech,” IEEE Open Journal of Signal Processing, 2024.