Biography/Background:
With the unprecedented development of IOT sensors that precisely collect various behavioral data from
users, new possibilities to investigate and uncover digital biomarkers of neurodegenerative diseases
based on AI and data science have emerged. Modalities like Voice, Face, and Handwriting can be
acquired from high quality low-cost sensors (e.g. microphones, cameras, tablets) which are ecological
and not obtrusive. Besides, they can be collected remotely, thus not requiring patient consultation at
the hospital each time. Developing such digital biomarkers, however, raises several challenges like the
inherent ambiguity of the modalities above, the scarcity of the training data, and the need of predictive
solutions than can be interpretable to the medical stakeholders. In this talk, we present our recent
research on assessing neurodegenerative diseases, namely Parkinson’s and Alzheimer’s, from noninvasive input, namely face, voice and on-tablet handwriting. In particular, we will show how our AI
techniques overcome the lack of training data through transfer learning and data augmentation, and
propose tools that make our models’ predictions understandable to physicians and other stakeholders
through sound explainable AI techniques. We will also show how the models’ predictions show good
correlations with traditional clinical and neuroimaging scores. Last but no least, we discuss safety, and
ethical and regulatory issues in the context of this topic.
Biography/Background:
Mounîm A. El-Yacoubi (Ph.D., University of Rennes, France, Sept. 1996) has been a professor in Artificial
Intelligence at Institut Polytechnique de Paris / Telecom SudParis since 2008. He was a researcher at
French Post Research Center (SRTP), from 1992-1996, a Visiting Scientist at the Centre for Pattern
Recognition and Machine Intelligence (CENPARMI, Concordia University), Montreal, in 1997-1998, and
then an Associate Professor at the Catholic University of Parana. From 2001 to 2008, he was a Senior
Researcher at Parascript, Boulder, Colorado. At SRTP and Parascript, he developed handwriting
recognition software for real-life automatic mail sorting, bank check reading, and form processing, still
running daily in automatic reading machines worldwide. He is Senior IEEE member, and Associate Editor
of IEEE Access, International Journal of Network Dynamics and Intelligence, and Journal of CyberPhysical-Social Intelligence. He has been Program Co-Chair of the International Conference on Pattern
Recognition and Artificial Intelligence (Paris, France, 2022), Program Co-Chair of IEEE Smart World
Congress (Portsmouth, UK, 2023) and International Conference on Cyber-physical Social Intelligence
(Xi'an, China, 2023). He served as General Chair of the International Conference on Human System
Interaction, Paris, France, 2024, and as Program Co-Chair of the International Conference on Computing
and Pattern Recognition (Tianjin, China, 2024). He edited two books: “Pattern Recognition and Artificial
Intelligence,” Springer, 2022, and “Emerging Topics in Pattern Recognition and Artificial Intelligence,”
World Scientific, 2023, and was the guest editor of Special Section: “Best papers of the international
conference on pattern recognition and artificial intelligence (ICPRAI) 2022” for “Pattern Recognition
Letters” in 2024.
Research Interest:
M.A. El-Yacoubi’sresearch interests include AI, machine learning & deep learning, modeling human user
data, especially behavioral signals like handwriting, voice, gesture and activity recognition. He has been
working recently in e-health applications, in particular detecting neurodegenerative and chronic
diseases, namely Alzheimer’s and Parkinson’s, and Diabetes . His other research includes indexing
historical documents, biometrics, sports analytics, smart agriculture, and human mobility for smart
cities.
