Data Science Week 2025

Lucas Noldus
Monday, 12:45 - 13:30. 
Location: Citadel T300.

Lucas Noldus is Professor of Behavior, Information Technology and Innovation at Donders Institute for Brain, Cognition and Behavior,  Radboud University and CEO of Noldus Information Technology. His research is aimed at the discovery and development of new techniques for automatic behavioral recording in animals and humans. The topics are quite diverse and include generic AI models for behavior recognition in rodents and human infants, vocalizations in mice, EEG and behavior in mice, learning tasks in zebrafish, eye tracking in MS patients, and motion analysis in visually impaired people. He tries to build bridges between the  University and companies in the field of behavior and technology.

Abstract

The desire of humans to observe and describe the behavior of other organisms – their behavioral phenotype – is thousands years old, as we know from the writings of Aristotle (4^th century BC). More recently, halfway the 20^th century, Nobel laureates Tinbergen, Lorenz and Von Frisch taught us the importance of systematic observation and registration as a way to understand the mechanism and development of behavior, and founded the scientific discipline of ethology. Since then, generations of behavioral biologists and psychologists have collected behavioral data through observation and manual annotation. With the advent digital image processing in the 1980s, it became possible to automate this labor-intensive and error-prone process, and digital phenotyping was born. The earliest applications of this novel technique were limited to movement tracking of small animals in controlled laboratory assays, such as insects and rodents.

Since then, developments have accelerated: recording with two cameras allowed the 3D tracking of flying insects or swimming fish, ultrawideband technology enabled accurate tracking of animals or humans in large spaces, and inertial sensing with 3D accelerometers and gyroscopes opened the door to posture estimation and behavior recognition. A recent example is the design of a Smart Baby Suit for monitoring the neurodevelopment of babies at risk of a genetic disorder. With increasingly powerful CPUs, GPUs and AI models, we can perform pattern recognition on just about any image stream, audio signal or motion data. These developments have brought countless opportunities to advance biomedical research, healthcare and affective computing, which has led to a thriving business of digital phenotyping tools.

However, these technologies can theoretically also be used for purposes that could harm people, which is why the European Commission has enacted the AI Act, which prohibits AI-based certain types of digital phenotyping. As a research community, we should keep a dialog with policymakers to make sure that society can reap the benefits of promising technologies while risks are mitigated.

Matthijs Noordzij
Tuesday, 12:45 - 13:30.
Location: Citadel T300.

Matthijs Noordzij is Full Professor of Health Psychology and Persuasive Technology and directs the Health Dynamics & Self Management Lab at the University of Twente in The Netherlands.

His research and education focuses on exploring the scientific foundations and design principles for integrating sensor technology in (mental) healthcare and self-management. His vision is to develop health technology that aligns with core human values such as compassion, while striving to create innovative solutions that enhance the way we interact with technology in healthcare settings. 

Advances in wearable technology have made it possible to continuously track physiological signals such as heart rate variability, skin conductance, and respiration in everyday life. This form of passive sensing provides valuable insights into stress, emotion regulation, and mental health, moving beyond the limitations of laboratory studies.

In this talk, Matthijs will introduce the principles of ambulatory physiological monitoring, discuss the validity of current measurement techniques, and explore their potential applications in research and clinical practice. While these methods offer exciting opportunities, challenges remain in ensuring data quality, interpreting physiological signals in context, and addressing ethical concerns. By critically examining both the promise and limitations of these tools, we can better understand how to integrate them into studies and applications in a meaningful and responsible way.

Arlene John and Ying Wang
Wednesday, 12:45 - 13:30.
Location: Citadel T300.

Arlene John and Ying Wang are both Assistant Professor at the Biomedical Signals and Systems (BSS) group. Dr. John's research interests include biomedical signal processing, machine learning and inference, explainable AI, and multisensor data fusion, while dr. Wang's main research interest is about remote continuous monitoring of physiological signs and body movement in daily life for personalized disease prevention and management.

This talk explores the journey from physiological time-series data to multimodal data analysis for digital phenotyping, emphasizing the transition from controlled semi-lab environments to real-life health monitoring. The challenges and some innovations in daily-life health monitoring required to sense information unobtrusively to enable the development of personalized phenotypes for continuous health tracking is discussed. 

Key topics include both wireless and wearable sensing techniques, multimodal feature extraction, identifying interrelationships amongst features, and connecting these insights to individual phenotypes. Additionally, we examine methods for monitoring health trends over extended periods. Practical applications discussed will include energy expenditure monitoring during daily physical activity for people with risk of obesity, cardiac function monitoring for people with long term diabetes, psychophysiological condition monitoring for people with knee osteoarthritis, recovery tracking post-colorectal surgery using patch sensors, and smartphone-based digital phenotyping for breast cancer survivors. 

Annemieke Witteveen and Jorge Piano Simoes
Thursday, 12:45 - 13:30.
Location: Citadel T300.

Annemieke Witteveen is Associate Professor at the Biomedical Signals and Systems (BSS) group and the Personalized eHealth Technology (PeHT) research program. Jorge Piano Simões is an Assistant Professor in the Psychology, Health, and Technology (PHT) section. They both do research on digital phenotyping to create adaptive interventions and prediction, monitoring, and optimization to support personalized clinical treatment strategies.

In this workshop, participants will be introduced to key concepts related to digital phenotyping, including definitions, existing frameworks linking behavior, emotions, and physical states and digital traces, as well as clinical opportunities and barriers for implementing this tool. During the workshop, participants will also have the opportunity to critically evaluate how digital traces can be translated into clinical applications with a hands-on assignment. Lastly, participants will be presented to the state-of-the-art overview of how machine learning methods are being leveraged with digital phenotyping to improve and personalize care.

Peter Slijkhuis
Monday, March 10th, 12:45 - 13:30.
Location: Citadel T300.

Peter Slijkhuis is a psychologist who focuses on human behavior and technology, with a special interest in how people use and interact with technology (Usability Testing and UX Research). He is the educational coordinator at BMS-Lab, responsible for creating course materials explaining and using the equipment that BMS-Lab provides.

He will share his extensive experience with eye-tracking, virtual and augmented reality, facial and body motion tracking, wearables and more in an overview of equipment available at BMS-Lab, and how it can be useful for social science. [Note: exact contents may still change].

Tuesday, 13:45 - 15:30. Cleaning and combining data from different sources. Exploring and visualizing data and patterns.

Wednesday, 13:45 - 15:30. Going from raw signal data to useful metrics.

Thursday, 13:45 - 15:30. Introduction to modelling and machine learning.

Friday, 13:45 - 15:30. More in-depth continuation of modelling, implications of different modeling choices and 'fairness' in AI models. [Note: exact contents may still change]

Tuesday, 13:45 - 15:30. Cleaning and combining data from different sources. Exploring and visualizing data and patterns.

Wednesday, 13:45 - 15:30. Going from raw signal data to useful metrics.

Thursday, 13:45 - 15:30. Introduction to modelling and machine learning.

Friday, 13:45 - 15:30. More in-depth continuation of modelling, implications of different modeling choices and 'fairness' in AI models. [Note: exact contents may still change]