Workshops

Wearable Sensing, Adaptive Robotics, and AI for Personalized Neurorehabilitation and Assistive Health Technology

Short title: SensAI-Rehab 2026

General description

Wearable sensing, rehabilitation robotics, and assistive health technologies are transforming personalized healthcare, yet significant challenges remain in translating these innovations from research laboratories to routine clinical practice and home environments. To address these challenges, the Marie Skłodowska-Curie Doctoral Networks TAILOR and AERIALIST are training a new generation of researchers developing intelligent, AI-enabled rehabilitation and assistive technologies. This workshop will bring together doctoral candidates from both networks to present their latest research and foster interdisciplinary discussion on wearable sensing, artificial intelligence, and rehabilitation robotics, in close alignment with the IEEE BSN 2026 theme: “From Lab to Life: Scaling Intelligent Sensors for Personalized Health Ecosystems.”

Agenda

Opening session: Welcome and introduction to TAILOR and AERIALIST

Speaker: Eduardo Rocon

Session 1: Data Management and Safe Frameworks for Rehabilitation and Assistive Robotics

Speakers: Mohammed Tloul – AERIALIST Doctoral Candidate; Juan Pablo González Sánchez – TAILOR Doctoral Candidate

Duration: 30 minutes

Ensuring the safety of rehabilitation and assistive technologies requires not only robust safety engineering methods but also effective data management practices throughout the product lifecycle. This presentation provides an overview of current approaches to system safety, highlighting the importance of integrating safety considerations from the earliest stages of technology development and illustrating how these principles are being applied within the AERIALIST project. It also addresses practical challenges encountered during safety analyses, particularly those related to managing heterogeneous data generated by multiple stakeholders. Building on these concepts, the presentation introduces key principles of data management and the data lifecycle in biomedical engineering, emphasizing their role in human-centered research. Through interactive audience participation, attendees will explore common data management challenges before being introduced to the TAILOR data management framework, which builds on existing solutions for rehabilitation robotics benchmarking and clinical data to support the development of safe, interoperable, and reproducible assistive technologies.

Session 2: Wearable Sensing for Intelligent Assistive Technologies: From Monitoring to Classification and Intervention

Speakers: Harry Morland – AERIALIST Doctoral Candidate; Matteo Caruso – AERIALIST Doctoral Candidate; Ricardo González Valiño – TAILOR Doctoral Candidate

Duration: 1 hour

Wearable sensing is a key enabler for advancing assistive and rehabilitation technologies from laboratory research to clinical and home environments. This session brings together complementary perspectives from the TAILOR and AERIALIST Doctoral Networks to demonstrate how wearable devices support diagnosis, continuous monitoring, and adaptive intervention. Topics include embedded edge AI for the early detection of movement disorders, multimodal sensing for objective assessment of gait and motor performance, and the integration of wearable sensing into closed-loop rehabilitation systems, such as functional electrical stimulation (FES). Together, these approaches illustrate how wearable sensing, combined with artificial intelligence, enables more personalized, responsive, and scalable assistive technologies for real-world healthcare.

Session 3: Smart Control for Powered Lower-Limb Devices for Rehabilitation and Assistance

Speakers: Johannes König – TAILOR Doctoral Candidate; Ramez Iskandar – TAILOR Doctoral Candidate; Riccardo Rocco – AERIALIST Doctoral Candidate

Duration: 45 minutes

Adaptive control is a fundamental requirement for the next generation of rehabilitation and assistive robotic systems. This session explores how AI-driven and model-based control strategies can personalize robotic assistance by adapting to the user’s motor capabilities, therapeutic needs, and changing physical state. Drawing on complementary research from the TAILOR and AERIALIST Doctoral Networks, the presentations cover adaptive control for lower-limb exoskeletons after stroke, “virtual therapist” strategies that adjust robotic support according to user contribution, and adaptive control of powered prostheses for safe real-world use. The session concludes with an interactive demonstration, allowing participants to compare manual parameter tuning with adaptive control strategies in a simulated exoskeleton-assisted gait rehabilitation scenario.

Session 4: Smart Interfaces and Shared Learning for Assistive Devices

Speakers: Max-Ole Bastian von Waldow – TAILOR Doctoral Candidate; Lucia Palumbi – AERIALIST Doctoral Candidate; Muhammad Shakeel Ishtiaq – AERIALIST Doctoral Candidate

Duration: 45 minutes

Artificial intelligence is becoming a key enabler of next-generation rehabilitation and assistive technologies, supporting the entire continuum of care from initial device setup to long-term personalized assistance. This session presents a patient-centered perspective on how AI can facilitate intuitive human–robot interaction, enable real-time monitoring and adaptive support in home-based rehabilitation, and leverage knowledge accumulated across users to continuously improve device performance. Drawing on complementary research from the TAILOR and AERIALIST Doctoral Networks, the session highlights how shared learning, wearable sensing, and intelligent control can create safer, more responsive, and scalable rehabilitation technologies while addressing the challenges of data management and trustworthy AI.

Closing cross-network panel: From body sensors to clinical impact – closing the loop

Speaker: Florence Thirlwell – TAILOR Doctoral Candidate

Duration: 30 minutes

Unobtrusive and Contactless Monitoring Technologies

General description

Recent advances in sensing technologies have enabled a wide range of unobtrusive and contactless approaches for monitoring vital signs in everyday and clinical environments. Such technologies offer the potential for continuous, long-term assessment of physiological parameters, thereby supporting a more comprehensive understanding of a patient’s current health status and its temporal dynamics. In addition, the fusion of complementary sensing modalities can increase robustness, provide redundancy, and reveal additional information about vital body functions that may remain hidden when using single measurement techniques alone. Despite their potential, many of these technologies are not yet widely adopted in clinical practice and still face challenges regarding validation, usability, reliability, standardization, and clinical acceptance. This workshop aims to provide an overview of emerging unobtrusive and contactless monitoring technologies, including capacitive ECG, mechanocardiography, magnetic induction measurement, camera-based photoplethysmography, and infrared thermography. Based on selected application examples from current research, the workshop will discuss technological principles, opportunities, limitations, and translational challenges toward practical use in healthcare.

Agenda

The workshop will combine concise theoretical introductions with hands-on demonstrations of selected unobtrusive and contactless monitoring technologies. For each sensing modality, we will first provide a brief introduction to the underlying measurement principle, technical implementation, and potential biomedical applications from current research topics (driver state monitoring, geriatric patients, neonatal intensive care unit). This will be followed by an interactive demonstration, allowing participants to experience and discuss the technology in practice.

The programme will cover the following measurement modalities:

Capacitive ECG, mechanocardiography, and magnetic induction measurement

These modalities will be demonstrated using a sensor cushion that enables unobtrusive monitoring of cardiac activity and respiration. The theoretical introduction will cover the principles of capacitive ECG acquisition, mechanical cardiac activity sensing, and magnetic induction-based respiratory monitoring.

Camera-based photoplethysmography

A dedicated camera setup will be used to demonstrate contactless pulse and vital sign monitoring based on optical signal extraction from video recordings.

Infrared thermography

An infrared camera setup will be presented to illustrate non-contact temperature and physiological monitoring, including potential applications in clinical and everyday environments.

The workshop will conclude with a session on sensor fusion, exploring how complementary measurements can improve robustness, increase redundancy, and reveal additional physiological information.

Speakers: Markus Lueken, Nico Blass, Jannik Pruessmann, Steffen Leonhardt

Programme

Introduction

Duration: 10 minutes

Theory: Capacitive ECG

Duration: 20 minutes

Theory: Mechanocardiography

Duration: 15 minutes

Theory: Magnetic induction measurement

Duration: 15 minutes

Hands-on: Seat-integrated unobtrusive sensing

Duration: 20 minutes

Theory: Camera-based photoplethysmography

Duration: 15 minutes

Hands-on: Camera-based photoplethysmography

Duration: 10 minutes

Theory: Infrared thermography

Duration: 15 minutes

Hands-on: Infrared thermography

Duration: 10 minutes

Theory and applications: Sensor fusion

Duration: 20 minutes

From Signals to Systems: Hands-On Robust AI for Respiratory Sound Analysis

Short title: AI4LUNGS 2026

General description

Despite rapid progress in Artificial Intelligence for healthcare, translating respiratory sound analysis and physiological signal processing from controlled laboratory environments to real-world clinical deployment remains a fundamental challenge. Models that perform well on curated datasets often degrade significantly under real clinical conditions due to cardiac interference in lung sounds, variability across wearable sensors, and limited representation of complex or noisy physiological states.

Although emerging approaches such as multimodal learning and robust signal processing show promise, a critical translational gap persists. Researchers and practitioners frequently lack practical strategies to adapt their models to unconstrained, heterogeneous, and clinically realistic data.

Aligned with the “From Lab to Life” theme of the BSN 2026, this workshop is motivated by the need to bridge this gap. It aims to bring together researchers working on respiratory and physiological signal analysis to identify real-world bottlenecks and discuss actionable solutions systematically. The focus will be on moving beyond proof-of-concept models toward methodologies that are robust, generalizable, and clinically viable.

In addition to fostering discussion around real deployment challenges, this workshop is designed to provide a hands-on, experiential component. Participants will engage directly with respiratory sound data through guided exercises, allowing them to develop an intuitive understanding of (i) the acoustic characteristics of lung sounds, (ii) the type of clinically relevant information that can be extracted from such signals, (iii) the practical challenges associated with noise, variability, and annotation, and (iv) the comparative strengths and limitations of different signal processing and modern AI techniques. By combining conceptual insights with interactive experimentation, the workshop aims to bridge not only the methodological gap but also the perceptual and practical gap between algorithm development and real-world clinical data.

This workshop will also showcase recent results on lung auscultation developed within the AI4LUNGS project, a European initiative focused on AI-driven respiratory health monitoring. These results will serve as concrete case studies to illustrate the challenges of deploying machine learning models on real-world respiratory data, including variability across acquisition settings, annotation constraints, and generalization across patient populations. By grounding the discussion in ongoing research efforts, the workshop will connect methodological advances with practical clinical applications.

Agenda

Session 1: Clinical and technical foundations of lung sound analysis

Speaker: Francesco Renna

Duration: 60 minutes

This keynote introduces the clinical and technical foundations of respiratory sound analysis, covering challenges and current solutions in signal acquisition, signal processing, and AI. It highlights recent advances in machine learning for respiratory signals, with particular attention to domain shift and generalizability, and prepares participants for the practical sessions.

Session 2: Hands-on development of robust AI pipelines

Speakers: Francesco Renna, Alireza Babei, Pedro Oliveira

Duration: 2 hours and 30 minutes

Using an interactive notebook and the publicly available ICBHI 2017 dataset, participants will develop a respiratory sound analysis pipeline from raw signal inspection to AI classification. Guided examples and hands-on exercises will cover:

– Data loading and curation

– Signal preprocessing and denoising

– Signal segmentation

– Robust respiratory signal representations

– Adventitious sound classification

Session 3: Interactive XAI platform for lung sound classification

Speakers: Francesco Renna, Alireza Babei, Pedro Oliveira

Duration: 30 minutes

Participants will explore a customized platform for interactive visualization of lung sound classification and explainable AI (XAI). They will load different AI models, examine their performance on the ICBHI dataset, and apply example-matching XAI algorithms to explore model predictions.