Pre-Conference Workshops

This workshop brings together psychiatrists, clinical researchers, and machine learning experts to explore practical and translational applications of artificial intelligence for psychiatric assessment, monitoring, and intervention. The workshop emphasizes clinically grounded AI approaches that are designed to integrate seamlessly into real-world psychiatric practice. Through a series of focused presentations, speakers will demonstrate how personalized, agentic chatbots can be combined with wearable and mobile sensor data to support patient engagement and longitudinal assessment; how multimodal and translational AI methods can be developed to assess multiple mental and physical health outcomes; and how to design end-to-end AI workflows that address the unique challenges of psychiatric research, including heterogeneous data, missing modalities, and ethical deployment. A central goal of the workshop is to foster meaningful dialogue between clinicians and AI researchers, enabling shared understanding of clinical needs, modeling constraints, and validation standards that accelerate translation from research to practice. The workshop concludes with a hands-on session in which participants follow an interactive notebook demonstrating a real-world AI application in psychiatry, providing practical exposure to data processing, model development, and interpretation. This interactive component is intended to demystify AI tools and empower participants to apply these methods within their own clinical or research contexts.

Workshop presenters:

• Ervin Sejdic
• Amir Rahmani
• Matthew Flathers, Andrew Byun
• Mai Ali
• Christopher Lucasius

The increasing accessibility of mobile electroencephalography (EEG), together with wearable biometric devices (e.g. electrocardiography (ECG), electromyography (EMG), and respiration), have made repeated at-home recordings feasible in generating large, longitudinal, and multi-modal data. Such large-scale data is a key enabler for personalized “digital twins” to address the inter-individual variabilities in neural and physiological dynamics to improve health monitoring and disease research. Following this paradigm, mechanistic brain modeling (mean-field, neural mass, biophysical, and whole-brain models) is an emerging tool for EEG analysis: Latent model parameters can serve as subject-specific descriptors of neural dynamics or be analyzed in tandem with additional biometrics through machine learning models. Furthermore, simulations of brain models enable systematic testing of perturbations to explore personalized interventions.

This workshop presents recent theory and applications of mean-field and neural mass models, alongside machine learning approaches for fitting these models to EEG data, as well as novel personalized machine learning methods to leverage wearables in predicting clinical outcomes. It also includes hands-on tutorials on collecting and interpreting mobile EEG in combination with multimodal biometrics (ECG, EMG, and breathing) and using the collected data during the workshop to generate and simulate whole-brain models.

Workshop presenters:

• John David Griffiths
• Ervin Sejdic
• Heng Kang Yao
• Minarose Ismail

Data-driven deep learning is now widely used in medical imaging, yet purely data-driven models can become unreliable when measurements are noisy, acquisition settings differ from training (domain shift), or training data are dominated by simulations rather than diverse clinical examples. Physics-inspired and physics-informed learning addresses these challenges by embedding imaging physics and known signal structure into model design and training, improving robustness, reducing reliance on broad training-data diversity, and often enhancing interpretability.
This workshop introduces key concepts and practical, hands-on techniques for integrating physics into deep models for medical imaging. We will cover a broad spectrum of approaches, including known-operator and model-based layers, fusion of physics-derived features, deep supervision using intermediate physical quantities, physics-based regularization, and physics-informed loss formulations. Participants will learn when to use each strategy, how to implement them in modern architectures, and how these choices impact stability and generalization under real-world variability. Attendees will leave with actionable design patterns and implementation guidance for building imaging pipelines that transfer reliably across domains.

Workshop presenters:

• Ali Kafaei Zad Tehrani
• Peng Guo
• Ion Candel

This workshop aims to provide participants with opportunities for hands-on physiological biosignal acquisition and interpretation with specific focus on Electromyography (EMG)/Photoplethysmography (PPG), Electrocardiography (ECG), and Electroencephalography (EEG). These experiences were designed for translation of physiological concepts to practical data, which is challenging to teach, allowsing student understanding of signal variation over time (accessing biosignals to infer physiological response).

Participants will engage in interactive sessions that combine theory with hands-on data collection using portable biosignal acquisition systems. The workshop will begin with a brief overview of the learning outcomes: i) to understand the principles behind ECG/PPG, EEG, and EMG; ii) to gain appreciation of sensor placement, signal recording techniques, and data acquisition; iiii) to perform basic signal processing and biosignal interpretation using open-source tools; and iv) translate relevant biosignals to cardiac/respiratory, neural, and skeletal muscle physiology.

Activities include:

• ECG/PPG: Capture heart rhythms, identify key waveform components, capture physiologic responses using easy to implement interventions and discuss clinical relevance.
• EEG: Measure brainwave patterns during rest and simple cognitive tasks, exploring artifact removal and basic interpretation.
• EMG: Record muscle activity during peripheral stimulation of involuntary skeletal muscle contractions to analyze signal amplitude, frequency, and latency.

Workshop presenters:

• Chris Bouwmeester
• Lindsey Fiddes
• Dawn Kilkenny

In silico medicine utilizes advanced computational models to emulate human anatomy and disease processes, enabling safer and faster development of medical devices and drugs. Alongside, machine intelligence technologies like deep learning and generative AI significantly improve the analysis of complex biomedical signals, providing accurate, real-time, and clinically relevant insights. Together, these innovations foster a shift toward integrated, data-driven biomedical research and healthcare.
Through a combination of hands-on exercises and guided tutorials, attendees will build and refine a high-fidelity cardiac electrophysiology model, simulate pathological conditions such as regional ischemia, and apply deep learning approaches (e.g., LSTMs, CNNs) for automated detection of cardiac anomalies. Attendees will learn about Verification and Validation (V&V) of AI models to ensure the models behave reliably, safely, and transparently.

Highlights:

• Build and refine high-fidelity simulation models of human physiological systems, with a specific focus on cardiac electrophysiology.
• Learn to parameterize models for simulating a variety of diseased states – e.g., generate electrocardiogram (ECG) and cardiac electrogram (EGM) by simulating regional ischemia.
• Classification of ECG signals using time-frequency-based AI models and GenAI approaches. Verification of AI models for reliable, safe, and robust AI models.
• Learn FDA certification guidelines pertinent to In Silico models and AI-enabled medical devices.

Workshop presenter:

• Reza Fazel-Rezai
• Ahmed Mekky
• Garima Sharma

Brain–computer interfaces (BCIs) and related neurotechnologies are rapidly advancing, offering new ways to measure, analyze, and influence brain activity. Despite growing interest, the diversity of available tools and methods can make it challenging for researchers and clinicians to understand their capabilities, applications, and limitations.

This workshop provides a comprehensive overview of modern brain–computer interface (BCI) and neurotechnology tools, spanning signal acquisition, biomarker analysis, and clinical application. The session will introduce the BCI technology landscape, including low- and high-density EEG systems, invasive and non-invasive neural recordings, and multimodal approaches.
By the end of the workshop, attendees will gain a practical understanding of the current BCI landscape, key applications, and emerging directions, equipping them to explore new research opportunities or clinical implementations.

Workshop presenters:

• Micah Ching, BSc.
• Dean J. Krusienski, Ph.D.
• Nuri F. Ince, Ph.D.
• Kei Masani, Ph.D.
• Christoph Guger

Workshop Agenda:

• Talk 1 – Micah Ching: The Brain-Computer Interface Toolbox: An Overview of Available Technologies (75min)
• Talk 2 – Dean Krusienski: Analysis of local field potentials for neurological disorders (30min)
• Talk 3 – Nuri F. Ince: Tracking High Frequency Neurobiomarkers for the Improvement of Neurosurgical Therapy in Epilepsy (30min)
• Talk 4 – Kei Masani: EEG-related approaches in the context of neurorehabilitation and balance control: From application to clinical perspective (30min)
• Talk 5 – When Technology Becomes Therapy: Brain-Computer Interfaces in Medicine (75min)

The adoption of wearable devices is profoundly impacting on numerous healthcare sectors, ranging from rehabilitation and assistive technology to vital-sign monitoring and point-of-care diagnostics. Commercial products have been demonstrated to be capable of collecting biopotentials and physiological markers, with medical-grade signal quality. However, they are typically closed stacks, which limits the possibility of accessing and modifying the proprietary hardware and software. Conversely, the open-source ecosystem has reached a sufficient level of maturity to encompass the entire pipeline, from sensing hardware to edge intelligence. However, the landscape remains somewhat fragmented and navigating it without concrete examples and reusable workflows can be challenging.

This workshop highlights how open-source solutions can support the full end-to-end workflow of a healthcare wearable device design. Starting from customizable acquisition platforms for biopotentials (ExG) and wearable ultrasound in various form-factors, we then move to standardized software workflows for myoelectric control, showing how open analysis and benchmarking practices reduce engineering overhead and accelerate iteration. To broaden the perspective on advanced edge deployments, we include a neuromorphic open-source EMG viewpoint focused on ultra-low-power, low-latency inference. Finally, we introduce a device-agnostic integration layer that unifies heterogeneous devices through modular Python connectors for real-time streaming, visualization, and rapid experimentation.

Workshop presenters:

• Simone Benatti
• Tobias Röddiger
• Andrea Cossettini
• Erik Scheme
• Elisa Donati
• Mattia Orlandi

The development of image-guided medical robotics is transforming the landscape of healthcare today. These tools offer several clinical advantages, such as dexterous manipulation of surgical instruments, surgical task automation, accurate digital patient models, and intraoperative tissue characterization. However, creating such systems requires the integration of AI, robots, imaging devices, and tracking equipment. This integration is a challenging barrier to entry for researchers to overcome before they can even begin innovating. Furthermore, many medical robotic systems rely on custom APIs or expensive research licenses that make them inaccessible.

Recently, several open-source platforms have gained momentum to facilitate the integration of various software and hardware components into a single system. These platforms enable modular and component-based development for rapid prototyping, reducing the initial startup burden. These tools can also incorporate dynamic robot simulation, allowing for a seamless transition between the virtual and physical environments for testing, which is crucial for the development of intelligent and AI-driven medical robotics.

This workshop will introduce participants to the landscape of image-guided robotics while providing a practical hands-on tutorial and demonstration. Our aim is for participants to leave with an understanding of how to begin prototyping medical robotic systems tailored to their own clinical use cases.

Workshop presenters:

• Pedro Moreira
• Laura Connolly
• Michelle Song
• Kaito Hara-Lee
• Lueder A. Kahrs

Smart monitoring represents an emerging paradigm for patient diagnosis, follow-up, and rehabilitation. This approach is becoming increasingly relevant as demographic trends indicate a growing number of individuals requiring care, alongside a declining availability of caregivers.
Imagine an environment in which every step taken and every surface touched can be seamlessly tracked. Within such a setting, patients can be monitored holistically, from eye movements to plantar pressure, while they move and interact in an ecological, real-life context.

This workshop explores how sensor networks integrated with artificial intelligence and machine learning algorithms can effectively support clinicians in delivering personalised and efficient care for neurological patients. Results from different clinical trials will be presented and discussed, enabling participants to reflect on the correlations between standard clinical assessments and sensor-derived metrics. The workshop will also encourage brainstorming on new application scenarios and foster interdisciplinary networking among professionals from different fields, including clinicians, occupational therapists, and engineers.

Through three representative use cases: multiple sclerosis, elderly populations, and pediatric patients, and contributions from invited speakers, the workshop will highlight the potential of smart monitoring technologies to enhance assessment, continuous monitoring, and rehabilitation outcomes.

Workshop presenters:

• Giulia Bodo
• Andrea Piccardo
• Megan O’Brien
• Joachim Hermsdörfer
• Dr. Ekaterini Zigoura