ABCP 2026 Annual Conference: Life Sciences and Health

Abstract:

Photoplethysmography (PPG) is widely used in wearable devices for continuous physiological monitoring, but its performance can be strongly affected by real-world conditions. In this talk, I will present our work on improving the robustness of wrist PPG sensing under varying contact pressure and dynamic sports activities. I will first introduce a new PPG dataset for studying pressure-induced waveform variations and a deep learning model for improving PPG waveform quality under different contact pressures. I will then present a new sports PPG dataset and our ongoing work on improving heart rate estimation during dynamic sports activities.

Abstract:

Surface electromyography (sEMG) signal understanding holds great potential for decoding human movement and intention, yet its progress is often constrained by limited data. Large language models (LLMs) contain rich knowledge relevant to human motion, but they are inherently designed to process language rather than physiological signals. In this talk, we present a method that constructs an “sEMG language,” bridging this modality gap and enabling LLMs to serve as effective solvers for sEMG signal understanding.

Abstract:

Recent advances in machine learning are transforming audio from an unstructured signal into a rich source of physiological and behavioural information. Self-supervised learning, foundation models and multimodal AI are enabling robust representations that extract meaningful health insights from everyday sounds recorded by commodity wearable devices. Among these, earbuds provide a particularly attractive sensing platform: their unique position within the ear canal offers access to rich acoustic and motion signals that capture multiple aspects of human physiology and behaviour.

In this talk I will use examples from our recent work on AI for wearable audio sensing. I will also show how in-ear microphones enable fine-grained respiratory and cardiovascular monitoring, supporting the estimation of breathing patterns, heart activity, stroke volume and cardiac output. I will then demonstrate how earable sensing extends beyond physiology to nutrition-aware food monitoring through chewing acoustics and biomechanical assessment using inertial sensing to estimate gait dynamics and ground reaction forces. Together, these examples illustrate how advances in machine learning are transforming commodity wearables into comprehensive platforms for continuous, personalised health monitoring, and point towards a future of multimodal wearable AI that seamlessly integrates sensing, representation learning and health intelligence.

Bio:

Cecilia Mascolo is Full Professor of Mobile Systems in the Department of Computer Science and Technology, University of Cambridge, UK. She is also Chief Scientific Officer of auryx, a start up focusing on physiology sensing via in-ear microphones of commercial earbuds. She is a Fellow of the Royal Academy of Engineering. Prior joining Cambridge in 2008, she was a faculty member in the Department of Computer Science at University College London. She holds a PhD from the University of Bologna. Her research interests are in wearable systems and machine learning for mobile health. She has published in top tier conferences and journals in the area and her investigator experience spans projects funded by Research Councils and industry. She has served as steering, organizing and programme committee member of wearable and sensor systems and machine learning conferences. More details at www.cl.cam.ac.uk/users/cm542 and www.auryx.ai.

Abstract:

This talk will introduce the research activities in the WSA E-textile Innovation Lab at the University of Southampton. The talk focuses on the design and development of e-textile-based wearable devices used for monitoring health conditions and managing chronic disease symptoms such as osteoarthritis pain, as well as supporting stroke rehabilitation. I will share the research journey from proof of concept to product development and clinical trials. The talk will highlight how co-design, stakeholder engagement, and partnerships have shaped the direction and outcomes of the research. Key considerations and recommendations will be shared to support the future development of e-textile-based medical devices.

Abstract:

Foundation models have reshaped language and vision, but medical sensing data behave differently. Physiological signals from wearable and clinical devices are continuous, noisy, and device-dependent, so simply scaling data and parameters can amplify noise, bias, and distribution shift. This talk introduces “sensing intelligence”, the capabilities needed to make physiological signals learnable, scalable, and clinically useful. Using cardiac biosignals (ECG and PPG) as an example, I will walk through our efforts on (i) fundamental methodologies for learning with imperfect sensing data; (ii) building a cardiac sensing foundation model from large-scale, multi-device data; and (iii) adapting pretrained models for real clinical use, from cross-modality reconstruction to longer-horizon trajectory modelling.

Abstract:

Acute Coronary Syndromes (ACS), including ST- and non-ST-segment elevation myocardial infarction (STEMI, NSTEMI), remain a leading cause of global mortality. Traditional Cardiovascular Risk Scores (CVRS) provide important insights but mainly rely on clinical data, often neglecting environmental factors (e.g. air pollution, climate) that significantly influence cardiovascular health. Integrating complex time-series environmental and clinical datasets also presents substantial challenges. We propose TabulaTime, a multimodal deep learning framework integrating clinical risk factors with environmental data to enhance ACS risk prediction. TabulaTime delivers three innovations: multimodal integration of time-series environmental and clinical data; PatchRWKV for extracting complex temporal patterns with linear computational complexity; and enhanced interpretability through attention mechanisms. TabulaTime improves prediction accuracy by 20.5% over CatBoost, with environmental data contributing a 10.1% gain. PatchRWKV outperforms state-of-the-art models (MLP-, CNN-, RNN- and Transformer-based models). Feature analysis highlights key clinical and environmental predictors. This approach advances personalised prevention and strengthens public health against cardiovascular risks.

Abstract:

Ulcerative colitis is a chronic inflammatory bowel disease that causes recurrent inflammation in the colon and requires careful endoscopic assessment for diagnosis. Artificial intelligence is rapidly reshaping how we analyse endoscopic data in ulcerative colitis, moving beyond score prediction tasks toward richer and more clinically meaningful understanding. In this talk, I will present our recent work across this progression. I will first discuss discriminative or encoder-based models based on convolutional neural networks and transformer architectures for automated severity scoring, focusing on clinically important indices. I will then introduce our more recent exploration of multimodal large language models using a mixture of experts for ulcerative colitis captioning, where visual findings are translated into natural language descriptions. Together, these studies illustrate a broader shift from scoring disease severity to enabling multimodal interpretation, and highlight the promise of building more expressive and clinically useful AI systems for endoscopic analysis.

Bio:

Dr Xinqi Fan is a Lecturer in Artificial Intelligence at Manchester Metropolitan University. He received his BEng from Southwest University, his MEng from the University of Western Australia, and his PhD from City University of Hong Kong. He also conducted research at King Abdullah University of Science and Technology and the Chinese University of Hong Kong. His research interests include deep learning, computer vision and multimodal learning, with applications in affective computing and medical image analysis. He has published research in leading conferences and journals, including CVPR, ICCV, MM, IEEE TAFFC and IEEE TIP. He has organised challenges at FG 2026 and MM 2025, as well as a workshop at ICME 2025. His team won first place in the ISBI 2026 Multimodal Ulcerative Colitis Grading Challenge.

Abstract:

In this talk I will first present my recent research work in personalization. Personalization is popularly used in different applications including e-commerce, e-learning, and e-health. Personalized recommender systems based on user behaviours, heterogenous information network with multi-modal data will be presented. After that, I will present work in conversational and multi-modal AI (e.g., text, image, audio, tabular, gaits). The example applications include Healthcare (e.g., Sarcopenia, dementia, Aphasia, mental health, thyroid nodule diagnosis), AI for science, Research & Development of product innovation.