Multimodal Signal Processing for Objective Human Pain Assessment

Pain is an essential biological signal, yet its assessment in clinical practice remains largely subjective and dependent on self-report or behavioural observation. This approach becomes particularly unreliable in individuals who are unable to communicate effectively, including stroke survivors, infants, patients with neurological disorders, or those under sedation. The aim of this project is to develop an accurate and objective framework for pain assessment by analysing multiple physiological signals that reflect the body’s automatic response to painful stimuli. These signals originate from the skin, heart, and brain, each capturing a distinct dimension of the pain response. By integrating these complementary sources of information, the project seeks to provide a more comprehensive and reliable assessment than any single modality alone.

This research focuses on multimodal signal processing and advanced artificial intelligence models to identify robust and generalisable biomarkers of pain across individuals. By integrating information from the skin, heart, and brain, the project aims to improve classification accuracy while reducing reliance on behavioural cues. Controlled laboratory-induced pain is used for systematic evaluation and model development, with the long-term objective of facilitating clinical validation. The work contributes to the development of accurate, data-driven, and physiologically grounded tools for objective human pain assessment.