Masters Thesis Presentation - Andrei-Stefan Istudor

Title: Automating Pain Score Assessment in Neonatal Intensive Care Units

Abstract:

With millions of infants born preterm each year worldwide, automating manual pain assessment processes in neonatal intensive care units (NICUs) is critical for clinical monitoring and decision-support. However, current visual and multi-modal methods often fail due to medical equipment, data availability and imbalance, infant swaddling, and visual obstructions during procedures. This study explores a purely parameter-driven alternative approach by analysing high-resolution vital sign data to predict pain during acute procedures, such as heel prick blood tests. We have developed an advanced deep learning framework to process continuous physiological data from NICU patients, capturing pain-related changes in vital signs. Our study demonstrates that this non-invasive, objective method provides an imbalance-aware baseline for continuously tracking neonatal pain without the limitations encountered when relying solely on visual monitoring. Finally, this paper provides a foundation for future clinical tools that can automatically monitor pain and be used to support decision-making within NICU environments.

Supervisors: George Azzopardi, Jirí Kosinka