Abstract
Pain is defined as an unpleasant sensory and emotional experience associated with actual or potential tissue damage, yet it remains an entirely subjective state for which no simple objective marker exists. This study utilizes machine learning techniques to uncover pain signatures within resting-state EEG data at the sensor level. It reveals that simple, costeffective, and clinically relevant brain signatures associated with chronic pain can be derived from raw EEG data. These signatures are characterized by connectivity patterns in the gamma band, particularly evident in the central EEG electrodes. This study helps build a foundation for a potential AI-based approach to creating better neuromodulation treatments for chronic pain, which has profound implications for improving pain management strategies and enhancing patient care. This thesis comprises three primary line studies and two validation studies. The first study aims to explore the presence of sex differences in resting-state raw EEGs at a sensor level. Subsequently, the second study focuses on identifying unisex neural signatures for neuropathic pain, fibromyalgia, and their combined condition. Next, as sex differences are observed, the research proceeds to investigate sex-specific neural signatures. Furthermore, validation of the pain signatures on other datasets is conducted to ensure the generality of the identified neural signatures. Lastly, a customized feature selection algorithm is presented and validated by comparing it with other benchmark schemes