Exploring physiologic regulatory factors in traumatic brain injury (TBI) through Correlation Analysis and Graph Neural Network

Abstract Purpose Traumatic brain injury (TBI) is one of the most common cause of mortality and disability globally. Intensive care unit (ICU) management poses significant challenges for medical practitioners, primarily because of the complex interplay between biomarkers and hidden interactions. This study aimed to uncover subtle interconnections between biomarkers and identify the key factors contributing to TBI characteristics and ICU severity scores. Methods A total of 29 patients with TBI who were admitted to the ICU were selected and analysed using monitoring electrocardiography (ECG), vital signs, Glasgow Coma Scale (GCS) and electronic medical records. This study utilized a methodology that integrates correlation-based network analysis and graph neural network (GNN) techniques to uncover hidden relationships between various biomarkers and identify the most critical monitoring biomarkers for patients with TBI within the first 12 hours of ICU stay. Results The analysis revealed significant associations within the dataset. Specifically, MeanRR exhibited notable connections with alterations in systolic blood pressure and heart rate variations. Moreover, the final GCS showed a strong correlation, including long-term correlation with heart rate variability (HRV) feature alpha2, variability in atrial blood pressure means and diastolic blood pressure, gender, and age. Variability of diastolic blood pressure, GCS ICU scoring values, and pNN50 (an HRV measure) demonstrated strong association with other biomarkers during the first 12 hours following ICU admission. Conclusion HRV as an electronic biomarker and the variability in physiological variables during first 12 hours in the ICU are equally important factors for TBI severity assessment and can offer valuable insights into the patient's health prognosis.