This is a prospective, observational cohort study aimed at constructing a machine learning-based prognostic model for severe brain-injured patients. The study will synchronously collect continuous glucose monitoring (CGM), electroencephalography (EEG), near-infrared spectroscopy (fNIRS), transcranial Doppler (TCD), and serum neuronal injury biomarkers (NSE, S100β) within 72 hours post-injury. The goal is to investigate the correlation between glycemic variability (GV) and neurological function and to develop an integrated model for early prediction of 3-6 month neurological outcomes (GOSE score).

This study intends to enroll 50 adult patients with brain injury admitted to the ICU. Multimodal monitoring data will be collected prospectively. Machine learning algorithms will be used to integrate the data and build a predictive model. The study will test whether integrated metabolic-neurological monitoring outperforms traditional single-parameter prognostic methods.

Inclusion Criteria:

• Diagnosis of severe TBI, large-volume stroke, or HIE
• Expected ICU stay >72 hours
• Informed consent from legal surrogate

Exclusion Criteria:

• Terminal organ failure
• Pre-existing severe neurological disease
• Skull defect preventing monitoring
• Pregnancy or lactation

The study population will consist of adult patients (≥18 years old) admitted to the Intensive Care Unit (ICU) of Chenzhou First People's Hospital with a diagnosis of brain injury. This includes patients with:Severe Traumatic Brain Injury (TBI), Expected ICU stay >72 hours,Informed consent from legal surrogate.