Compact Optical-EEG for Recovery Prediction in the ICU

Background

Assessing consciousness levels in neuro intensive care units is a critical challenge for planning optimal interventions, particularly for patients suffering from severe traumatic brain injury and disorders of consciousness. Current diagnostic and prognostic tools are frequently imprecise, susceptible to the effects of sedation and clinical treatments, and can lead to misdiagnosis. While electroencephalography (EEG) offers an objective alternative, its real-time analysis requires highly skilled specialists, and changes in EEG traces can be slow or confounded by ongoing clinical treatments and sedation. Moreover, EEG provides only a partial view of the brain's state, as it primarily monitors electrical activity and may not fully capture crucial changes in cerebral blood flow, which are essential for sustaining consciousness and can occur independently or more rapidly than electrical changes. This limitation also affects many existing machine learning models that predominantly rely on EEG data alone, highlighting a significant need for more comprehensive and robust methods to accurately evaluate consciousness.

Technology

Researchers at Stony Brook University developed a machine learning-enabled, non-invasive brain monitoring system that integrates standard electroencephalography (EEG) with continuous-wave diffuse correlation spectroscopy (CW-DCS) to classify patient consciousness at the bedside. The system simultaneously acquires and co-registers electrical signals from EEG scalp electrodes and cerebral blood flow measurements from CW-DCS at the same anatomical locations. An analysis pipeline extracts spectral EEG features, alongside low-frequency cerebral blood flow oscillations and a blood flow index from CW-DCS. These combined multimodal features are then processed by a machine learning model to classify the patient's state as conscious or unconscious. The setup is compact, portable, and designed for synchronized data acquisition.