Detection of Vasospasm in Comatose Patients Using Eeg Spectral Characteristics

Abstract

Subarachnoid hemorrhage (SAH) is usually caused by the rupture of a brain aneurysm. Vasospasm, the constriction of cerebral vessels reducing blood flow to the brain (leading to ischemia or infarction), is a major cause of morbidity/mortality following SAH. Transcranial Doppler and computed tomography are used to detect vasospasm but cannot be used continuously. However, electroencephalography (EEG) is a sensitive indicator of brain ischemia and also can be monitored continuously.

We therefore developed a new algorithm to calculate and detect changes in quantitative EEG features such as alpha band spectral flatness ratio (SFR; ratio of geometric mean of power to the arithmetic mean of power expressed in dB), as an indicator of vasospasm.

SFR was calculated from auto-regressive model coefficients obtained using Burg’s algorithm. Discriminant analysis was used to classify vasospasm and non-vasospasm segments from the EEG. In a pilot study, classification performance was evaluated in 6 patients using the leave-one-out method. The ‘True Positive Fraction’ and ‘True Negative Fraction’ were 83% in both cases. We suggest this algorithm may be used clinically to monitor a patient’s EEG following SAH, for the early detection of vasospasm. As a result, earlier treatment of vasospasm may result in better patient outcome.