Transcranial cerebral oximetry, transcranial Doppler sonography, and heart rate variability: useful neuromonitoring tools in anaesthesia and intensive care?

Abstract

Sophisticated software algorithms and miniaturized hardware components have opened new non-invasive vistas to monitor the central nervous system. Electrophysiological modalities (electroencephalography (EEG), evoked potentials) can be used to assess the integrity (or compromise) of neuronal structures at different levels. Spectroscopic methods are used to evaluate oxygen metabolism in the brain and ultrasound techniques can depict cerebral perfusion. Quantifying oscillations of the heart-rate sequence sheds light on the regulatory systems governing complex autonomic feedback loops. Transcranial cerebral oximetry Transcranial near infra-red spectroscopy (NIRS) is a fascinating technique that promises to provide information on the balance between oxygen supply and demand in the brain through the intact skull. It can detect situations in which the oxygen status of the brain can change dangerously and where the peripheral systemic haemodynamics and oxygen saturation would not predict the changes. Transcranial NIRS is a technique based on the Beer-Lambert Law [1,2]. The values obtained with cerebral oximetry depict primarily the oxygen status of the chromophores (haemoglobin-deoxyhaemoglobin) in the venous compartment (75%) [3] and on the intracellular redox state (cytochrome aa3) [4]. NIRS monitoring is used in a number of surgical procedures (e.g., carotid, neuroendovascular, open heart and aortic arch surgery) [5-10]. It is also applied in the critical care setting for detecting cerebral hypoxia in patients with severe brain injury [11-17], aneurysmal subarachnoid haemorrhage [18,19], low cardiac output states, pulmonary and vascular diseases, sepsis and anaemia [20]. Unfortunately the appealing prospect of simply placing a sensor on the forehead and obtaining a numeric readout of the oxygen status of the brain has led to over simplifications and premature expectations that could not be fulfilled. NIRS data have to be interpreted in the context of the underlying pathophysiology. Information is required on systemic arterial pressure, peripheral oxygen saturation, oxygen carrying capacity, body temperature, carbon dioxide, cerebral arterial or venous obstruction, and cerebral seizures. Mistakes by the user (e.g., insufficient light shielding, ineffective probe fixation and incorrect positioning of optodes) affect the results [21]. The problem of NIRS is the hydra-like quality of the technique: NIRS can have remarkably high sensitivity for minimal physiological [22] or pathophysiological [23] shifts and therapeutic effects [21]. By the same token the technique is limited by its reach: the saturation values are representative only of the region directly beneath the sensor and may not be sensitive to changes in other locations. Furthermore, the computational algorithms used in several NIRS devices assume that the infra-red signal exclusively reflects intravascular haemoglobin. Admixture of this signal with that obtained from a stagnant pool of deoxygenated blood can result in values of no clinical significance. The spatial orientation of the optode to the underlying healthy or abnormal anatomical structures is critical [24]. Measurements over regions of infarct or absent brain tissue can produce spurious readings. Metal plates implanted after craniotomy make monitoring impossible and the absence of frontal bone can result in overscale reflected signals. A further methodological problem with NIRS is the extracerebral contribution to cerebral oximetry. Results from carotid surgery show that the contribution of the extracranial circulation to the measured oxygen values is insignificant [25]. In contrast, changes in scalp oxygenation or in extracerebral perfusion of the head have a significant effect on NIRS readings [26-29]. Numerous studies show a close correlation between changes in cerebral oxygenation assessed with NIRS and other monitoring modalities under varying clinical conditions. But correlation does not prove causation, and m