Non-invasive Intracranial Pressure estimation using Support Vector Machine

Abstract

Intracranial Pressure (ICP) measurements are of great importance for the diagnosis, monitoring and treatment of many vascular brain disturbances. The standard measurement of the ICP is performed invasively by the perforation of the cranial scalp in the presence of traumatic brain injury (TBI). Measuring the ICP in a noninvasive way is relevant for a great number of pathologies where the invasive measurement represents a high risk. The method proposed in this paper uses the Arterial Blood Pressure (ABP) and the Cerebral Blood Flow Velocity (CBFV) - which may be obtained by means of non-invasive methods - to estimate the ICP. A non-linear Support Vector Machine was used and reached a low error between the real ICP signal and the estimated one, allowing an on-line implementation of the ICP estimation, with an adequate temporal resolution.

Review: pathophysiology of intracranial hypertension and noninvasive intracranial pressure monitoring

“…Due to the somewhat complicated interplay between ICP, arterial pressure, and cerebral hemodynamics, a large number of methods have attempted to incorporate arterial pressure measurements, which may provide complementary information to assist in measuring ICP [125,154,155,[157][158][159][160][161][162][163][169][170][171][172][173][174][175]. These methods are not technically noninvasive as they require the placement of a radial artery catheter for monitoring ABP; however, this procedure is typically already performed as part of standard care in neurointesive care units, and the risks associated with monitoring ABP via radial artery catheter are considered significantly less risky than invasive ICP monitoring.…”

“…Kim et al proposed a method also based on the morphological analysis of CBFV waveforms using both supervised and semi-supervised learning approaches and reported predictive accuracies as high as 82% and 92%, respectively, in classifying normal vs. hypertensive intracranial pressure [167]. Additional data-based learning techniques including the use of support vector machines, linear discriminant analysis, and random forests using features extracted from ABP and CBFV waveforms have also been shown to achieve low error and promising classification accuracies in isolated cases [168][169][170].…”

Non-Invasive Intracranial Pressure Monitoring

“…A fuzzy logic algorithm has also been applied to TCD and arterial blood pressure [ 84 ], and support vector machines [ 85 ] have also been used.…”

Intracranial Pressure and Intracranial Elastance Monitoring in Neurocritical Care

“…Continuous ABP measurement is routinely available in a wide spectrum of critical care patients, mostly through radial artery catheterization, and CBFV measurement can be safely acquired over long periods of time, especially in comatose or heavily sedated patients, through transcranial Doppler (TCD) ultrasonography and is indicated in NCC patients with neurovascular pathologies. Previous approaches to estimating ICP from the ABP and CBFV waveforms applied learning methods, such as nested regressions (63, 64, 67), neural networks (65), or support-vector machines (66,70), that map features of the ABP and CBFV input data to ICP. In such data-driven approaches, the mappings are learned from a set of training data in which invasive ICP measurements are available.…”