Classification of morphological features extracted from intracranial pressure recordings in the diagnosis of Normal Pressure Hydrocephalus (NPH)

Abstract

The intracranial pressure (ICP) monitoring is a common procedure in neuro-intensive care for pathologies such traumatic brain injuries or hemorrhages, but also for chronic ones as the Normal Pressure Hydrocephalus (NPH). The only available treatment for NPH is the surgical implantation of a shunt with the aim of routing cerebrospinal fluid (CSF) away from the brain to another part of the body. In this study, using the classification software WEKA, an intensive investigation of ICP signals has been conducted. In particular we studied 14 ICP recordings of different patients who underwent an infusion test, with the aim of investigating the presence of NPH through the ICP recording. More precisely, 20 morphological features are extracted from the ICP pulsed wave, the trend have been computed and, for each one, 9 statistical functions determined. The 180 features have been selected and passed for the classification. The results obtained shows how, among the 14 patients, a number of 12 out of 14 (85.7%) have been correctly classified, looking at just 3 features. In particular 8 out of 9 not-NPH-affected patients were correctly identified (88.89%) while 4 out of 5 NPH-affected patients were correctly identified (80%).