Purpose: In recent years, radiofrequency identification has been used for the continuous measurement of intracranial pressure (ICP) in patients with a cerebrospinal fluid (CSF) shunt for hydrocephalus . Unlike ICP monitoring in an inpatient setting, measurements in mobile patients outside the hospital provide ICP data that take into account the everyday activities of each individual patient. Common methods of ICP monitoring and analysis cannot be used for those patients. In addition, ICP measurements in mobile patients require considerably longer observation times than in-hospital monitoring. For this reason, ICP measurements over a period of 7 to 10 days must be analyzed effectively and efficiently.
Methods: A possible approach is to analyze ICP data graphically. Pathologic changes can be expected to be associated with specific patterns that can be detected graphically (e.g., Lundberg A waves). Patients without pathologic ICP values and without intracranial pathologies usually show an approximately normal distribution of ICP data. By contrast, patients with pathologic ICP values are likely to show major deviations from a normal distribution such as changes in minimum and maximum values and multimodal distributions. Against this background, we present a new graphical method for detecting pathologic conditions. This novel method is based on the distribution of ICP data that is assessed using GNU R, a free software package for statistical computing and graphics.
Results: A left-skewed distribution indicates CSF shunt overdrainage and a right-skewed distribution suggests CSF shunt underdrainage. In addition, an additive analysis of the number of physiologic ICP values can be helpful in detecting possible causes of CSF shunt overdrainage or underdrainage. The approach presented here shows that patients with hydrocephalus objectively benefited from ICP-guided adjustments of the opening pressure of a shunt valve or the insertion of a valve. This objective improvement was confirmed by the patients' subjective perception of well-being.
Conclusions: Further investigations should be performed to examine the influence of multimodal ICP distributions and to assess how data analysis is affected by a drift that can occur when a sensor has been in place for an extended period of time.
Georg Thieme Verlag KG Stuttgart ยท New York.