Klinische Neurophysiologie 2006; 37 - A196
DOI: 10.1055/s-2006-939279

Non-invasive ICP assessment – approaches of improvement

B Schmidt 1, B Schmidt 1, M Czosnyka 2, SF Bocklisch 3, M Päßler 3, C Leege 4, JJ Schwarze 1, J Klingelhöfer 1
  • 1Klinikum Chemnitz
  • 2Addenbrookes Hospital, Cambridge, UK
  • 3TU Chemnitz
  • 4Universitätsklinikum der TU-München

Introduction: In a recently introduced mathematical model ICP was assessed non-invasively (nICP) by usage of arterial blood pressure (ABP) and cerebral blood flow velocity (FV) curves.

Well defined haemodynamic parameters (TCD characteristics) derived from both signals, were taken to control a linear signal transformation of ABP into ICP. In the current work attempts were made to improve former results by application of advanced data analysis techniques.

Methods: In 201 patients (3–78 years, mean: 37±19 years) with either severe traumatic brain injury (N=176) or stroke (N=25) signal data consisting of FV, ABP and (invasively assessed) ICP was used to validate nICP results. Modifications of a basic linear nICP model were made by means of a method called Fuzzy Pattern Classification which allowed an individual model adaptation in heterogeneous patient groups. In a second approach Fuzzy Pattern Classification was used for a pre-selection of the best fitting linear model. The results of these modified methods were compared to the former results of nICP assessment.

Results: Two different linear models which were using different TCD characteristics for nICP calculation were evaluated. The models showed mean absolute differences between ICP and nICP (ΔICP) of 6.41±5.74mm Hg (median=4.92mm Hg) and 6.39±5.76mm Hg (median=5.04mm Hg). The Fuzzy Pattern approach of individual model adaptation resulted in a ΔICP of 6.54±5.84mm Hg (median=5.29mm Hg). The three results did not differ significantly (P>0.05). If Fuzzy Pattern Classification was used to control application of linear procedures by selecting the best fit procedure the ΔICP could be reduced to 6.21±5.50mm Hg (median=4.83mm Hg) (P<0.001).

Conclusions: Although advanced methods of data analysis were used only minor improvements of nICP assessment accuracy could be achieved. With regard to the underlying data only low potential seems to be left for further improvement. Therefore, realisation of this method should be considered now.