Dynamic validation of the estimates obtained by lumbar infusion test in hydrocephalic patients

The infusion test is a routine clinical tool to study the efficacy of intracranial volume compensation in neurosurgical practice. Traditional approach requires measurement of two relatively long steady states, totaling up to an hour of invasive testing. Therefore, we investigate here the consistency of estimates arising from shorter monitoring times. We developed an off-line software, based on widely accepted model of cerebrospinal fluid circulation, and used it to validate the dynamically evolving test estimates in 25 recordings in patients referred for the infusion test, to confirm diagnosis of hydrocephalus. Our results show that, typically, 4 minutes is required to reach the inflection point in the response curve and, about twice that time to produce estimates within 10% error from the final results. This effect depends on number of free parameters allowed in model estimation and varies between the clinical cases depending on their status and amount of the measurement noise. The observation of the dynamical convergence in the model parameters provides an important information regarding the stability and accuracy of the latest estimates. Real-time application of our method to clinical practice could potentially save 30% of monitoring time or provide an alternative means to validate the accuracy of the test.