A mathematical model that estimates intracranial pressure using measurements of arterial blood pressure and middle cerebral artery blood flow velocity provided accurate readings in a retrospective, proof-of-concept study of patients with severe traumatic brain injury.

This model-based approach for noninvasive intracranial pressure (ICP) monitoring, which allows near-real-time estimation, does not require patient-specific calibration or training on a reference population. These features make it an improvement over previously tested methods for noninvasively measuring ICP, such as sensing tympanic membrane displacement, measuring the diameter of the optic nerve sheath, or applying external pressure to the eyeball while monitoring ophthalmic artery flow.

Thomas Heldt, Ph.D., and Faisal M. Kashif, Ph.D., of the Massachusetts Institute of Technology, Cambridge, worked with colleagues at the University of Cambridge (England) and at Beth Israel Deaconess Medical Center, Boston, to obtain and analyze a data set of 35 hours of simultaneous recordings of arterial blood pressure from radial artery catheterization (used as a surrogate for arterial blood pressure in the brain), cerebral blood flow velocity from transcranial Doppler ultrasound, and ICP measurements from an indwelling parenchymal probe in 37 comatose patients with severe closed head injury (Sci. Transl. Med. 2012;4:129ra44). None of the patients had a significant stenosis in the vasculature between the radial artery and the middle cerebral artery, which would affect arterial blood pressure measurements. Within each time window of 60 heartbeats that the investigators used to estimate ICP, they ignored the production of cerebrospinal fluid, but they accounted for it when providing longer estimates of ICP across nonoverlapping time windows.

In an estimate of the ICP over 10 consecutive 60-beat windows of data for each patient, there was an error of 1.6 mm Hg and a standard deviation error of 6.9 mm Hg. In 30 patients who had bilateral, rather than just unilateral, cerebral blood flow velocity recordings, the results improved to an error of 1.5 mm Hg and a standard deviation error of 4.9 mm Hg. The individual ICP estimates for each of the patient records, rather than just the data windows, gave results of 0.9 mm Hg and 6.5 mm Hg, respectively. The estimates performed well with pressures up to 100 mm Hg.

The model could identify ICP greater than 20 mm Hg – a commonly used threshold for treatment in traumatic brain injury – with 83% sensitivity and 70% specificity for all data sets combined. On a patient-record basis, a threshold of greater than 20 mm Hg gave 90% sensitivity and 80% specificity.

"The feedback that we’ve gotten from our clinical colleagues is that they are very encouraged by the statistics and accuracy that our method shows, that we’re getting within range of what is clinically acceptable, but we’re not there yet," Dr. Heldt said in an interview.

The technique for estimating ICP makes it possible to perform spot assessments or intermittent monitoring, or obtain "accurate estimation of the trend in ICP" over a period of hours, Dr. Heldt said.

Dr. Stephan A. Mayer, head of the division of critical care neurology at the Neurological Institute of New York at Columbia University Medical Center, agreed, noting that "noninvasive ICP monitoring doesn’t even have to be that precise," particularly if it is accurate enough to screen patients for elevated ICP who may need invasive monitoring.

But the many attempts to develop a noninvasive approach to ICP monitoring over the past 20 years have suffered from problems of reproducibility of results in different settings, said Dr. Mayer, who was not involved in the study. "The real trick is going to be who can make it simple and work in the hands of anybody with minimal expertise," he said in an interview.

Dr. Heldt and his colleagues are planning to work with colleagues at Beth Israel Deaconess Medical Center to validate their model further in subarachnoid hemorrhage patients. Future work would also try to examine other injuries to the brain to see if there are particular pathologies or patient profiles where the model breaks down. If the method is validated in those studies, Dr. Heldt said the researchers plan to try it in conditions where ICP data are not already available for comparison, such as in chronic migraine, headaches, or certain kinds of visual disorders.

"We currently do not have a way of assessing ICP in these kinds of conditions, and don’t really know if ICP is a contributing factor or might even be a causative factor," Dr. Heldt said.

The study was supported by grants from the National Institutes of Health and the Center for Integration of Medicine and Innovative Technology. Massachusetts Institute of Technology has filed patent applications for the algorithms involved, listing several authors as coinventors. One of the authors has a financial interest in a software package for multimodal neurointensive care monitoring, which has a commercially available noninvasive ICP plug-in. Dr. Mayer said that he is on the scientific advisory board for Orsan Medical Technologies, an Israeli start-up company that is developing noninvasive ICP technology. He also receives research support from Noninvasive ICP Technologies.

This column, "Neuroscience Today, Neurology Tomorrow," appears regularly in Clinical Neurology News, a publication of Elsevier.

Simple Model May Provide Useful Noninvasive ICP Estimates