Algorithm produces patient-specific estimates, with no calibration or training
FRIDAY, April 13 (HealthDay News) -- A model-based, patient-specific, calibration-free approach has been developed for estimation of intracranial pressure (ICP), according to a study published in the April 11 issue of Science Translational Medicine.
Faisal M. Kashif, Ph.D., from the Massachusetts Institute of Technology in Cambridge, and colleagues used a model-based approach to continuous estimation and tracking of ICP. Estimates were based on time-synchronized, routinely obtainable, noninvasive or minimally invasive measurements of peripheral arterial blood pressure and blood flow velocity of the middle cerebral artery, both obtained at intra-heartbeat resolution.
The researchers found that, with no calibration or training, their algorithm produced patient-specific ICP estimates. ICP estimates for 2,665 non-overlapping 60-beat data windows were generated using 35 hours of data from 37 patients with traumatic brain injury. These estimates were referenced against simultaneously recorded invasive parenchymal ICP and achieved a mean error (bias) of 1.6 mm Hg and standard deviation of error (SDE) of 7.6 mm Hg. Blood flow velocity recordings were available from the left and right middle cerebral arteries over 22 hours; for the 1,673 data windows, averaging the resulting bilateral ICP estimates decreased the bias and SDE to 1.5 and 5.9 mm Hg, respectively. The accuracy was similar to that seen in some of the invasive methods currently used.
"Overall, our results suggest that noninvasive, continuous, calibration-free and patient-specific estimation of ICP with clinically acceptable accuracy is feasible," the authors write. "Such technology has the potential to dramatically improve neuromonitoring in a variety of conditions in which ICP cannot be assessed currently."
Several authors are listed as co-inventors on a patent related to the study subject matter. One author disclosed a financial tie to the ICM+ software package used for multimodal neuro-intensive care monitoring.
Abstract (http://stm.sciencemag.org/content/4/129/129ra44 )Full Text (subscription or payment may be required) (http://stm.sciencemag.org/content/4/129/129ra44.full )