Abstract: |
INTRODUCTION: Morphologic features of computed tomography (CT) scans of the brain can be used to estimate intracranial pressure (ICP) via an image-processing algorithm. Clinically, such estimations can be used to prognosticate outcomes and avoid placement of invasive intracranial monitors in certain patients with severe traumatic brain injury. Features on a CT scan that may correlate with measurements of low ICP are sought. METHODS: A measure is proposed that is a function of the distribution of cerebrospinal fluid (CSF) in and around the brain. In our method, we present an algorithm that semiautomatically segments brain parenchyma from CSF, and apply standard image processing calculations. The ratio of CSF volume to the size of the intracranial vault (ICV) or volume inside the skull, csf(v) /icv(v) is calculated and then plotted against the actual recorded ICP, yielding a relationship between the image features and ICP. RESULTS: We analyzed a total of 45 scans from 20 patients with severe traumatic brain injury (TBI). We showed that a ratio csf(v)/icv(v) > .034 correlates with an ICP 20 mmHg (P = .0046). For csf(v)/icv(v) = .034, a distinction between low and high ICP cannot be effectively estimated by this univariate measure. CONCLUSION: This method permits a noninvasive means of identifying patients who are low risk for having elevated ICP; by following Brain Trauma Foundation guidelines strictly such a patient may be subjected to an unnecessary, invasive procedure. This work is a promising pilot study that will need to be analyzed for a larger population. |