Reliability assessment of computerized tomography scanning measurements in intracerebral hematoma

As one of the aspects of the International Surgical Trial in Intracerebral Haemorrhage (STICH), prerandomization computerized tomography (CT) scans were collected. In the present study the authors determined the inter-and intraobserver variability of various parameters pertinent to CT scans obtained in patients with intracerebral hematomas (ICHs). A protocol was devised to analyze CT scans in a uniform and systematic manner. Each observer evaluated the same set of scans twice, with a minimum 2-month interval between assessments. In addition to noting the side and the sites of involvement, the observers measured the scale present on the scan itself and the length, breadth, height, and depth of the spontaneous ICH as well as the midline shift. The intraclass correlation was very high (kappa 0.8-1) for the measurements of volume, depth, and midline shift. Good interobserver agreement (kappa 0.8-1) was demonstrated with regard to involvement of basal ganglia or thalamus, presence of intraventricular extension, and the side of the hematoma. Agreement was substantial (kappa 0.61-0.8) with regard to identifying primary involvement of particular lobes. Agreement was moderate (kappa 0.41-0.6) on the presence or absence of hydrocephalus. When comparing the first and the second sets of readings, the intraobserver agreement was good (80-100%). The study quantifies the degree of inter- and intraobserver agreement regarding evaluation of CT scans in patients with ICH when conducted in accordance with a set protocol.