[Quantitative evaluation of low contrast detectability in a brain computed tomography: investigation for the effect of window width on recognition of hyperacute ischemic stroke].

Abstract

In diagnostic brain computed tomography (CT) imaging of hyperacute ischemic stroke, the recognition of subtle lesions is difficult and may be missed. The aim of this study was to evaluate quantitatively the effect of the window width on low contrast detectability by using digital phantom images intended for detection of hyperacute ischemic stroke, and clinical CT images. We first derived the digital phantom images by subtracting the simulated signal data (intensity=1-3 hounsfield units (HU), diameter=10-30 mm) from the water phantom images at various mAs values. Observer studies were then performed under the various window widths at 20, 40, 60, and 80 HU by using the 30 digital phantom images and the 30 water phantom images. In addition, observer studies on brain CT images with 30 abnormal cases and 30 normal cases were performed under the window widths at 20 and 80 HU. Studies were also performed under the simultaneous display of the set of brain CT images on each window width. As a result of evaluation by receiver operating characteristic analysis, narrowing of the window width can improve the low contrast detectability in CT images with noticeable noise, and can decrease the variation in the interpretation skills between observers in clinical CT images. Moreover, the interpretation accuracy was improved by displaying simultaneously the clinical CT image set on window widths at 20 and 80 HU.