Improvement of detection of hypoattenuation in acute ischemic stroke in unenhanced computed tomography using an adaptive smoothing filter.

BACKGROUND

Much attention has been directed toward identifying early signs of cerebral ischemia on computed tomography (CT) images. Hypoattenuation of ischemic brain parenchyma has been found to be the most frequent early sign.

PURPOSE

To evaluate the effect of a previously proposed adaptive smoothing filter for improving detection of parenchymal hypoattenuation of acute ischemic stroke on unenhanced CT images.

MATERIAL AND METHODS

Twenty-six patients with parenchymal hypoattenuation and 49 control subjects without hypoattenuation were retrospectively selected in this study. The adaptive partial median filter (APMF) designed for improving detectability of hypoattenuation areas on unenhanced CT images was applied. Seven radiologists, including four certified radiologists and three radiology residents, indicated their confidence level regarding the presence (or absence) of hypoattenuation on CT images, first without and then with the APMF processed images. Their performances without and with the APMF processed images were evaluated by receiver operating characteristic (ROC) analysis.

RESULTS

The mean areas under the ROC curves (AUC) for all observers increased from 0.875 to 0.929 (P = 0.002) when the radiologists observed with the APMF processed images. The mean sensitivity in the detection of hypoattenuation significantly improved, from 69% (126 of 182 observations) to 89% (151 of 182 observations), when employing the APMF (P = 0.012). The specificity, however, was unaffected by the APMF (P = 0.41).

CONCLUSION

The APMF has the potential to improve the detection of parenchymal hypoattenuation of acute ischemic stroke on unenhanced CT images.