Performance of computer-aided diagnosis for detection of lacunar infarcts on brain MR images: ROC analysis of radiologists' detection

The detection and management of asymptomatic lacunar infarcts on magnetic resonance (MR) images are important tasks for radiologists to ensure the prevention of sever cerebral infarctions. However, accurate identification of lacunar infarcts is a difficult. Therefore, we developed a computer-aided diagnosis (CAD) scheme for detection of lacunar infarcts. The purpose of this study was to evaluate radiologists' performance in detection of lacunar infarcts without and with use of CAD scheme. 30 T1- and 30 T2- weighted images obtained from 30 patients were used for an observer study, which were consisted of 15 cases with a single lacunar infarct and 15 cases without any lacunar infarct. Six radiologists participated in the observer study. They interpreted lacunar infarcts first without and then with use of the scheme. For all six observers, average area under the receiver operating characteristic curve value was increased from 0.920 to 0.965 when they used the computer output. This CAD scheme might have the potential to improve the accuracy of radiologists' performance in the detection of lacunar infarcts on MR images.