Accurate localization of incidental findings on the computed tomography attenuation correction image

This observer performance study assessed lesiondetection in the computed tomography attenuationcorrection image, as would be produced for myocardialperfusion imaging over a tube current (mA) range. A staticanthropomorphic chest phantom containing simulatedpulmonary lesions was scanned using the four availablemA values (1, 1.5, 2 and 2.5) on a GE Infinia Hawkeye 4.All other computed tomography acquisition parametersremained constant throughout. Twenty-seven casesshowing zero to four lesions were produced for afree-response receiver-operating characteristic method.Image observations were completed using our novelweb-based ROCView software under controlled conditions.The Jackknife alternative free-response receiver-operatingcharacteristic (JAFROC) figure of merit was used forsignificance testing, wherein a difference in lesiondetection performance was considered significantat P values less than 0.05. Twenty readers with varyingcomputed tomography experience (0–24 years) evaluated108 images using an ordinal scale to score confidence.The JAFROC analysis showed that there was nostatistically significant difference in performance betweenmA values (P= 0.439) for this sample of observers.In conclusion, no significant difference in lesion detectionperformance was seen between the mA values. Thissuggests that there is no value in using anything otherthan the lowest mA value for the investigation of incidentalextracardiac finding