Effect of Attenuation Correction on Lesion Detection Using a Hybrid PET System
Chiraporn Tocharoenchai, PhD*, Benjamin MW Tsui, PhD**, Eric C Frey, PhD**, Wen-Tung Wang, PhD***
Affiliation : * Department of Radiological Technology, Faculty of Medical Technology, Mahidol University ** Department of Radiology, Johns Hopkins University, USA *** Biomedical Engineering Department, University of Virginia, USA
Objective : The purpose of this study was to investigate the effect of attenuation correction (AC) on lesion
detection for a hybrid PET system.
Materials and Methods : Experimental list-mode data were acquired from hot spheres inside a uniform cylin-
drical phantom with an elliptical cross-section using a Siemens E.CAM+ dual-camera hybrid PET system.
Spheres with inner diameters of 0.8- and 1-cm and the cylindrical phantom were filled with F-18 to simulate
lesions with lesion-to-background (L/B) ratios of 14:1 and 8:1, respectively, found in clinical PET studies.
The list-mode data of each sphere size were regrouped into sinograms with peak-to-peak energy window
settings at 30% and 20% for the 0.8- and 1-cm diameter lesion, respectively. They were then rebinned using
the single slice rebinning method. Attenuation correction was applied assuming uniform attenuation. The
sinograms with and without AC were reconstructed using 5 iterations of OS-EM algorithm with 8 angles/
subset and postfiltered with a Butterworth filter with n = 5 and fc = 0.52 cycles/cm. Human observer perfor-
mance study and localization receiver operating characteristic (LROC) analysis were used to evaluate the
reconstructed images for maximum lesion detection. Average areas under the LROC curves (ALROC) across 8
observers obtained with and without AC were determined. The null hypothesis that there was no difference
between with AC and without AC was tested using a two-tailed t-test with 95% confidence interval.
Results : The results indicated that for the 0.8-cm lesion with 14:1 L/B ratio, the ALROC decreases from 0.66 to
0.62 when AC is applied as compared to without AC and from 0.69 to 0.63 for the 1.0-cm lesion with 8:1 L/
B ratio, but no statistical significant difference (p > 0.05).
Conclusion : The authors conclude that for a phantom with hot lesions embedded in a uniform background,
AC decreases lesion detectability compared to without AC using a hybrid PET system for small lesion sizes.
Keywords : Hybrid PET, Attenuation correction, LROC study
