Assessment of PET imaging devices: the case of a LSO/NaI PET-SPECT prototype

Purpose: The aim of this study was to relate lesion detectability to image contrast and noise for fi ve diff erent PET scanners, by performing an observer study of thorax phantom images. Methods: A thorax phantom containing four fi llable spheres with diameters of 10, 13, 17 and 23 mm in the mediastinal region, as well as cold lung, liver and spine inserts, was fi lled with 20 MBq F. The activity concentration in the spheres was 2.5, 5 or 10 times the background concentration. Emission and transmission measurements were made at each concentration ratio with each of the scanners using routine clinical acquisition settings. Images were reconstructed using routine clinical protocols both with and without attenuation correction. Images were evaluated by ten observers and either rated positive or negative. Actual image contrasts and image noise were calculated for each scanner and sphere to background ratio and related to lesion detectability. Results: Total lesion detectability for all sphere sizes and contrasts ranged from 35% to 70%, using attenuation corrected images. Lesion detectability was lower for images without attenuation correction. Lesion detectability for each scanner correlated well with contrast to noise ratio averaged over all lesions and sphere to background ratios (r 0.90). Conclusion: Overall, lesion detectability was signifi cantly better for attenuation corrected images and 3D acquisitions. Lesion detectability was predicted well by contrast to noise ratios in the images. 070168 Urvi Joshi.indb 20 21-05-2007 15:15:09 Lesion detectability predicted by a simple phantom study