Melanoma patients evaluated by four different positron emission tomography reconstruction techniques.

One hundred and nineteen patients with malignant melanoma were studied using 2-[ F]-fluoro-2-deoxy-D-glucose positron emission tomography (FDG PET). The images were reconstructed using ordered subset expectation maximization with and without attenuation correction and filtered backprojection with and without attenuation correction. The most probable draining lymph node chains were surgically explored and the tumour volume was quantified at histology. The four different reconstructions of the PET images were retrospectively graded on a five-point scale by two blind readers and compared with the tumour volume. The readers agreed within +/-1 grade 93% (529/568) of the time. Comparing the areas under the receiver operating characteristic (ROC) curves gave 0.698, 0.668, 0.694 and 0.684 for the four reconstruction techniques. The lowest value comparing any pair of the four reconstruction techniques was P=0.371. Thus, none of the reconstruction techniques gave significantly better results than any of the others. The sensitivity of detection was 85% for tumour volumes of 113 m or more (about 6 mm in diameter), but only 4% for tumours less than this size. It can be concluded that the use of attenuation correction gives aesthetically more pleasing images, but the sensitivity and specificity are not significantly improved.