Feasibility study of the use of similarity maps in the evaluation of oncological dynamic positron emission tomography images

A preliminary study is presented on the potential role of similarity mapping (SM) in the evaluation of oncological dynamic 18F-fluorodeoxyglucose positron emission tomography studies, mainly in lesion localisation and detectability. Similarity maps were calculated using previously described (correlation coefficient (COR) and normalised correlation coefficient (NCOR)) and newly introduced similarity measures (sum of squares coefficient (SSQ), squared sum coefficient (SQS), sum of cubes coefficient (SC) and cubed sum coefficient (CS)). The results were evaluated using simulated and clinical data. The study revealed that the best-suited similarity measure for such applications was the CS similarity coefficient, which provided the best parametric images, delineating structures of interest and supporting the visual interpretation of data sets. It was shown that SM and standardised uptake value (SUV) images had comparable diagnostic performance, although SM was able to offer additional time-related information in a single image. For the case of colorectal recurrences (17 cases), the measured contrast values for the CS and SUV images were 2.36 +/- 0.47 and 4.12 +/- 0.42, respectively, whereas, for three cases of giant cell tumours, these values were 11.6 +/- 2.1 and 11.9 +/- 1.8, respectively.