On the fractal nature of dynamic positron emission tomography (PET) studies

Quantification of dynamic PET Studies is generally based on compartmental methods. This is a report about the implementation and assessment of a new non-compartmental method, the fractal dimension (FD), a parameter based on the box counting (BC) procedure of the chaos theory for the analysis of dynamic PET data. The evaluation included 200 malignant lesions in 159 patients with different tumour entities as well as 57 benign lesions for comparison. 101/200 malignant lesions were treated with chemotherapy, whereas 99/200 malignant lesions as well as all 57 benign lesions were untreated within the last six months prior to the PET study with F-18-Fluorodeoxyglucose (FDG). The evaluation of the FDG kinetics was performed using the BC based FD for the time-activity data. Visual assessment demonstrated generally different FDG uptake patterns in the conventional images and the parametric images of FD. FD estimates depended on the number of boxes and the maximum cut-off value used for calculation. Based on the discriminant analysis for benign and malignant lesions, FD demonstrated an accuracy of 76.65% for all patients, 67.7% for the untreated patients and 83.44% for the treated group. The use of the BC based FD is a reliable, new method for the quantification of dynamic PET studies and seems to be in particular helpful for the evaluation of treated malignant lesions. (author)