Abstract
363
Objectives Quantification of metastatic osseous tumor volume on 18F-fluoride PET/CT has been correlated with patient prognosis. Total osseous tumor volume and patient outcomes could be further evaluated by a multi-center clinical trial if a rapid, reproducible and accurate method becomes widely available that is patient specific to allow for institutional variability. Prior quantification techniques were fast and highly reproducible by utilizing a single SUV threshold and the quantitative aspect of PET. However, the single threshold (ST) limits accuracy in detection of metastatic disease due to variability in normal uptake and partial volume effect of smaller lesions. We previously demonstrated a significantly lower normal SUV activity in the ribs and hypothesize that automated metastatic rib lesion detection could be improved if a lower regional threshold was applied. The purpose of this study is to determine the feasibility of a novel, rapid, reproducible and patient-individualized regional method of detecting and quantifying bone metastases based on patient and regional specific thresholds (RT).
Methods 18F-fluoride PET/CT scans (n=20 patients) were anonymized and transferred to a dedicated analysis workstation with commercially available software (MIM 6.4; MIM Encore). Workflows were then created for the RT and ST quantification methods in the following sequence. Skeletal osseous activity was identified based on SUV > 1.5, and separated into regions (head, thorax, ribs, spine, sternum and pelvis/lower extremities). For each patient, region-specific baseline values of normal osseous mean activity were obtained. The results were used to create thresholds that were then applied to each region. An experienced radiologist identified and removed the regions of degenerative joint disease (DJD) and urinary activity from analysis by visual inspection based on characteristic CT appearance. The remaining contours were used to measure metastatic lesion tumor volume (TV) and activity index (AI=TV x 18F-fluoride uptake). These measurements were compared to those obtained by a single threshold method (ST method; SUV > 10). Each patient was analyzed twice to evaluate reproducibility. Time required for analysis was noted.
Results Both the RT method and the ST method detected no metastases in 10 patients. The intra-observer percentage difference for patient (n=20) total fluoride bone volumes and total bone activity for the patient specific RT method were 2.2% and 4.8%; for ST method there was no difference. In 1 patient, RT detected a solitary metastasis that the ST method did not; physician reader visually confirmed this metastasis. Of the remaining 9 patients, RT and ST methods detected metastases in all patients. Four patients had low volume of metastatic disease (<6ml), which resulted in high variability of both TV and AI due to concurrent DJD, for the RT (47% and 48%) and for ST (24% and 22%) methods, respectively. Excluding these, the variability in TV and AI were relatively low for the RT (4.6% and 3.3%) and for the ST (2.9% and 2%, respectively) methods. In the 5 patients with physician reader visually confirmed multiple metastatic rib lesions, the patient specific RT method captured 28 rib metastases, while ST method captured 15 (p=0.04). Additional time for segmentation of the regional analysis was approximately 2-3 min/patient.
Conclusions With the exception of low volume disease with concurrent DJD, the patient and region specific threshold method is feasible as a rapid and reproducible method that could potentially improve the accuracy of NaF PET/CT skeletal tumor burden quantification.