Comparison of manual versus semi-automatic quantification of skeletal tumor burden on 18F-Fluoride PET/CT

Objectives: Skeletal tumor burden parameters on ¹⁸F-Fluoride PET/CT images have shown prognostic value in prostate cancer patients. However, the calculation of the skeletal tumor burden is not routinely performed because the current existing tools render whole-body quantification a time-consuming process; furthermore, there are no available platforms for adequate reporting of quantification in busy clinical setting.This study aimed to correlate two quantification methods (manual vs semi-automatic) to calculate skeletal tumor burden on ¹⁸F-Fluoride PET/CT and additionally, to evaluate skeletal tumor burden with clinical outcomes.

Methods: A total of 107 female breast cancer patients that underwent ¹⁸F-Fluoride PET/CT whole-body image for investigation of bone metastases and were retrospectively reviewed. ¹⁸F-Fluoride PET/CT images demonstrating osteoblastic bone metastases were quantified blindly both manually and semi-automatically. For semi-automatic quantification, we used the Multi Foci Segmentation (MFS) tool in syngo.via MM Oncology (Clinical Applications, Oncology, Siemens ® Medical Solutions, Molecular Imaging, Hoffman Estates, IL USA Inc.). Skeletal tumor burden was calculated (by both methods) using the threshold SUVmax蠅10 to exclude normal bone. The semi-automatic method allowed easily the manual exclusion of abnormal sites not related to metastases (ex. urinary bladder, degenerative processes, etc). The skeletal tumor burden was defined as the total uptake of fluoride-avid bone metastasis (TLF₁₀) for each patient. The primary endpoint was overall survival (OS), established from the date of ¹⁸F-Fluoride PET/CT until death or last follow-up. Secondary aims were time-to-bone event (TTBE) and progression free survival (PFS). The tumor characteristics Ki-67, HER-2, histology and nuclear grades were correlated with the skeletal tumor burden.

Results: Bone metastases were present in 49 patients. The results presented by both quantification methods were highly correlated (intraclass correlation coefficient = 0.9). On multivariable analysis, TLF₁₀ calculated by both methods was significantly and independently associated with OS (manual quantification: p<0.0001; HR=1.136; 95%CI = 1.062 -1.216 vs semi-automatic quantification: p<.0001; HR = 1.315; 95%CI = 1.049 -1.648). Furthermore, TLF₁₀ calculated by both methods was also significantly and independently associated with PFS (manual quantification: p=0.0002; HR=1.120; 95%CI = 1.055-1.189 vs semi-automatic quantification: p = 0.0002; HR = 1.137; 95%CI = 1.033 - 1.229). Approximate times for calculating bone tumor burden manually and semi-automatically were, respectively, 30 minutes’ vs 30 seconds (in patients with less than six metastases); 60 minutes’ vs 2 minutes (in patients with >6 - 20 metastases; > 120 minutes’ vs 4 minutes (in patients with >20 - 50 metastases).

Conclusion: MFS semi-automatic calculation of bone tumor burden in breast cancer patients is practical and fast and provides the same information obtained with the time-consuming manual method. This tool might be an independent prognostic imaging biomarker. More studies are necessary to confirm our findings.

Research Support: None