Multi-level otsu method to define metabolic tumor volume in positron emission tomography

This study was to validate reliability and clinical utility of a PET tumor segmentation method using multi-level Otsu (MO-PET) in standard National Electrical Manufacturers Association (NEMA) image quality (IQ) phantom and patients with osteosarcoma. The NEMA IQ phantom was prepared with a lesion-to-background ratio (LBR) of either 8:1, 4:1, or 1.5:1. The artificial lesions in the phantom were segmented using MO-PET, gradient-based method (PETedge), relative threshold methods, and background threshold methods. Metabolic tumor volumes (MTVs) using MO-PET and PETedge were named as MTV (MO-PET) and MTV (PETedge), respectively. Among the MTVs using multiple methods, only MTV (MO-PET) and MTV (PETedge) showed excellent agreements with the actual volume of NEMA IQ phantom across the different LBRs (intraclass correlation coefficient, ICC = 0.987, 0.985 in LBR 8:1, 0.981, 0.993 in LBR 4:1 and 0.947, 0.994 in LBR 1.5:1). Repeated measurements of MTV (MO-PET) of the primary tumors showed excellent reproducibility with ICC of 0.994 (0.989-0.997) in patients with osteosarcoma. Also, MTV (MO-PET) was found to be predictive of Event Free Survival (EFS) [Hazard ratio (95% CI) = 6.1 (2.1-17.2), log rank P = 0.0003] in patients with osteosarcoma. We have validated in NEMA IQ phantom that the MTV (MO-PET) is accurate, and importantly, stable and consistent across a range of lesion sizes and LBRs representative of clinical tumor lesions. Furthermore, MTV (MO-PET) showed excellent reproducibility and was predictive for EFS in patients with osteosarcoma.