A thresholding approach to improve metabolic volume assessment in PET/CT: Initial validation of methodology

Objectives It is common practice that an individualized threshold rather than a constant cutoff is used for assessing the tumor metabolic volume (MV). The study proposed a standardized methodology for accurate and reproducible MV determination at individualized cutoffs validated by phantom and patient PET/CT.

Methods Specific phantoms & 44 lung cancer patients were scanned on a HI-Res PET/CT. 4 approaches were used to set delineation of MV at cutoffs of: (I) 50%, (II) 2.5SUV, (III) background and (IV) individualized thresholds determined by automatic background subtraction and correction. Known phantom volumes (PV) and CT volumes (CTV) were used for MV evaluation. A target threshold (TT) at which MV=CTV was used as the 2^nd comparator.

Results Significant differences were found between MVs determined by different approaches (p<0.001). The proposed method (IV) was consistently more accurate and reproducible than other approaches. The comparison to PV & CTV by (I) led to underestimation of MVs, while it revealed better results by(II). Method (III) yielded the largest variability. Compared to TT, it revealed threshold deviation of only 4±3% by (IV) however significantly increased by other methods (average: 18±9%).

Conclusions The proposed method was shown to be superior in accuracy and robustness compared to current constant cutoff approaches. MV as a quantitative biomarker incorporating tumor metabolic activity might reflect overall disease burden more accurately than SUV and is of increasing importance to assess biological activities of therapies.