Improvement of Hepatic Lesion Characterization by 18F-FDG PET/CT with the Use of the Lesion to Background Liver Activity Ratio

OBJECTIVE: This study aimed to develop a reproducible method to improve (18)F-FDG PET/CT's diagnostic accuracy when characterizing hepatic lesions. METHODS: This was a retrospective study of 76 patients with various types of cancer who had 134 liver lesions described on diagnostic CT, MRI, or (18)F-FDG PET/CT. All 134 lesions were imaged by PET/CT. For comparison, all 134 lesions were also imaged either by MRI (58 lesions) or by diagnostic CT (107), with some imaged by both. The original reader's interpretation was used to determine the diagnostic validity of each modality. The (18)F-FDG PET/CT findings were then re-stratified using the ratio of the SUV(max) of the liver lesion to the SUV(avg) of liver background (SUV(TLR)). The standard of reference was histopathology and/or imaging follow-up. Eighty-nine out of the 134 lesions were positive for metastases. Two-tailed P values were calculated using a chi-square test. RESULTS: A SUV(TLR) greater than 2.0 strongly correlated with metastasis, while less than 2.0 strongly correlated with benign activity. Based on the original interpreter's visual analysis (PET(VA)), (18)F-FDG PET/CT had an accuracy of 74.6% with a high sensitivity (95.5%), but limited specificity (33.3%). Using the SUV(TLR) cutoff value of 2.0 to objectively re-stratify these cases, there was improvement in accuracy to 97.8%, sensitivity to 100%, and specificity to 93.3%. These values were statistically significant by chi-square analysis, with P = 0.0001 for both PET(VA) and SUV(TLR). The 58 MRI lesions and 107 diagnostic CT lesions that were reviewed had accuracies of 70.7% (P = 0.008) and 73.8% (P = 0.0001), respectively. CONCLUSION: Using a cutoff value of 2.0 for SUV(TLR) to objectively evaluate for potential hepatic metastases on (18)F-FDG PET results in a significant improvement in diagnostic accuracy over subjective visual analysis alone. In this limited study, use of SUV(TLR) more accurately characterized hepatic lesions than the original CT or MRI interpretations.