Purpose: To validate semiquantitative analysis of positron emission mammography (PEM).
Methods: Fifty women with histologically confirmed breast lesions were retrospectively enrolled. Semiquantitative uptake values (4 methods), the maximum PEM uptake value (PUVmax), and the lesion-to-background (LTB) value (3 methods) were measured. LTB is a ratio of the lesion's PUVmax to the mean background; LTB1, LTB2, and LTB3 (which were calculated on different background) were used to designate the three values measured. Interobserver reliability between two readers for PUVmax and the LTBs was tested using the interobserver correlation coefficient (ICC). The likelihood ratio test was used to evaluate the relationship between ICCs. Receiver operating characteristic (ROC) curves were calculated for all methods. Diagnostic accuracy in differentiating benign tissue from malignant tissue was compared between PUVmax and LTB1.
Results: The ICC rate was 0.971 [95 % confidence interval (CI) 0.943-0.986] for PUVmax, 0.873 (95 % CI 0.758-0.935) for LTB1, 0.965 (95 % CI 0.925-0.983) for LTB2, and 0.895 (95 % CI 0.799-0.946) for LTB3. However, there were some technical difficulties in the practical use of LTB2 and LTB3. The likelihood ratio test between PUVmax and LTB1 was statistically significant (p < 0.001). ROC curves of the 4 methods had similar characteristics. The median PUVmax was 1.39 for benign lesions and 3.70 for malignant lesions. LTB1 was 1.92 for benign lesions and 4.78 for malignant lesions. Significant differences (p < 0.001) in both PUVmax and LTB1 were observed between groups.
Conclusion: Due to its simplicity and reproducibility, PUVmax is superior to LTB as an indicator for PEM in semiquantitative analysis.