부산대학교 도서관

The immunohistochemical determination of the Ki67 expression is regarded as a reliable method to monitor nonsmall cell lung cancer (NSCLC) proliferation. However, cancer tissues are not uniform and the use of a small tumor sample in biopsy may not be accurate. Although positron emission tomography (PET) represents an alternative to biopsy, by providing the 3-D functional and morphological distribution of cancer cells across the entire tumor volume, standardized uptake values (SUVs) exhibit poor accuracy. The purpose of this study is to identify quantitative features in 2-[18F]FDG PET images uniquely correlated with proliferating lung cancer cells that may perform better than SUVs in the evaluation of lung cancer proliferation. By using a xenograft NSCLC model implanted in 19 mice, we found that 89 textural parameters extracted from 3-D 2-[18F]FDG uptake can be described by 6 independent principal components accounting for 95% of the total variance. One of these components features a significant and univocal correlation with 3-D-stacked Ki67 staining immunohistochemical images ( $|r|>0.6; P< 0.05/6$ , including compensation for multiple comparisons). Furthermore, the values of these components extracted from the PET images of in vivo and ex vivo explanted NSCLC cancers exhibit a strong correlation, verifying the stability of the segmentation method.