The cross-sectional areas of SF and VF were quantified from images of serial CT abdomen/pelvis scans at a superior and at an inferior L3 vertebra level using a fat segmentation software (Zhao 2006)15 that separated SF from VF by the user-demarcated AM boundary (Fig. 3D). The superior and inferior values for each scan were averaged for an L3 estimate at that timepoint.
Pearson’s (r) and Spearman’s (ρ) coefficients were calculated for correlations among each of the tissue types for the raw cross-sectional area data. Correlation matrices were generated with these coefficients (Tables 2 and 3). Different timescales were used for the tissue types, as compared to the tumor burden, to assess for possible lag in time that may delay a change in one tissue with respect to the change in another tissue type (Table 4). Conditional formatting for the highest values in pale yellow and the lowest values in bright green was applied for green-to-yellow coloration using Microsoft Excel to observe for associations between the outcomes of different tissue types.
(To view a larger version of Table 2, click here.)
(To view a larger version of Table 3, click here.)
(To view a larger version of Table 4, click here.)