OBJECTIVE: The purposes of this study were to construct a model for estimation of splenic volume from standardized one-dimensional diameters of the spleen and to compare that model with the ellipsoid model for estimation of splenic volume. MATERIALS AND METHODS: In this retrospective study, segmentation software was used for semiautomated quantification of splenic volume by counting CT voxels in 193 consecutively registered patients. For standardization of one-dimensional measurements, the software was used to measure transaxial diameter in the slice with the largest splenic cross-sectional area. By incorporation of splenic volume and the product of width, thickness, and length into the linear regression equation, a model for estimation of splenic volume was constructed, and its performance was externally assessed. Splenic volume also was calculated with the formula for a prolate ellipsoid. The ellipsoid volume and best-fit volumes were compared with segmented splenic volume by use of Bland-Altman plot and Lin concordance correlation. A value of p < 0.05 denoted statistical significance. RESULTS: Splenic width was the best one-dimensional predictor of splenic volume (r = 0.84, p < 0.05). The linear regression fitted model for estimation of splenic volume (V(R)) in the initial 100 patients was V(R) = (0.36 × W × T × L) + 28, where W is width, T is thickness, and L is length (R(2) = 0.91, p < 0.05) and was externally validated by estimation of splenic volume in the other 93 patients. Compared with that observed with use of the ellipsoid formula, mean bias decreased from 22.57% to 0.93%, and the Lin coefficient increased from 0.81 to 0.96 with application of the best-fit model for calculation of splenic volume. CONCLUSION: The best-fit model V(R) = (0.36 × W × T × L) + 28 is more optimized than the ellipsoid formula and is associated with less bias for estimation of splenic volume.
OBJECTIVE: The purposes of this study were to construct a model for estimation of splenic volume from standardized one-dimensional diameters of the spleen and to compare that model with the ellipsoid model for estimation of splenic volume. MATERIALS AND METHODS: In this retrospective study, segmentation software was used for semiautomated quantification of splenic volume by counting CT voxels in 193 consecutively registered patients. For standardization of one-dimensional measurements, the software was used to measure transaxial diameter in the slice with the largest splenic cross-sectional area. By incorporation of splenic volume and the product of width, thickness, and length into the linear regression equation, a model for estimation of splenic volume was constructed, and its performance was externally assessed. Splenic volume also was calculated with the formula for a prolate ellipsoid. The ellipsoid volume and best-fit volumes were compared with segmented splenic volume by use of Bland-Altman plot and Lin concordance correlation. A value of p < 0.05 denoted statistical significance. RESULTS: Splenic width was the best one-dimensional predictor of splenic volume (r = 0.84, p < 0.05). The linear regression fitted model for estimation of splenic volume (V(R)) in the initial 100 patients was V(R) = (0.36 × W × T × L) + 28, where W is width, T is thickness, and L is length (R(2) = 0.91, p < 0.05) and was externally validated by estimation of splenic volume in the other 93 patients. Compared with that observed with use of the ellipsoid formula, mean bias decreased from 22.57% to 0.93%, and the Lin coefficient increased from 0.81 to 0.96 with application of the best-fit model for calculation of splenic volume. CONCLUSION: The best-fit model V(R) = (0.36 × W × T × L) + 28 is more optimized than the ellipsoid formula and is associated with less bias for estimation of splenic volume.
Authors: Adriano R Tonelli; Ruchi Yadav; Amit Gupta; Andrea V Arrossi; Gustavo A Heresi; Raed A Dweik Journal: Respiration Date: 2012-08-03 Impact factor: 3.580