Abstract
PURPOSE: Ellipsoid tumour models are used to approximate the tumour volume of uveal melanomas, as the conventionally used ultrasound does not provide a three-dimensional visualization of the tumour. However, these models are a simplification of the actual tumour geometry. The aim of this study was to determine to what extent several of these frequently used ellipsoid tumour models accurately describe uveal melanoma volume. METHODS: Tumours were delineated on contrast-enhanced T1-weighted MRI for 70 uveal melanoma patients. The MRI-delineated volume was compared with three ellipsoid models, which used two-dimensional measurements such as thickness and basal diameters as input: half ellipsoids with round (V(roundbase)) or oval base (V(ovalbase)) and a paraboloid consisting of two parts, also incorporating the curvature of the eye wall (V(twoparts)). RESULTS: Statistically significant relative differences between MRI-delineated and model volume of 53 ± 32% (V(roundbase)), 26 ± 24% (V(ovalbase)) and 15 ± 24% (V(twoparts)) were observed (p < 0.001). Tumour volume and shape did not influence the difference between the model volumes and MRI-delineated tumour volume. CONCLUSION: All tumour models result in considerable systematic overestimations of tumour volume, with large variations in overestimation between patients. Adding the perpendicular basal diameter to the model decreases this variation. Although ellipsoid tumour models have been shown to be valuable on a group level, they should be used with caution for individual patients.