Rationale and objectives: High-radiofrequency energy is used clinically to ablate pathologic tissue with interventional magnetic resonance (MR) imaging. For many tissues, resulting lesions have a characteristic appearance on contrast-enhanced T1- and T2-weighted MR images, with two boundaries enclosing an inner hypointense region and an outer hyperintense margin. Geometric modeling of three-dimensional thermal lesions in animal experiments and patient treatments would improve analyses and visualization.
Materials and methods: The authors created a model with two quadric surfaces and 12 parameters to describe both lesion surfaces. Parameters were estimated with iterative optimization to minimize the sum of the squared shortest distances from segmented points to the model surface. The authors validated the estimation process with digital lesion phantoms that simulated varying levels of segmentation error and missing surface information. They also applied their method to in vivo images of lesions in a rabbit model.
Results: For simulated phantom lesions, the lesion geometry was accurate despite manual segmentation error and incomplete surface data. Even when 50% of the surface was missing, the median error was less than 0.5 mm. For all in vivo lesions, the median distance from the model surface to data was no more than 0.58 mm for both inner and outer surfaces, less than a voxel width (0.7 mm). The interquartile range was 0.89 mm or less for all data.
Conclusion: The authors' model provides a good approximation of actual lesion geometry and is highly resistant to missing segmentation information. It should prove useful for three-dimensional lesion visualization, volume estimation, automated segmentation, and volume registration.