Intensity-Based Nonoverlapping Area Registration Supporting "drop-Outs" in Terms of Model-Based Radiostereometric Analysis

Abstract

A model-based radiostereometric analysis (MBRSA) is a method for precisemeasurement of prosthesis migration, which does not require marking the implantwith tantalum beads. Instead, the prosthesis pose is typically recoveredusing a feature-based 2D-3D registration of its virtual model into a stereo pairof radiographs. In this study, we evaluate a novel intensity-based formulation ofpreviously published non-overlapping area (NOA) approach. The registrationis capable to perform with both binary radiographic segmentations or non segmentedX-ray images. In contrast with the feature-based version, it is capableto deal with unreliable parts of prosthesis. As the straightforward formulationallows efficient acceleration using modern graphics adapters, it is possible toinvolve precise high-poly virtual models. Moreover, in case of binary segmentations,the non-overlapping area is simply interpretable, useful for indicating theaccuracy of the registration outcome. In silico and phantom evaluations wereperformed using a cementless Zweymüller femoral stem and its reverse engineered(RE) model. For initial pose estimates with difference from the ground-truth limited to 4 mm and 4° respectively, the mean absolute translationalerror was not higher than 0.042+-0.035 mm. The error in rotation around theproximodistal axis was 0.181+-0.265°, error for remaining axes was not higherthan 0.035+-0.037°.A model-based radiostereometric analysis (MBRSA) is a method for precisemeasurement of prosthesis migration, which does not require marking the implantwith tantalum beads. Instead, the prosthesis pose is typically recoveredusing a feature-based 2D-3D registration of its virtual model into a stereo pairof radiographs. In this study, we evaluate a novel intensity-based formulation ofpreviously published non-overlapping area (NOA) approach. The registrationis capable to perform with both binary radiographic segmentations or non segmentedX-ray images. In contrast with the feature-based version, it is capableto deal with unreliable parts of prosthesis. As the straightforward formulationallows efficient acceleration using modern graphics adapters, it is possible toinvolve precise high-poly virtual models. Moreover, in case of binary segmentations,the non-overlapping area is simply interpretable, useful for indicating theaccuracy of the registration outcome. In silico and phantom evaluations wereperformed using a cementless Zweymüller femoral stem and its reverse engineered(RE) model. For initial pose estimates with difference from the ground-truth limited to 4 mm and 4° respectively, the mean absolute translationalerror was not higher than 0.042+-0.035 mm. The error in rotation around theproximodistal axis was 0.181+-0.265°, error for remaining axes was not higherthan 0.035+-0.037°.