Evaluation of Articulated Registration and Application to Metastatic Bone Lesions Identification
S Yip*, C Morrison, S Perlman, G Liu, R Jeraj, University of Wisconsin, Madison, WITU-A-WAB-6 Tuesday 8:00AM - 9:55AM Room: Wabash Ballroom
Accurate skeleton registration is required to identify corresponding metastatic bone lesions over multiple scans to assess therapeutic responses of individual lesions. In articulated registration, whole-body skeletons are registered by auto-segmenting individual bones, then rigidly aligning them. We evaluated if articulated registration outperforms other registration algorithms in skeleton registration and lesion identification.
Six prostate cancer patients with 128 total bone lesions were treated with molecular targeted therapy. Serial whole-body [¹⁸F]NaF PET/CT images were acquired throughout treatment. Skeletons were extracted from the CT images by thresholding (HU>150). Serial skeletons were registered using rigid, deformable, and articulated registration algorithms. The registration performance was quantified as the overlapped volume (OV) of the registered skeletons. The corresponding lesions were contoured and identified on serial PET images under the guidance of a nuclear medicine physician. Each algorithm was evaluated for its ability to accurately align corresponding lesions via skeleton registration. A lesion matching score (LMS) was measured for each lesion, which quantified the percent overlap between a lesion's two corresponding contours.
Articulated (OV=90+/-1%) and deformable registration algorithms (OV=90+/-1%) outperformed rigid registration algorithm (OV=50+/-7%) in the skeleton registration as the rigid registration algorithm failed to capture the skeleton flexibility in the joints. The performance of articulated registration (LMS=75+/-23%) in lesion identification was also significantly better than rigid (LMS=59+/-32%, p=0.018) and deformable (LMS=46+/-39%, p=0.0003) registration algorithms. Despite superior skeleton registration performance, deformable registration algorithm failed to preserve the local rigidity of bones leading to undesirable lesion deformation and poor performance in lesion identification, leaving articulated registration as the best choice.
Articulated registration is superior to rigid and deformable registrations by capturing global flexibility while preserving local rigidity inherent in skeleton registration. Articulated registration facilitates treatment response assessment of individual bone lesions by potentially reducing physician workload and error in lesion identification throughout treatment.