3D prostate histology reconstruction: An evaluation of image-based and fiducial-based algorithms

摘要: Purpose: Evaluation ofin vivo prostate imaging modalities for determining the spatial distribution and aggressiveness of cancer ideally requires accurate registration images to an accepted reference standard, such as histopathological examination radical prostatectomy specimens. Three-dimensional (3D) reconstruction histology facilitates these registration-based evaluations by reintroducing 3D information lost during processing. Because accuracy may constrain clinical questions that can be answered with data, it is important assess tradeoffs between minimally disruptive methods based on intrinsic image potentially more robust extrinsic fiducial markers. Methods: Ex magnetic resonance (MR) digitized whole-mount from 12 specimens were used evaluate four algorithms. reconstructions computed registering each corresponding ex MR using one two similarity metrics (mutual or error) search domains (affine transformations a constrained subset thereof). The algorithms evaluated mean target error (TRE) homologous point landmarks (3–16 per section; 232 total) identified images, sensitivity TRE rotational, translational, scaling initialization errors. Results: The mutual had ± standard deviation TREs 0.7 0.4 1.2 mm, respectively, algorithm affine transforms negligible sensitivities errors. postoptimization values information-based metric showed evidence errors due both optimizer metric, variation parameters did not improve its performance. Conclusions: The fiducial-based lower than intensity-based information. A model relating statistical power certain validation study designs estimated mm would require 27% fewer subjects method initialize (mean 1.3 mm), suggesting choice technique have substantial impact design studies, their overall cost.