Methodology to register prostate B-mode and ARFI images to MR and histology
Acoustic Radiation Force Impulse (ARFI) imaging is being developed for guiding needle biopsy and focal therapy of Prostate cancer (PCa). In vivo ARFI images portray internal structures in the prostate with higher contrast than matched B-mode images. Given the heterogeneity of the prostate and the poor visualization provided by B-mode, another gold standard for determining what is being visualized in ARFI images is necessary. In this study, we present image registration techniques that facilitate correlation of in vivo ARFI, B-mode ultrasound (US), and Magnetic Resonance (MR) images obtained prior to radical prostatectomy with whole mount histology data. Pathology and structures were identified and segmented in the different datasets. The segmented datasets were used to form 3D mesh models of the prostate and the node and element information was extrapolated into a 3D image matrix of equivalent size for all modalities. Non-rigid registration of the different models was performed and the registered images were evaluated for co-localization of confirmed pathology. The methodology was validated using simulated prostate anatomy and finite-element techniques and found to improve the average displacement of registration markers by 76% in the MR simulation and 58% in the US simulation. When implemented on the patient data, the registration methodology was found to simplify multi-modality image comparison and analysis. Confirmed pathology was found to align with similarly suspicious regions in both ARFI and MR images. With its improved anatomical visualization over traditional B-mode imaging, ARFI holds promise for providing targeted image guidance of prostate focal therapy and needle biopsy. © 2011 IEEE.
