1 0.82 / 0.81 / 0.85 0.73 / 0.73 / 0.78 0.72 / 0.79 / 0.82 195 2 0.79 / 0.82 / 0.87 0.77 / 0.80 / 0.85 0.60 / 0.65 / 0.75 94 3 0.85 / 0.88 / 0.89 0.66 / 0.73 / 0.76 0.85 / 0.88 / 0.89 55 4 0.88 / 0.90 / 0.90 0.90 / 0.91 / 0.90 0.63 / 0.67 / 0.69 63 5 0.88 / 0.90 / 0.91 0.89 / 0.92 / 0.91 0.66 / 0.72 / 0.79 92 6 0.92 / 0.90 / 0.90 0.90 / 0.90 / 0.89 0.72 / 0.69 / 0.57 104 7 0.84 / 0.85 / 0.85 0.79 / 0.80 / 0.81 0.73 / 0.73 / 0.72 65 8 0.89 / 0.89 / 0.90 0.87 / 0.86 / 0.87 0.73 / 0.75 / 0.76 41 Table 1: DSC between the structures in the intra-procedural and registered diagnostic MRI (rigid / affine / B-spline registration steps), and the registration execution time. Results Image registration was successfully accomplished in each of the 8 cases within time compatible with the clinical protocol. The average registration computation time was 2 min. For the purposes of registration, contouring of the prostate in each of the cases was completed by a computer scientist with 3 hours of prior training in radiologic appearance of the prostate in MRI. The average time required for the preparation of these approximate manual contourings of the prostate was within 2 min in each of the cases. For the evaluation of registration, accurate TG, PZ and CG contours were prepared by an abdominal radiologist with over 10 years of experience, which required on average 10 min per case. Contouring was done using the 3D Slicer software (1). Transformations produced by the registration were applied to the multiparametric diagnostic images and to the planned biopsy targets identified before the procedure. The quality of the registration was examined visually by the interventional radiologist and found to be acceptable. Quantitative assessment showed that in most cases the accuracy of alignment of TG, CG and PZ segmented by the expert improved between the rigid and affine, and between affine and b-spline registration steps. In the cases where no ERC was used (1 and 7), the improvement was less prominent. No improvement due to deformable registration was observed in Case 6, where the difference in the shape of the gland was not significant due to the presence of rectal filling during the biopsy procedure. On average, registration resulted in 0.88 DSC and 4.4 mm 95% HD for the total gland contour after registration. Discussion and Conclusions We presented a novel approach to the registration of diagnostic T2w MRI with the intra-procedural MRI for improved targeting during MR-guided transperineal PCa biopsy, and its preliminary evaluation on 8 cases. Our approach allows to register the images within the constraints of the clinical workflow, and facilitates fusion of the diagnostic data for improved PCa targeting. The advantages of the developed methodology over the surface-based registration methods is in its ability to leverage the intensity information throughout the whole prostate gland, which enables improved alignment of the features corresponding to the internal structures, as demonstrated in Fig.1. Unlike the previously developed intensity-based approaches to recover similar prostate deformations (4), we significantly reduced the involvement of the operator: once the approximate contours of the prostate gland are identified, our method is fully automatic and does not require further interaction. Additionally, we improved the robustness of the registration procedure by incorporating inhomogeneity correction and automatic initialization into the registration process. Acknowledgments