Deformable registration between MRI/MRSI and ultrasound images for targeted robotic prostate biopsy

Development of a prostate biopsy robotics can alleviate urologists\´ labor and ensure high accuracy. However, intervention from urologists is still needed to define the target cores of biopsy. It is being recognized that magnetic resonance imaging (MRI) and magnetic resonance spectroscopy imaging (MRSI) are potential to provide cancer distribution preoperatively. So guiding the biopsy needle towards those suspected cancer sites could improve its detection rate and reduce the possibility of false negative results. In this paper, a deformable registration method is presented for the purpose of merging MRI/MRSI and transrectal ultrasound (TRUS) images. Considering the ultrasound (US) image quality and the cross-modal deformation, a framework including a global rigid alignment and a nonrigid transformation is established to match the two prostate surfaces and thereafter their volumes. For validation purpose, a deformable prostate phantom that simulates the human condition was set up. Fifteen fiducial markers were implanted inside the phantom prostate to act as the "ground truth". The phantom study shows that our method can achieve an accuracy around 1.28±0.50mm, under voxel dimension of 0.5mm×0.5mm×0.5mm. This result was promising for blind knowledge about the interior prostate was utilized in the algorithm. Experimental result on patient data was also demonstrated.