Elastic image registration for guiding focal laser ablation of prostate cancer: preliminary results

Abstract

Purpose: To guide ultrasound-driven prostate photodynamic therapy using information from MRI-based treatment planning.

Methods: Robust points matching (RPM) and thin plate splines (TPS) are used to solve correspondences and to map optimally positioned landmarks from MR images to transrectal ultrasound (TRUS) images. The algorithm uses a reduced number of anatomical markers that are initialized on the images.

Results: Both phantom and patient data were used to evaluate precision and robustness of the method. Mean registration error (±standard deviation) was of 2.18±0.25 mm and 1.55±0.31 mm for patient prostate and urethra, respectively. Repeated tests with different markers initialization conditions showed that the quality of registration was neither influenced by the number of markers nor to the human observer.

Conclusion: This method allows for satisfyingly accurate and robust non rigid registration of MRI and TRUS and provides practitioners with substantial help in mapping treatment planning from pre-operative MRI to interventional TRUS.