Development of three-diimensional transrectal ultrasound for biopsy of the prostate.

摘要

Biopsy of the prostate is the sole clinical investigation capable of providing definitive diagnosis of prostate adenocarcinoma (PCa), or prostate cancer. Two-dimensional (2D) transrectal ultrasound (TRUS) is the current imaging modality used to guide biopsy needles into the prostate however, 2D TRUS suffers from poor visibility of early stage PCa and is confined to using 2D information to plan and guide biopsy needles to specific, non-visible biopsy targets within the prostate and record the needle locations. A three-dimensional (3D) TRUS prostate biopsy system was developed to provide biopsy planning, needle guidance and biopsy record all in three-dimensions. It is hypothesized that the 3D biopsy system will improve the biopsy accuracy over 2D TRUS, particularly in cases of repeat prostate biopsy, which might ultimately increase detection of PCa.The display of the 3D biopsy core locations during a biopsy procedure would provide verification that biopsy targets were adequately sampled and would allow for appropriate intra-procedure biopsy target modification. An automatic TRUS needle segmentation algorithm was developed specifically for the prostate biopsy procedure, to identify the location of the biopsy core for each biopsy. The needle segmentation algorithm demonstrated a > 99% accuracy in identifying the temporal TRUS image containing the biopsy needle at full insertion from a collection of real-time TRUS images throughout the insertion and withdrawal of the needle. The axis of the segmented biopsy needles were accurate to within 2.5 +/- 2.2° and 0.51 +/- 0.46 mm of the gold standard. Identification of the needle tip to within approximately half of the biopsy core length ( 10 mm) was 94% successful with a mean error of 2.7 +/- 5.6 mm.The complete 3D TRUS prostate biopsy system was assembled and the system accuracy was quantified in vitro by performing targeted biopsies on patient-specific prostate phantoms. The 3D biopsy system successfully reconstructed 3D prostate surface models with a mean volume error of -3.2 +/- 7.6% and the surface of the prostate was accurate to a mean error of 0.9 +/- 0.2 mm. Using the 3D TRUS biopsy system, biopsy needles were accurately guided to within 2.3 +/- 1.0 mm of pre-defined biopsy targets and the 3D location of the biopsy cores were recorded to within 1.5 +/- 0.8 mm.Finally, the accuracy of our 3D TRUS biopsy system was compared to both 2D TRUS and conventional 3D TRUS for performing repeat biopsy procedures. Repeat biopsies were performed by experts and radiology residents on a clinically-realistic prostate biopsy simulator composed of 3D TRUS images of 10 patient's prostates. The ability to biopsy a repeat biopsy target using 2D TRUS, conventional 3D TRUS and our developed 3D TRUS system was compared for all users. The comparative study demonstrated that our 3D TRUS biopsy system was more accurate (0.86 +/- 0 47 mm) than 2D TRUS (3.68 +/- 2.60 mm) for biopsy of the repeat targets, while the other conventional 3D TRUS (3.60 +/- 2.57 mm) did not significantly improve biopsy accuracy over 2D.A prostate surface reconstruction algorithm was developed to provide 3D context through a prostate surface model for the 3D TRUS biopsy system. The surface reconstruction algorithm conforms to the constraints of the current prostate biopsy procedure and can produce a 3D patient-specific prostate model without requiring major changes to the current procedural workflow. From a sparse sample of 2D TRUS images, 3D prostate models were accurately reconstructed to a mean surface boundary error of 1.34 +/- 0 20 mm and a volume error of 5.3 +/- 4.1%.Keywords. prostate biopsy, ultrasound, three-dimensional ultrasound, transrectal ultrasound, prostate cancer, surface reconstruction, needle segmentation, biopsy simulator.