Mieke Bus

90 Chapter 6 re-positioned 50mm distally in the ureter and a new pullback was performed until the com- plete renal pelvis and ureter were visualized. After each pullback, the probe position was recorded using co-registered 3D CT. Endoluminal Ultrasound (ELUS) Imaging ELUS images were recorded using the Volcano intravascular ultrasound Imaging system interfaced to the Revolution® 45MHz 3.5Fr (1.7mm) IVUS probe (Volcano Corporation, San Diego, California, USA) with an axial resolution of 200µm and an lateral resolution of 200-250µm. (10) Following introduction though the ureteral orifice, the imaging probe was advanced into the renal pelvis. Subsequently, ELUS images were acquired by retracting the imaging probe using the Spinvision® automatic pullback system (Volcano Corporation, San Diego, California, USA) at a pullback speed of 0.5mm/s, while rotationally acquiring 30 frames/s, across a trajectory of 90mm. This resulted in 5400-frame datasets obtained in 180 seconds. After each pullback, the imaging probe was manually re-positioned 80mm distally in the ureter and a new pullback was performed until the complete renal pelvis and ureter were visualized. After each pullback, the probe position was recorded using co-reg- istered 3D CT. Computed Tomography Imaging (CT) A Brilliance 64-slice CT scanner (Philips Medical System; Best; The Netherlands) was used to make CT‑scans of each of the nephroureterectomy specimens. The scans were made at a tube voltage of 120 kV and tube charge of 22 mAs. The total detector collimation was 64x0.625 mm. The filter used in the CT-reconstruction was a medium smooth filter (filter B) and the final voxel spacing was 0.4x0.4x0.7 mm. Image Reconstruction, 3D rendering and deformable manual CT based co-registration CT datasets (DICOM) each depicting a different OCT or ELUS probe position were loaded into the 3D visualization software. An AMIRA embedded, automatic, rigid co-registration algorithm was used to accurately align the datasets in three dimensions. For separate depic- tion of 1) the probe, 2) kidney and ureter and 3) peri-renal fat, Hounsfield value (HU) based automatic segmentations were performed on the CT data. Subsequently, corresponding OCT and ELUS datasets (TIFF stacks) were loaded, displayed as 3D-volume reconstruction, and visually aligned to the probe position as seen on CT, using the probe tip as the dataset’s starting point. Rotational orientation of the OCT/ELUS datasets was determined based on mutual visible image features in both imaging modalities, such as lumen contour and air bubbles. To correct for the curvature of the probe (and ureter) and to facilitate registration with pathologic slides, manual 3D deformation of the OCT/ELUS datasets was performed. For this reason, the individual OCT/ELUS datasets were separated