Mieke Bus

6 91 into 50-frame segments and manually aligned perpendicular to the CT based probe seg- mentation in three dimensions. After alignment of the individual segments, recombination of the segments into a new, 3D deformed dataset was conducted. This process was repeated for all datasets to aim for a virtually complete 3D reconstruction of the ureter based on OCT and ELUS data only. A B C D E F G H Figure 2: Co-registration steps as executed in AMIRA. A: Create and visualize segmented centerline of OCT probe in CT data, B: Read in OCT volume, C: Split total bounding box volume into 10 bounding boxes, D: Manually place individual OCT bounding boxes perpendicular to centerline probe in Y,Z plane, E: Manually rotate each individual OCT bounding box in X,Y plane to fit inner ureter shape, F & G: Merge individual OCT bounding boxes, H: Visualize fused OCT – CT dataset Pathology preparation The standard pathological report of nephroureterectomy specimens was considered the ref- erence standard for comparison with OCT/ELUS imaging. Nephroureterectomy specimens dissected and examined at the pathology department according to a standardized protocol. Image analysis and matching with pathology. OCT and ELUS datasets were reviewed slice by slice for identification of ureter wall archi- tecture, focusing on identification of ureter wall layers consisting of urothelium, lamina pro- pria, muscularis propria and periureteral fat. Ureteral abnormalities seen as visible lesions were divided in either non-invasive tumors if underlying ureteral wall layer architecture was still visible, or invasive tumors when the wall architecture was interrupted or no visible underlying architecture was present. Images were scored inconclusive when imaged lesion transcended the field of view. Matching of histopathology with the corresponding regions of OCT/ELUS was achieved based on available histopathology slides with sufficient ureter tumor architecture and information provided by the uro-pathologist.