Date of Award
Master of Science
Epidemiology and Biostatistics
Dr. George B. Rodrigues
The purpose of the study was to assess the accuracy, validity and time-savings of automated atlas-based segmentations (AABS). Further, the study assessed the inter- and intra observer variability in the delineation of the prostate bed (PB) and the five regions of interest for postoperative conformal radiotherapy in prostate cancer patients. Finally, the study reports on the development of an appropriate methodology for similar studies. Seventy-five DICOM Computed Tomography (CT) datasets were obtained to create the prostate bed atlas and another five datasets were retrospectively contoured by the atlas builder, the expert panel and the AABS tool. Consensus segmentations (CS) were also generated. The mean dice similarity coefficient comparing the edited AABS and CS was 0.67, 0.88, 0.93, 0.92, 0.54 and 0.78 for the PB, bladder, left- and right femoral head, penile bulb and rectum, respectively. Significant inter observer variation was observed in the PB and bilateral femoral heads. Significant time savings were obtained using the average AABS editing time (p = 0.003) versus the manual contouring time. We successfully developed a methodology and validated the AABS tool for routine clinical use.
Hwee, Jeremiah, "METHODOLOGY TO ASSESS AUTOMATED ATLAS-BASED SEGMENTATIONS AND INTER- AND INTRA-OBSERVER VARIABILITY IN THE DELINEATION OF THE PROSTATE BED" (2011). Digitized Theses. 3424.