Premium
Sci‐Thur PM Therapy‐08: Intra and Inter Observer Variability and Systematic Error in Prostate Delineation
Author(s) -
Gerig L,
Gao Z,
Eapen L,
Wilkin D,
Morash C,
Wassef Y
Publication year - 2006
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.2244622
Subject(s) - prostate , medicine , gold standard (test) , prostate cancer , population , radiation therapy , nuclear medicine , standard deviation , medical physics , radiology , statistics , mathematics , cancer , environmental health
Individual sub‐processes of external beam radiation therapy can affect patient outcome. Specifically geometric uncertainties play a major role in tumor control and normal tissue complications. It is known that tumor definition varies among different observers but importantly, at least to our knowledge, few studies have reported their findings against a true gold standard, most relying on the population average as the reference value. This work uses the Visible Human Project® data for a set of cross‐indexed CT and anatomical photographic digital images of a human male cadaver. We use the anatomical images to define a gold standard for the prostate as defined by a group of experts. Six radiation oncologists then repeatedly (20 times over several weeks) defined the prostate alone on the corresponding CT images allowing us to quantify inter and intra observer random and importantly to measure systematic error. Individual and population means were tested for systematic error against the gold standard. We found that the observers routinely over estimated the prostate volume, and that our intra and inter observer variability is not significantly different than that reported in the literature. Significantly we report a systematic error in target definition, with the physicians failing to include all of the posterior prostate volume near the rectum. In contrast the observers rarely missed true prostate volume in the left, right or anterior quadrants, but we observe a systematic error in that normal tissue anterior to the prostate was routinely included as target tissue.