Double reading for gross tumor volume assessment in radiotherapy planning

Abstract

Background/Objective: The precise definition of the gross tumor volume (GTV) that takes into account intra-and interobserver variability is necessary for high-precision radiotherapy (RT) techniques. The purpose of this study was to demonstrate the practical GTV assessment by a "double reading" approach.Methods: Pretreatment magnetic resonance (MR) imaging, including the post-contrast 3D magnetization-prepared rapid-gradient echo (MP-RAGE) sequence (section thickness 1.0 mm) was performed on a 3T superconducting imager in 50 patients with glioblastoma. MR images were transferred to a RT planning system (RTPS) that provides many opportunities for GTV contouring, e.g., at diagnosis, surgical navigation, and RT deliberations. Independent 2 observers preliminarily contoured the GTV on MR images. After planning-CT scanning, CT images with a 1.0 mm slice interval were transferred to the RTPS, registered with the diagnostic images, and then the preliminarily-contoured strictures were copied onto the CT images and used for GTV assessment. The practical GTV on the planning CT was determined by integrating the interpretations and adding information on postoperative changes. The interobserver variability in GTV contouring was assessed by Bland-Altman analysis and the concordance index.Results: There was substantial interobserver variability in GTV contouring (95% limits of agreement: -29.4%, 16.8%). The mean interobserver concordance rate for the GTV was 82.1% (range 56.5-91.2%). The practical GTVs were significantly larger than the preliminarily-contoured GTVs by both observers (p < 0.01).Conclusions: Considering interobserver variability, "double reading" is necessary for practical GTV assessment. This approach for volume assessment may facilitate the standardization of treatments, not only of RT but also of surgery and chemotherapy.

New approaches to the analysis of eye movement behaviour across expertise while viewing brain MRIs

“…Where studies have looked at the similarities or agreement in observers, they have tended to use brain tumour delineation tasks. Low inter- and intra-observer variability is well documented in clinicians delineating brain tumours (Crowe, Alderson, Rossiter, & Kent, 2017 ; Mazzara, Velthuizen, Pearlman, Greenberg, & Wagner, 2004 ; Murakami, Hirai, Toya, Nakamura, & Yamashita, 2012 ; Weltens et al, 2001 ). This line of research highlights the need to obtain high agreement in tasks relevant to medical image perception to assure consistency both within and between clinicians responsible for making diagnostic decisions.…”

Expertise Affects Inter-Observer Agreement at Peripheral Locations within a Brain Tumor

“…Low inter-observer agreement was also observed amongst nine physicians delineating five patients’ supratentorial inoperable brain tumors on CT scans (range 38–59%) and on CT combined with MRI scans (range 38–71%; Weltens et al, 2001 ). Murakami et al (2012) revealed an average concordance rate of 82% for one radiation oncologist and one neuroradiologist delineating glioblastomas using diagnostic MR images whilst Mazzara et al (2004) reported an average concordance rate of 28% for three radiation oncologists outlining gliomas on MR images. Researchers have also revealed high intra-observer variability with Mazzara et al (2004) reporting only a 20% concordance rate for three radiation oncologists delineating tumors at three 1-month intervals.…”

“…The area on which two clinicians agree (i.e., they draw around the same tissue) is commonly referred to as the intersection area. A concordance rate (note: researchers refer to this value differently but it reflects the same calculation) ranging from 0% (complete disagreement) to 100% (complete agreement) can be calculated by dividing the intersection area by the sum of the area contoured by both clinicians ( Murakami et al, 2012 ) to index inter-observer agreement. Leunens et al (1993) reported that the concordance rate of 12 radiation oncologists delineating 5 patients’ brain tumors on lateral orthogonal radiographs ranged from 25 to 73%.…”