Virtual Library

Start Your Search

J. Jochem

Author of

  • +

    MO23 - Radiotherapy II: Lung Toxicity, Target Definition and Quality Assurance (ID 107)

    • Event: WCLC 2013
    • Type: Mini Oral Abstract Session
    • Track: Radiation Oncology + Radiotherapy
    • Presentations: 1
    • +

      MO23.07 - Impact of a gradient-based FDG-PET auto-contouring method on non-small cell lung cancer delineation (ID 1993)

      10:30 - 12:00  |  Author(s): J. Jochem

      • Abstract
      • Presentation
      • Slides

      Manual target volume delineation using CT/FDG-PET is the standard method used for radiotherapy treatment planning of non-small cell lung cancer (NSCLC) patients. Since manual delineation is prone to inter-observer variability and is time consuming, many FDG-PET auto-contouring methods were proposed in literature. The purpose of this study was to investigate to what extent a gradient-based FDG-PET auto-contouring method reduces observer variation, reduces delineation time and influences delineation behavior in radiotherapy treatment planning for NSCLC patients.

      Seven radiation oncologists (observers) dedicated to lung cancer treatment delineated the primary tumor (PT) and involved lymph nodes (LN) for 10 patients with stage IIA-IIIB NSCLC on a co-registered CT/FDG-PET scan. The study was separated in two phases. In the first phase, the observers manually delineated the PT and LN for all patients. For the second phase (four months later), auto-contours were generated for both the PT and LN using a gradient-based FDG-PET segmentation method. Bone and air tissue were removed from these auto-contours using CT thresholding. These auto-contours were provided as initial delineation and were adapted by the observers. Delineation times, delineated contours and agreement with the auto-contour were analyzed. Delineated contours were analyzed based on volume, the ratio between the common volume and the encompassing volume (C/E), Dice Index (DI), local standard deviation (SD) and the local distance between median surface and delineated surface. Regions were identified where the observers did or did not change the provided auto-contours.

      The observers agreed with the provided auto-contour for 37.3% of the PT and for 42.6% of the LN. Notable regions of agreement were the tumor/bone and tumor/air interfaces. The mean delineation time was reduced by 23.9% from 25.5 minutes in phase 1 to 19.4 minutes for phase 2 (p=0.000). The mean delineated volume was smaller in phase 2 compared to phase 1: 8.9% for the PT (155.8 to 142.0 cm[3], p=0.000) and by 9.1% for the LN (13.2 to 12.0 cm[3], p=0.001), respectively. The C/E ratio and DI both did not change significantly and were 0.79 and 0.88 for the PT and 0.54 and 0.67 for the LN in both phases. The mean local SD for the PT was 1.7 mm and 1.5 mm and for the LN was 1.5 mm and 1.4 mm and both did not change significantly, for both phases respectively. The mean distance between the median surface and PT delineations was slightly reduced from 2.1 to 1.8 mm for phase 2, and was 2.0 mm for the LN in both phases.

      The gradient-based FDG-PET auto-contouring method reduced delineation time by 24%, but was sufficient in only 37.3% of the primary tumors and 42.6% of the involved lymph nodes; most notably at the tumor/bone and tumor/air interfaces segmented using the CT scan. The results suggest the FDG-PET auto-contour is currently primarily used for localization, and not so much for delineation. Multi-modal auto-contouring has the potential to reduce inter-observer variation when further developed in close collaboration with radiation oncologists.

      Only Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login, select "Add to Cart" and proceed to checkout. If you would like to become a member of IASLC, please click here.

      Only Active Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login or select "Add to Cart" and proceed to checkout.