Author of

• 25936

  +

  P1.16 - Treatment of Early Stage/Localized Disease (Not CME Accredited Session) (ID 948)

  • Event: WCLC 2018
  • Type: Poster Viewing in the Exhibit Hall
  • Track:
  • Presentations: 1
  • Moderators:
  • Coordinates: 9/24/2018, 16:45 - 18:00, Exhibit Hall

  • 26723

    +

    P1.16-16 - Automatic Intratumor Segmentation in CT of NSCLC: An Alternative to PET Metabolic Subregions (ID 12188)

    16:45 - 18:00  |  Author(s): Yong Yin
    • Abstract

    Background
    

    PET images provide heterogeneous metabolic information in precision radiation treatment planning and the radiation dose given to high metabolic volumes should be escalated. However, PET scanning will increase the radiation dose received by patients. The aim of this study was to evaluated the feasibility of automatic intratumor segmentation in CT of NSCLC patients based on level-set evolution and cell automaton algorithm and assess the consistency with PET metabolic subregions.

    a9ded1e5ce5d75814730bb4caaf49419 Method
    

    The PET and plan-CT imaging data set of 17 patients who have diagnosed with NSCLC were randomly collected. First, the gross tumor volume (GTV) was defined using a threshold of 40% SUV_max on PET. Then, rigid registration was used to align PET images to plan-CT images and the GTV was mapped to the CT image subsequently. The subregions which describe heterogeneity of voxel gray-level intensities were then automatic segmented using an algorithm combined level-set evolution and cell automaton in GTV_CT. Meanwhile, the three metabolic subregions in GTV_PET were delineated using threshold interval a) 40%-60% SUV_max, b) 60%-80% SUV_max, and c) 80%-100% SUV_max. To evaluate the consistency with PET metabolic subregions, we calculated the spatial overlap by Dice’s similarity coefficient (DSC).

    4c3880bb027f159e801041b1021e88e8 Result
    

    

    In total, 21 GTV pairs acquired from CT and PET data set were used to evaluate the feasibility of method proposed in this study. The GTV_CT was automatically divided into three heterogenous subregions based on its difference on grays-level intensities and regional connectivity of voxel and 63 subregions were acquired. The average DSC value calculated from subregion in CT and PET is 0.725 with interval (0.321,0.905).

    8eea62084ca7e541d918e823422bd82e Conclusion
    

    Our study revealed that there is a significant correlation between PET metabolic information and CT gray-level intensity. The automatic segmentation of subregion in CT images may can serve as an alternative to the metabolic region delineated in PET.

    6f8b794f3246b0c1e1780bb4d4d5dc53