Lung Tumor Segmentation Using Electric Flow Lines for Graph Cuts

Christian Hollensen, George Cannon, Donald Cannon, Søren Bentzen, Rasmus Larsen

    Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

    328 Downloads (Pure)


    Lung cancer is the most common cause of cancer-related death. A common treatment is radiotherapy where the lung tumors are irradiated with ionizing radiation. The treatment is typically fractionated, i.e. spread out over time, allowing healthy tissue to recover between treatments and allowing tumor cells to be hit in their most sensitive phase. Changes in tumors over the course of treatment allows for an adaptation of the radiotherapy plan based on 3D computer tomography imaging. This paper introduces a method for segmentation of lung tumors on consecutive computed tomography images. These images are normally only used for correction of movements. The method uses graphs based on electric flow lines. The method offers several advantages when trying to replicate manual segmentations. The method gave a dice coefficient of 0.85 and performed better than level set methods and deformable registration.
    Original languageEnglish
    Title of host publicationImage Analysis and Recognition : 9th International Conference, ICIAR 2012 Aveiro, Portugal, June 25-27, 2012 Proceedings, Part II
    Publication date2012
    ISBN (Print)978-3-642-31297-7
    ISBN (Electronic)978-3-642-31298-4
    Publication statusPublished - 2012
    EventInternational Conference on Image Analysis and Recognition, ICIAR 2012 - Aveiro, Portugal
    Duration: 25 Jun 201227 Jun 2012


    ConferenceInternational Conference on Image Analysis and Recognition, ICIAR 2012
    Internet address
    SeriesLecture Notes in Computer Science


    • Electric flow line
    • Segmentation
    • Lung tumor
    • Graph cut


    Dive into the research topics of 'Lung Tumor Segmentation Using Electric Flow Lines for Graph Cuts'. Together they form a unique fingerprint.

    Cite this