1 / 19

1 University of Rouen, LITIS EA4108, Rouen, France 2 Centre Henri-Becquerel, Rouen, France

1 University of Rouen, LITIS EA4108, Rouen, France 2 Centre Henri-Becquerel, Rouen, France. Objective. Conventional radiotherapy (RT) treatment “ Fractionated ” radiotherapy total dose: 60-70 Gy 2 Gy per fraction (day) 5 fractions per week Disappointing outcome

porter
Download Presentation

1 University of Rouen, LITIS EA4108, Rouen, France 2 Centre Henri-Becquerel, Rouen, France

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. 1University of Rouen, LITIS EA4108, Rouen, France 2Centre Henri-Becquerel, Rouen, France

  2. Objective • Conventional radiotherapy (RT) treatment • “Fractionated” radiotherapy • total dose: 60-70 Gy • 2 Gy per fraction (day) • 5 fractions per week • Disappointing outcome • 12 months’ survival [*] • Variation of response among patients • in terms of tumor volume regression • Adapt the treatment individually • Iterative tumor segmentationduring the RT • Sequential medical images [*]: Dubray et al., in IRBM(2013)

  3. Objective • To segment lung tumors in longitudinal PET images during radiotherapy • FDG-PET: metabolic activity baseline treatment

  4. Challenge weak contrast baseline treatment small tumor CHALLENGE low spatial resolution 4x4x2 mm3

  5. State of the art • Methods using a fixed threshold value, in terms of SUV (Standardized Uptake Value) • 40% of the SUVmax (T40, Ye et al., 2002) • Adaptive thresholding (TAD, Vauclin et al., 2009) • Advanced segmentation methods • statistics (FLAB, Hatt et al., 2009) • random walks (RW, Bagci et al., 2011) • ... SUV - La Valeur de fixation normalisée = (fixation dans le tissu d'intérêt)/((dose injectée)/(poids du patient))

  6. Tumor Growth Model[*] • Advection Reaction Equation • tumor cell density • advection: describing the advective flux transport of tumor cells • proliferation: representing the tumor cell proliferation • treatment: quantifying the radiotherapeuticefficiency [*]: Mi et al., in IEEE Trans. Med. Imaging (2014)

  7. Prediction • Prediction of tumor region at time

  8. Proposed Segmentation Method • Random Walks (Grady, 2006) • Graph: nodes & edges • Probability belonging to the tumor D Intensity Similarity Input image

  9. Proposed Segmentation Method • Random Walks (Grady, 2006) • Graph: nodes & edges • Probability belonging to the tumor • + prediction Intensity Similarity Input image

  10. Solution • The matrix form: • Laplacian matrix, such that where , is the geometric distance of node iand j. Subscripts U: unlabeled nodes Subscripts L: labeled nodes [Onoma et al., 2012]

  11. Definition of labeled nodes • Illustration ROI Label Seeds Input PET ROI boundary Non tumor seeds Define ROI Tumor seeds Prediction Dilate • Fuzzy C-Means: • low: C1 • moderate: C2 • high : C3 • Tumor Seeds: • intensity ≥ mean(C2, C3) Unlabeled nodes [Onoma et al., 2012]

  12. Solution • Illustration ROI Label Seeds Solve Dirichlet Problem Input PET Probability map Define ROI Output Prediction Threshold Dilate Segmentation

  13. Experimental Results Our Method • Data: • Three types of tumors: largeand smallsizes, heterogeneoustumors Input Expert RW T40 TAD FLAB large

  14. Experimental Results Our Method Input Expert RW T40 TAD FLAB heterogeneous small

  15. S E Experimental Results • Data: • 7 patients • 15 follow-up PETs Comparison of different methods (S) with the Expert (E) delineation spatial location volume

  16. Experimental Results • Data: • 7 patients • 15 follow-up PETs Comparison of different methods (S) with the Expert (E) delineation

  17. Tumor longitudinal Volumes PET Images Expert Our Method baseline treatment

  18. Conclusion • Automatic tumor segmentation method • Follow up tumor response • Good performance for small tumor • Validation on larger patient images

  19. Thank you !

More Related