MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Testing and Implementation of Atlas Based Segmentation for Radiotherapy Treatment Planning

M Jameson*, J Hellyer , C Choong , G Dinsdale , Ingham Institute & Liverpool Cancer Centre, Sydney, NSW

Presentations

(Monday, 7/30/2018) 9:30 AM - 10:00 AM

Room: Exhibit Hall | Forum 2

Purpose: To assess the quality and efficiency of atlas based segmentation for a number of treatment sites using a commercial software package

Methods: The MiM® software package was used for atlas based segmentation. This software ships with a number of anatomical atlases. Local atlases were created for prostate, breast and brain using 17, 12 and 20 patient datasets. The brain atlas used a combination of CT and MRI. The local atlases were compared to manual segmentations and the shipped atlas (where applicable). Contour comparison metrics included Dice coefficient, Hausdorff distance and volume. Efficiency was assessed by collecting timing data for manual contouring, atlas segmentation and editing atlas contours post automatic segmentation.

Results: The Dice coefficient when comparing the shipped and local atlases to manual segmentations was 0.72±0.11 and 0.85±0.04. The mean Hausdorff distance when comparing shipped and local atlases to manual segmentations was 0.35±0.14 and 0.18±0.05 cm. Average runtime for the local patient atlas for 5 prostate cases was 10.5 minutes, for the shipped prostate atlas 8.21 minutes and the time to manually contour these same 5 cases was 16.2 minutes. The brain atlas showed some deficiencies like due to the lack of signal from bone tissue in the skull.

Conclusion: The atlases created using local patient data were shown to have higher Dice coefficients and reduced Hausdorff distances compared to the shipped atlas. Both atlases were faster than manual contouring. We are currently investigating the use ultra-short TE sequences to improve the brain MRI atlas.

Funding Support, Disclosures, and Conflict of Interest: Licencing agreement with Standard Imaging Inc

Keywords

CT, MRI, Segmentation

Taxonomy

IM/TH- image segmentation: Multi-modality segmentation

Contact Email