Grading evaluation study of atlas based autosegmentationof organs at risk in thorax

Background: the grading evaluation of atlas based auto-segmentation (abas)of organs at risk (oars) in thorax was studied. materials and methods: fortypatients with thoracic cancer were included in this study, and for eachthirteen thoracic oars were delineated by an experienced radiationoncologist. the patients were randomly grouped into the training and the testdataset (20 each). the investigated abas strategies included single-atlas(single), majority voting with 5 atlas matches (mv5) and simultaneous truthand performance level estimation (staple) with 5 atlas matches (st5). thedice similarity coefficient (dsc), the difference of the euclidean distancebetween centers of mass (δcmd), the difference of volume (δv), maximumhausdorff distance (mhd) and average hausdorff distance (ahd) betweenauto-segmented and manual contours were calculated. results: most of theindex values (33/65) of st5 were optimal. there were differences in thegrading results for the five indexes. with dsc, five, four and four oars weregraded into level 3, level 2 and level 1, respectively. the mean dsc valuesranged from 0.88 to 0.96, from 0.73 to 0.79, and from 0.53 to 0.62 for thelevel 3, level 2 and level 1, respectively. conclusion: grading evaluation ofabas of thoracic oars based on the dsc proved to be feasible and relativelymore reliable. the thoracic oars auto-segmentation was divided into threelevels based on the dsc. level 3 oars can be auto-segmented, level 2 oarsdelineations need to be manually modified after the auto-segmentation, andlevel 1 oars are not recommended for the auto-segmentation.