(Sunday, 7/14/2019)
Room: ePoster Forums
Purpose: This paper aims to automatically segment the rectum on female pelvic 3D CT image based on low-sample training using Dense V-Network fusion network model.
Methods: Dense Net and V-Net network models were combined to generate Dense V-Network network model which has dense link and residual structure .Fusion network model can reduce the number of parameters and redundant calculation ,speed up the convergence speed. It also can effectively solves the gradient disappearing explosion problem that occurs with the increase of depth when training three-dimensional data. 100 female pelvic CT images were used for annotation, 80 of which were used as training sets and 20 of which were used as test sets. The accuracy of the fusion model was analyzed using 8 indicators such as DSC.
Results: The average values of the eight indicators were as follows: Dice Similarity Coefficient (0.92), Hausdorff Distance (2.11mm), Jaccard Distance (0.63mm), Deviation of Volume (15, 91%) , Sensitivity Index (0.77), Inclusive Index (0.78), Minimum Distance Average (2.46mm), Deviation of Centroid (0.71) and segmentation time within 30s.
Conclusion: The Dense V-Network fusion model algorithm can quickly achieve to accurate delineation of rectal cancer patients with a small sample size.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
