Advanced search
1 file | 623.90 KB Add to list

Binary Glioma grading : radiomics versus pre-trained CNN features

Milan Decuyper (UGent) , Stijn Bonte and Roel Van Holen (UGent)
Author
Organization
Abstract
We compare the predictive performance of hand-engineered radiomics features with features extracted through a pre-trained CNN for discriminating glioblastoma from lower- grade glioma. The BRATS 2017 database was used containing MRI data of 285 patients. State-of-the-art performance was achieved (AUC of 96.4%) with radiomics features extracted from manually segmented tumour volumes. With pre-trained CNN features extracted from the tumour bounding box, an AUC of 93.5% was obtained.
Keywords
MRI, Radiomics, CNN, Glioma Grading

Downloads

  • (...).pdf
    • full text
    • |
    • UGent only
    • |
    • PDF
    • |
    • 623.90 KB

Citation

Please use this url to cite or link to this publication:

MLA
Decuyper, Milan, et al. “Binary Glioma Grading : Radiomics versus Pre-Trained CNN Features.” Medical Imaging Summer School 2018 : Medical Imaging Meets Deep Learning, 2018, pp. 13–13.
APA
Decuyper, M., Bonte, S., & Van Holen, R. (2018). Binary Glioma grading : radiomics versus pre-trained CNN features. In Medical Imaging Summer School 2018 : Medical Imaging meets Deep Learning (pp. 13–13). Favignana, Sicily, Italy.
Chicago author-date
Decuyper, Milan, Stijn Bonte, and Roel Van Holen. 2018. “Binary Glioma Grading : Radiomics versus Pre-Trained CNN Features.” In Medical Imaging Summer School 2018 : Medical Imaging Meets Deep Learning, 13–13.
Chicago author-date (all authors)
Decuyper, Milan, Stijn Bonte, and Roel Van Holen. 2018. “Binary Glioma Grading : Radiomics versus Pre-Trained CNN Features.” In Medical Imaging Summer School 2018 : Medical Imaging Meets Deep Learning, 13–13.
Vancouver
1.
Decuyper M, Bonte S, Van Holen R. Binary Glioma grading : radiomics versus pre-trained CNN features. In: Medical Imaging Summer School 2018 : Medical Imaging meets Deep Learning. 2018. p. 13–13.
IEEE
[1]
M. Decuyper, S. Bonte, and R. Van Holen, “Binary Glioma grading : radiomics versus pre-trained CNN features,” in Medical Imaging Summer School 2018 : Medical Imaging meets Deep Learning, Favignana, Sicily, Italy, 2018, pp. 13–13.
@inproceedings{8616460,
  abstract     = {{We compare the predictive performance of hand-engineered radiomics features with features extracted through a pre-trained CNN for discriminating glioblastoma from lower- grade glioma. The BRATS 2017 database was used containing MRI data of 285 patients. State-of-the-art performance was achieved (AUC of 96.4%) with radiomics features extracted from manually segmented tumour volumes. With pre-trained CNN features extracted from the tumour bounding box, an AUC of 93.5% was obtained.}},
  author       = {{Decuyper, Milan and Bonte, Stijn and Van Holen, Roel}},
  booktitle    = {{Medical Imaging Summer School 2018 : Medical Imaging meets Deep Learning}},
  keywords     = {{MRI,Radiomics,CNN,Glioma Grading}},
  language     = {{eng}},
  location     = {{Favignana, Sicily, Italy}},
  pages        = {{13--13}},
  title        = {{Binary Glioma grading : radiomics versus pre-trained CNN features}},
  url          = {{http://iplab.dmi.unict.it/miss/posters.htm}},
  year         = {{2018}},
}