Advanced search
1 file | 1.04 MB Add to list
Author
Organization
Abstract
Recent studies have demonstrated the importance of neural networks in medical image processing and analysis. However, their great efficiency in segmentation tasks is highly dependent on the amount of training data. When these networks are used on small datasets, the process of data augmentation can be very significant. We propose a convolutional neural network approach for the whole heart segmentation which is based upon the 3D U-Net architecture and incorporates principle component analysis as an additional data augmentation technique. The network is trained end-to-end i.e. no pre-trained network is required. Evaluation of the proposed approach is performed on 20 3D CT images from MICCAI 2017 Multi-Modality Whole Heart Segmentation Challenge dataset, divided into 15 training and 5 validation images. Final segmentation results show a high Dice coefficient overlap to ground truth, indicating that the proposed approach is competitive to state-of-the-art. Additionally, we provide the discussion of the influence of different learning rates on the final segmentation results.
Keywords
computerised tomography, image segmentation, learning (artificial intelligence), medical image processing, neural net architecture, neural nets, CT images, 3D U-Net architecture, neural networks, medical image processing, segmentation tasks, training data, convolutional neural network approach, principle component analysis, pre-trained network, validation images, whole heart segmentation, 3D CT images, high Dice coefficient, Image segmentation, Training, Heart, Three-dimensional displays, Computer architecture, Computed tomography, Decoding, CT, data augmentation, heart segmentation, medical image segmentation, neural networks, volumetric segmentation

Downloads

  • Whole Heart Segmentation from CT images Using3D U-Net architecture-2.pdf
    • full text (Author's original)
    • |
    • open access
    • |
    • PDF
    • |
    • 1.04 MB

Citation

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

MLA
Habijan, Marija, et al. “Whole Heart Segmentation from CT Images Using 3D U-Net Architecture.” PROCEEDINGS OF 2019 INTERNATIONAL CONFERENCE ON SYSTEMS, SIGNALS AND IMAGE PROCESSING (IWSSIP 2019), edited by S RimacDrlje et al., 2019, pp. 121–26.
APA
Habijan, M., Leventic, H., Galic, I., & Babin, D. (2019). Whole heart segmentation from CT images using 3D U-Net architecture. In S. RimacDrlje, D. Zagar, I. Galic, G. Martinovic, D. Vranjes, & M. Habijan (Eds.), PROCEEDINGS OF 2019 INTERNATIONAL CONFERENCE ON SYSTEMS, SIGNALS AND IMAGE PROCESSING (IWSSIP 2019) (pp. 121–126). Osijek, Croatia.
Chicago author-date
Habijan, Marija, Hrvoje Leventic, Irena Galic, and Danilo Babin. 2019. “Whole Heart Segmentation from CT Images Using 3D U-Net Architecture.” In PROCEEDINGS OF 2019 INTERNATIONAL CONFERENCE ON SYSTEMS, SIGNALS AND IMAGE PROCESSING (IWSSIP 2019), edited by S RimacDrlje, D Zagar, I Galic, G Martinovic, D Vranjes, and M Habijan, 121–26.
Chicago author-date (all authors)
Habijan, Marija, Hrvoje Leventic, Irena Galic, and Danilo Babin. 2019. “Whole Heart Segmentation from CT Images Using 3D U-Net Architecture.” In PROCEEDINGS OF 2019 INTERNATIONAL CONFERENCE ON SYSTEMS, SIGNALS AND IMAGE PROCESSING (IWSSIP 2019), ed by. S RimacDrlje, D Zagar, I Galic, G Martinovic, D Vranjes, and M Habijan, 121–126.
Vancouver
1.
Habijan M, Leventic H, Galic I, Babin D. Whole heart segmentation from CT images using 3D U-Net architecture. In: RimacDrlje S, Zagar D, Galic I, Martinovic G, Vranjes D, Habijan M, editors. PROCEEDINGS OF 2019 INTERNATIONAL CONFERENCE ON SYSTEMS, SIGNALS AND IMAGE PROCESSING (IWSSIP 2019). 2019. p. 121–6.
IEEE
[1]
M. Habijan, H. Leventic, I. Galic, and D. Babin, “Whole heart segmentation from CT images using 3D U-Net architecture,” in PROCEEDINGS OF 2019 INTERNATIONAL CONFERENCE ON SYSTEMS, SIGNALS AND IMAGE PROCESSING (IWSSIP 2019), Osijek, Croatia, 2019, pp. 121–126.
@inproceedings{8633939,
  abstract     = {Recent studies have demonstrated the importance of neural networks in medical image processing and analysis. However, their great efficiency in segmentation tasks is highly dependent on the amount of training data. When these networks are used on small datasets, the process of data augmentation can be very significant. We propose a convolutional neural network approach for the whole heart segmentation which is based upon the 3D U-Net architecture and incorporates principle component analysis as an additional data augmentation technique. The network is trained end-to-end i.e. no pre-trained network is required. Evaluation of the proposed approach is performed on 20 3D CT images from MICCAI 2017 Multi-Modality Whole Heart Segmentation Challenge dataset, divided into 15 training and 5 validation images. Final segmentation results show a high Dice coefficient overlap to ground truth, indicating that the proposed approach is competitive to state-of-the-art. Additionally, we provide the discussion of the influence of different learning rates on the final segmentation results.},
  author       = {Habijan, Marija and Leventic, Hrvoje and Galic, Irena and Babin, Danilo},
  booktitle    = {PROCEEDINGS OF 2019 INTERNATIONAL CONFERENCE ON SYSTEMS, SIGNALS AND IMAGE PROCESSING (IWSSIP 2019)},
  editor       = {RimacDrlje, S and Zagar, D and Galic, I and Martinovic, G and Vranjes, D and Habijan, M},
  isbn         = {9781728132532},
  issn         = {2157-8672},
  keywords     = {computerised tomography,image segmentation,learning (artificial intelligence),medical image processing,neural net architecture,neural nets,CT images,3D U-Net architecture,neural networks,medical image processing,segmentation tasks,training data,convolutional neural network approach,principle component analysis,pre-trained network,validation images,whole heart segmentation,3D CT images,high Dice coefficient,Image segmentation,Training,Heart,Three-dimensional displays,Computer architecture,Computed tomography,Decoding,CT,data augmentation,heart segmentation,medical image segmentation,neural networks,volumetric segmentation},
  language     = {eng},
  location     = {Osijek, Croatia},
  pages        = {121--126},
  title        = {Whole heart segmentation from CT images using 3D U-Net architecture},
  url          = {http://dx.doi.org/10.1109/IWSSIP.2019.8787253},
  year         = {2019},
}

Altmetric
View in Altmetric
Web of Science
Times cited: