# Coronary Centerline Extraction from CCTA Using 3D-UNet

^{*}

## Abstract

**:**

## 1. Introduction

## 2. Related Research

#### 2.1. Rule-Based Centerline Extraction

#### 2.2. Machine Learning-Based Centerline Extraction

## 3. Proposed Method

#### 3.1. Neural Network Architecture

#### 3.2. Resize or Patches

#### 3.3. Loss Functions

#### 3.3.1. Local Loss

#### 3.3.2. Global Loss

#### 3.3.3. Combined Loss Functions

#### 3.3.4. Proposed Loss Function

#### 3.4. Generating the Full Output

## 4. Experimental Results and Discussion

#### 4.1. Coronary Dataset

#### 4.2. Visualization

#### 4.3. Experimental Setup

#### 4.4. Training the Network

- Input patch size;
- Layer reduction;
- Batch size;
- Feeding or not the network patches with no single voxel of ground truth.

#### 4.5. Results

## 5. Discussion

## 6. Conclusions

## Author Contributions

## Funding

## Institutional Review Board Statement

## Informed Consent Statement

## Data Availability Statement

## Conflicts of Interest

## Appendix A. Graph Results for Considered Hyperparameters

**Figure A1.**Graphs of the evaluation metrics for different parameter configurations for training and validation, respectively, (

**a**,

**b**) 128 × 128 × 96, batch size 1, reduction 1, only patches with ground truth; (

**c**,

**d**) 128 × 128 × 128, batch size 2, reduction 2, only patches with ground truth; (

**e**,

**f**) 256 × 256 × 128, batch size 1, reduction 2; (

**g**,

**h**) 320 × 320 × 64, batch size 1, reduction 4; and (

**i**,

**j**) 384 × 384 × 48, batch size 1, reduction 4.

**Figure A2.**Loss function switch for (

**a**) patch size 128 × 128 × 128 and (

**b**) Patch size 256 × 256 × 128.

## Appendix B. Visual Validation for Other Training Configurations

**Figure A3.**Visual validation for different parameter configurations. The ground truth centerline is depicted in red. The segmentation (after thinning) is seen in green for (

**a**) 128 × 128 × 96, batch size 1, reduction 1, only patches with ground truth; (

**b**) 128 × 128 × 128, batch size 2, reduction 2, only patches with ground truth; (

**c**) 256 × 256 × 128, batch size 1, reduction 2; (

**d**) 320 × 320 × 64, batch size 1, reduction 4; and (

**e**) 384 × 384 × 48, batch size 1, reduction 4.

## Appendix C. Input Size 256 × 256 × 128, Model Reduction 4

## References

- Galal, H.; Rashid, T.; Alghonaimy, W.; Kamal, D. Detection of positively remodeled coronary artery lesions by multislice CT and its impact on cardiovascular future events. Egypt Heart J.
**2019**, 71, 26. [Google Scholar] [CrossRef] [Green Version] - Pantos, I.; Katritsis, D. Fractional Flow Reserve Derived from Coronary Imaging and Computational Fluid Dynamics. Interv. Cardiol. Rev.
**2014**, 9, 145. [Google Scholar] [CrossRef] [Green Version] - Zhao, Y.; Ping, J.; Yu, X.; Wu, R.; Sun, C.; Zhang, M. Fractional flow reserve-based 4D hemodynamic simulation of time-resolved blood flow in left anterior descending coronary artery. Clin. Biomech.
**2019**, 70, 164–169. [Google Scholar] [CrossRef] [PubMed] - Williams, L.H.; Drew, T. What do we know about volumetric medical image interpretation? A review of the basic science and medical image perception literatures. Cogn. Res.
**2019**, 4, 21. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Zhao, F.; Chen, Y.; Hou, Y.; He, X. Segmentation of blood vessels using rule-based and machine-learning-based methods: A review. Multimed. Syst.
**2019**, 25, 109–118. [Google Scholar] [CrossRef] - Chen, X.; Fu, Y.; Lin, J.; Ji, Y.; Fang, Y.; Wu, J. Coronary Artery Disease Detection by Machine Learning with Coronary Bifurcation Features. Appl. Sci.
**2020**, 10, 7656. [Google Scholar] [CrossRef] - Danilov, A.; Pryamonosov, R.; Yurova, A. Image Segmentation for Cardiovascular Biomedical Applications at Different Scales. Computation
**2016**, 4, 35. [Google Scholar] [CrossRef] [Green Version] - Bates, R.; Irving, B.; Markelc, B.; Kaeppler, J.; Muschel, R.; Grau, V.; Schnabel, J.A. Extracting 3D Vascular Structures from Microscopy Images using Convolutional Recurrent Networks. arXiv
**2017**, arXiv:1705.09597. [Google Scholar] - Han, D.; Shim, H.; Jeon, B.; Jang, Y.; Hong, Y.; Jung, S.; Ha, S.; Chang, H.-J. Automatic Coronary Artery Segmentation Using Active Search for Branches and Seemingly Disconnected Vessel Segments from Coronary CT Angiography. PLoS ONE
**2016**, 11, e0156837. [Google Scholar] [CrossRef] [PubMed] - Liu, L.; Xu, J.; Liu, Z. Automatic segmentation of coronary lumen based on minimum path and image fusion from cardiac computed tomography images. Clust. Comput
**2019**, 22, 1559–1568. [Google Scholar] [CrossRef] - Kong, B.; Wang, X.; Bai, J.; Lu, Y.; Gao, F.; Cao, K.; Xia, J.; Song, Q.; Yin, Y. Learning tree-structured representation for 3D coronary artery segmentation. Comput. Med. Imaging Graph.
**2020**, 80, 101688. [Google Scholar] [CrossRef] - Lv, T.; Yang, G.; Zhang, Y.; Yang, J.; Chen, Y.; Shu, H.; Luo, L. Vessel segmentation using centerline constrained level set method. Multimed Tools Appl.
**2019**, 78, 17051–17075. [Google Scholar] [CrossRef] - Gao, Z.; Liu, X.; Qi, S.; Wu, W.; Hau, W.K.; Zhang, H. Automatic segmentation of coronary tree in CT angiography images. Int. J. Adapt. Control. Signal. Process.
**2019**, 33, 1239–1247. [Google Scholar] [CrossRef] - Sheng, X.; Fan, T.; Jin, X.; Jin, J.; Chen, Z.; Zheng, G.; Lu, M.; Zhu, Z. Extraction Method of Coronary Artery Blood Vessel Centerline in CT Coronary Angiography. IEEE Access
**2019**, 7, 170690–170702. [Google Scholar] [CrossRef] - Cui, H.; Xia, Y. Automatic Coronary Centerline Extraction Using Gradient Vector Flow Field and Fast Marching Method From CT Images. IEEE Access
**2018**, 6, 41816–41826. [Google Scholar] [CrossRef] - Guo, Z.; Bai, J.; Lu, Y.; Wang, X.; Cao, K.; Song, Q.; Sonka, M.; Yin, Y. DeepCenterline: A Multi-task Fully Convolutional Network for Centerline Extraction. In Information Processing in Medical Imaging; Chung, A.C.S., Gee, J.C., Yushkevich, P.A., Bao, S., Eds.; Lecture Notes in Computer Science; Springer International Publishing: Cham, Switzerland, 2019; Volume 11492, pp. 441–453. ISBN 978-3-030-20350-4. [Google Scholar]
- Wolterink, J.M.; van Hamersvelt, R.W.; Viergever, M.A.; Leiner, T.; Išgum, I. Coronary artery centerline extraction in cardiac CT angiography using a CNN-based orientation classifier. Med. Image Anal.
**2019**, 51, 46–60. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Sainath, T.N.; Vinyals, O.; Senior, A.; Sak, H. Convolutional, Long Short-Term Memory, fully connected Deep Neural Networks. In Proceedings of the 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP); IEEE: South Brisbane, Australia, 2015; pp. 4580–4584. [Google Scholar] [CrossRef]
- Salahuddin, Z.; Lenga, M.; Nickisch, H. Multi-Resolution 3D Convolutional Neural Networks for Automatic Coronary Centerline Extraction in Cardiac CT Angiography Scans. arXiv
**2020**, arXiv:2010.00925. [Google Scholar] - He, J.; Pan, C.; Yang, C.; Zhang, M.; Wang, Y.; Zhou, X.; Yu, Y. Learning Hybrid Representations for Automatic 3D Vessel Centerline Extraction. In Medical Image Computing and Computer Assisted Intervention—MICCAI 2020; Martel, A.L., Abolmaesumi, P., Stoyanov, D., Mateus, D., Zuluaga, M.A., Zhou, S.K., Racoceanu, D., Joskowicz, L., Eds.; Lecture Notes in Computer Science; Springer International Publishing: Cham, Switzerland, 2020; Volume 12266, pp. 24–34. ISBN 978-3-030-59724-5. [Google Scholar]
- Ronneberger, O.; Fischer, P.; Brox, T. U-Net: Convolutional Networks for Biomedical Image Segmentation. In Medical Image Computing and Computer-Assisted Intervention—MICCAI 2015; Navab, N., Hornegger, J., Wells, W.M., Frangi, A.F., Eds.; Lecture Notes in Computer Science; Springer International Publishing: Cham, Switzerland, 2015; Volume 9351, pp. 234–241. ISBN 978-3-319-24573-7. [Google Scholar]
- Irving, B.; Hutton, C.; Dennis, A.; Vikal, S.; Mavar, M.; Kelly, M.; Brady, J.M. Deep Quantitative Liver Segmentation and Vessel Exclusion to Assist in Liver Assessment. In Medical Image Understanding and Analysis; Valdés Hernández, M., González-Castro, V., Eds.; Communications in Computer and Information Science; Springer International Publishing: Cham, Switzerland, 2017; Volume 723, pp. 663–673. ISBN 978-3-319-60963-8. [Google Scholar]
- Jin, Q.; Meng, Z.; Pham, T.D.; Chen, Q.; Wei, L.; Su, R. DUNet: A deformable network for retinal vessel segmentation. Knowl. Based Syst.
**2019**, 178, 149–162. [Google Scholar] [CrossRef] [Green Version] - Zhuang, J. LadderNet: Multi-path networks based on U-Net for medical image segmentation. arXiv
**2019**, arXiv:1810.07810. [Google Scholar] - Guo, C.; Szemenyei, M.; Yi, Y.; Wang, W.; Chen, B.; Fan, C. SA-UNet: Spatial Attention U-Net for Retinal Vessel Segmentation. arXiv
**2020**, arXiv:2004.03696. [Google Scholar] - Çiçek, Ö.; Abdulkadir, A.; Lienkamp, S.S.; Brox, T.; Ronneberger, O. 3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation. arXiv
**2016**, arXiv:1606.06650. [Google Scholar] - Huang, C.; Han, H.; Yao, Q.; Zhu, S.; Zhou, S.K. 3D U2-Net: A 3D Universal U-Net for Multi-domain Medical Image Segmentation. In Proceedings of the Medical Image Computing and Computer Assisted Intervention—MICCAI 2019; Shen, D., Liu, T., Peters, T.M., Staib, L.H., Essert, C., Zhou, S., Yap, P.-T., Khan, A., Eds.; Springer International Publishing: Cham, Switzerland, 2019; Volume 11765, pp. 291–299. [Google Scholar] [CrossRef] [Green Version]
- Lin, T.-Y.; Goyal, P.; Girshick, R.; He, K.; Dollár, P. Focal Loss for Dense Object Detection. arXiv
**2018**, arXiv:1708.02002. [Google Scholar] - Cui, Y.; Jia, M.; Lin, T.-Y.; Song, Y.; Belongie, S. Class-Balanced Loss Based on Effective Number of Samples. arXiv
**2019**, arXiv:1901.05555. [Google Scholar] - Rotterdam Coronary Artery Algorithm Evaluation Framework. Available online: http://coronary.bigr.nl/centerlines/ (accessed on 11 January 2020).
- Schaap, M.; Metz, C.T.; van Walsum, T.; van der Giessen, A.G.; Weustink, A.C.; Mollet, N.R.; Bauer, C.; Bogunović, H.; Castro, C.; Deng, X. Standardized evaluation methodology and reference database for evaluating coronary artery centerline extraction algorithms. Med. Image Anal.
**2009**, 13, 701–714. [Google Scholar] [CrossRef] [Green Version] - Rotterdam Coronary Artery Challenge Categories. Available online: http://coronary.bigr.nl/centerlines/about.php (accessed on 11 January 2020).
- 3D Slicer. Available online: https://www.slicer.org/ (accessed on 11 January 2020).
- GitHub Slicer. Available online: https://github.com/Slicer/Slicer (accessed on 11 January 2020).
- Alexandru Dorobanțiu—GitHub. Available online: https://github.com/AlexDorobantiu (accessed on 11 May 2019).
- Jonker, P.P. Morphological Operations on 3D and 4D Images: From Shape Primitive Detection to Skeletonization. In Discrete Geometry for Computer Imagery; Borgefors, G., Nyström, I., di Baja, G.S., Eds.; Lecture Notes in Computer Science; Springer: Berlin/Heidelberg, Germany, 2000; Volume 1953, pp. 371–391. ISBN 978-3-540-41396-7. [Google Scholar]
- MIC-DKFZ/batchgenerators. Available online: https://github.com/MIC-DKFZ/batchgenerators (accessed on 29 August 2020).
- Frid-Adar, M.; Diamant, I.; Klang, E.; Amitai, M.; Goldberger, J.; Greenspan, H. GAN-based synthetic medical image augmentation for increased CNN performance in liver lesion classification. Neurocomputing
**2018**, 321, 321–331. [Google Scholar] [CrossRef] [Green Version] - AlexDorobantiu/CoronaryCenterlineUnet. Available online: https://github.com/AlexDorobantiu/CoronaryCenterlineUnet (accessed on 29 August 2020).
- Iglovikov, V.; Shvets, A. TernausNet: U-Net with VGG11 Encoder Pre-Trained on ImageNet for Image Segmentation. arXiv
**2018**, arXiv:1801.05746. [Google Scholar] - Manzo, M.; Pellino, S. Bucket of Deep Transfer Learning Features and Classification Models for Melanoma Detection. J. Imaging
**2020**, 6, 129. [Google Scholar] [CrossRef] - WHO—Cardiovascular diseases (CVDs) Fact Sheet. Available online: https://www.who.int/news-room/fact-sheets/detail/cardiovascular-diseases-(cvds) (accessed on 24 September 2020).

**Figure 2.**Downscaled and upscaled centerline ground truth (80 × 80 × 64) plotted in 3D to highlight the difficulty of using it for training.

**Figure 5.**Patch size 256 × 256 × 128: (

**a**) training only with ground truth patches and (

**b**) validation.

**Figure 6.**Patch size 256 × 256 × 128; a 3D Slicer rendering of ground truth (red) and centerline segmentation (green).

CTA | Size (Voxels) | Resolution (mm^{3}) |
---|---|---|

dataset00 | 512 × 512 × 272 | 0.363 × 0.363 × 0.4 |

dataset01 | 512 × 512 × 338 | 0.363 × 0.363 × 0.4 |

dataset02 | 512 × 512 × 288 | 0.334 × 0.334 × 0.4 |

dataset03 | 512 × 512 × 276 | 0.371 × 0.371 × 0.4 |

dataset04 | 512 × 512 × 274 | 0.316 × 0.316 × 0.4 |

dataset05 | 512 × 512 × 274 | 0.322 × 0.322 × 0.4 |

dataset06 | 512 × 512 × 268 | 0.320 × 0.320 × 0.4 |

dataset07 | 512 × 512 × 304 | 0.287 × 0.287 × 0.4 |

Patch Size | Batch Size | Model Reduction | Patches with Ground Truth Only | Epoch | Overlap | Accuracy |
---|---|---|---|---|---|---|

128 × 128 × 96 | 1 | 1 | true | 231 | 0.934 | 0.921 |

128 × 128 × 128 | 2 | 2 | true | 260 | 0.895 | 0.940 |

256 × 256 × 128 | 1 | 2 | true | 232 | 0.917 | 0.943 |

256 × 256 × 128 | 1 | 2 | false | 392 | 0.944 | 0.911 |

320 × 320 × 64 | 1 | 4 | false | 1250 | 0.883 | 0.953 |

384 × 384 × 48 | 1 | 4 | false | 1500 | 0.910 | 0.928 |

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |

© 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Dorobanțiu, A.; Ogrean, V.; Brad, R.
Coronary Centerline Extraction from CCTA Using 3D-UNet. *Future Internet* **2021**, *13*, 101.
https://doi.org/10.3390/fi13040101

**AMA Style**

Dorobanțiu A, Ogrean V, Brad R.
Coronary Centerline Extraction from CCTA Using 3D-UNet. *Future Internet*. 2021; 13(4):101.
https://doi.org/10.3390/fi13040101

**Chicago/Turabian Style**

Dorobanțiu, Alexandru, Valentin Ogrean, and Remus Brad.
2021. "Coronary Centerline Extraction from CCTA Using 3D-UNet" *Future Internet* 13, no. 4: 101.
https://doi.org/10.3390/fi13040101