磁共振掃描具有非侵入優勢與無輻射的特性，其原理是藉由組織對比度不同，去攝影出不同的成像，為心血管疾病的診斷提供一大利器，近年來磁共振心肌影像的T1值被視為重要的生物指標，因此想藉由心肌的T1值辨別疾病，必須要精準的圖像心肌切割系統，才能將心肌T1值計算出來，進而辨別疾病與分析，本研究將藉由兩組心臟的磁振造影資料集，深度學習演算法的語意分析架構，去完成自動切割心肌的系統，並在過程中使用，資料增強術、轉移式學習使模型變穩健，最後將語義分析架構預測的結果，結合多層分析法，探討多層分析法提供更有區辨力特徵的可能性，並將本研究結果，封裝成一個應用放在雲端系統上，致力於提供一個全自動心肌T1值分析系統的雛型。

Cardiovascular magnetic resonance imaging (CMR) is a medical imaging technology for non-invasive assessment of the function and structure of the cardiovascular system. Native T1 values of the myocardium are recently considered potential biomarkers of myocardial fibrosis. For this reason, we would like to build an automatically segmentation and analysis system for CMR.
In this study, we used two different datasets and deep learning networks for the segmentation task. In the training progress, we used data augmentation and transfer learning method to help model got more highly adaptation. Finally, we got a stable model for segmentation and combined with a multilayer analysis concept to provide distinctive features to analysis.

1. Cannao, P.M., et al., Novel cardiac magnetic resonance biomarkers: native T1 and extracellular volume myocardial mapping. Eur Heart J Suppl, 2016. 18(Suppl E): p. E64-E71.
2. Captur, G., C. Manisty, and J.C. Moon, Cardiac MRI evaluation of myocardial disease. Heart, 2016. 102(18): p. 1429-35.
3. Sarvari, S.I., et al., Layer-specific quantification of myocardial deformation by strain echocardiography may reveal significant CAD in patients with non-ST-segment elevation acute coronary syndrome. JACC Cardiovasc Imaging, 2013. 6(5): p. 535-44.
4. Nagata, Y., et al., Normal range of myocardial layer-specific strain using two-dimensional speckle tracking echocardiography. PLoS One, 2017. 12(6): p. e0180584.
5. Lee, H., et al., Unsupervised Learning of Hierarchical Representations with Convolutional Deep Belief Networks. Communications of the Acm, 2011. 54(10): p. 95-103.
6. LeCun, Y., Y. Bengio, and G. Hinton, Deep learning. Nature, 2015. 521(7553): p. 436-444.
7. Long, J., E. Shelhamer, and T. Darrell. Fully convolutional networks for semantic segmentation. in Proceedings of the IEEE conference on computer vision and pattern recognition. 2015.
8. Krizhevsky, A., I. Sutskever, and G.E. Hinton. Imagenet classification with deep convolutional neural networks. in Advances in neural information processing systems. 2012.
9. Badrinarayanan, V., A. Kendall, and R. Cipolla, SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation. IEEE Trans Pattern Anal Mach Intell, 2017. 39(12): p. 2481-2495.
10. Konukoglu, O.C.E., M.P.K. Rhode, and M.S.A. Young, Statistical Atlases and Computational Models of the Heart.
11. Messroghli, D.R., et al., Modified Look‐Locker inversion recovery (MOLLI) for high‐resolution T1 mapping of the heart. Magnetic resonance in medicine, 2004. 52(1): p. 141-146.
12. Kuetting, D.L.R., et al., Flip angle optimization for balanced SSFP: Cardiac cine imaging following the application of standard extracellular contrast agent (gadobutrol). J Magn Reson Imaging, 2018. 47(1): p. 255-261.
13. Maas, A.L., A.Y. Hannun, and A.Y. Ng. Rectifier nonlinearities improve neural network acoustic models. in Proc. icml. 2013.
14. Kingma, D.P. and J. Ba, Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014.
15. Golik, P., P. Doetsch, and H. Ney. Cross-entropy vs. squared error training: a theoretical and experimental comparison. in Interspeech. 2013.
16. Nathan, M., et al., Assessment of myocardial fibrosis with cardiac magnetic resonance. Journal of the American college of cardiology, 2011. 57(8): p. 891.
17. Ho, C.Y., et al., Myocardial Fibrosis as an Early Manifestation of Hypertrophic Cardiomyopathy. New England Journal of Medicine, 2010. 363(6): p. 552-563.
18. Kozak, M.F., et al., Diffuse myocardial fibrosis following tetralogy of Fallot repair: a T1 mapping cardiac magnetic resonance study. Pediatric radiology, 2014. 44(4): p. 403-409.
19. Xue, H., et al., Motion correction for myocardial T1 mapping using image registration with synthetic image estimation. Magnetic resonance in medicine, 2012. 67(6): p. 1644-1655.
20. Cerqueira, M.D., et al., Standardized myocardial segmentation and nomenclature for tomographic imaging of the heart: a statement for healthcare professionals from the Cardiac Imaging Committee of the Council on Clinical Cardiology of the American Heart Association. Circulation, 2002. 105(4): p. 539-542.
21. Zhao, H., et al. Pyramid scene parsing network. in IEEE Conf. on Computer Vision and Pattern Recognition (CVPR). 2017.
22. Chen, L.-C., et al., Rethinking atrous convolution for semantic image segmentation. arXiv preprint arXiv:1706.05587, 2017.