近年來深度學習快速的發展，深度學習被應用在各項領域上，當然也包含
磁振影像領域，像是腫瘤切割或者是一些病灶偵測等等。
在本實驗中，我們提出了 Up U-net 的上採樣深度學習模型架構，輸入是低
解析度的 2mm
2的 T1-Weighted MRI 大腦影像，輸出是高解析度的 1mm
2的 T1-
Weighted MRI 大腦影像，此做法可以有效降低 T1-Weighted MRI 的掃描時間，
進而得到高解析度的影像。我們的實驗使用 PSNR、SSIM 以及 MSE 這三項指標對
網路使用在測試資料上的表現進行評分，其結果顯示我們的作法可以使得所生
成出的高解析度影像與真實的高解析度影像有著高度的相似性。最後在使用
spm12 對所生成出來的 T1-Weighted MRI 大腦影像進行切割，切出灰質、白質
和腦脊髓液並與真實影像進行比較，統計結果同樣顯示得到了高度的相似性。

In recent years, deep learning has rapidly advanced and has been widely applied
across various fields, including the magnetic resonance imaging (MRI). Applications
of deep learning in MRI include tumor segmentation, lesion detection, and many other
areas.
In our experiment, we propose the Up U-net architecture, which is an up-sampling
deep learning model, the input to this model is low-resolution 2mm2 T1-Weighted brain
MRIs, and the output is high-resolution 1mm2 T1-Weighted brain MRIs, our approach
effectively reduces the scanning time of T1-Weighted MRIs while also achieving
getting high-resolution images. In our experiment, we using three metrics to evaluate
the performance of the network applying on test data, which are PSNR (Peak Signalto-Noise Ratio), SSIM (Structural Similarity Index), and MSE (Mean Squared Error),
the results indicated that our network successfully generated high-resolution MRIs that
exhibited a high degree of similarity to the ground truth high-resolution MRIs. Finally,
we segment the generated T1-Weighted brain MRIs by using spm12 into gray matter,
white matter, and cerebrospinal fluid, and compared these regions with ground truth,
the statistical results also indicated a high level of similarity.

[1] X. Glorot, A. Bordes, and Y. Bengio, "Deep sparse rectifier neural networks," in
Proceedings of the fourteenth international conference on artificial
intelligence and statistics, 2011: JMLR Workshop and Conference
Proceedings, pp. 315-323.
[2] S. S. Basha, S. R. Dubey, V. Pulabaigari, and S. Mukherjee, "Impact of fully
connected layers on performance of convolutional neural networks for image
classification," Neurocomputing, vol. 378, pp. 112-119, 2020.
[3] M. Brant-Zawadzki, G. D. Gillan, and W. R. Nitz, "MP RAGE: a threedimensional, T1-weighted, gradient-echo sequence--initial experience in the
brain," Radiology, vol. 182, no. 3, pp. 769-775, 1992.
[4] Z. Hongtao, Y. Shinomiya, and S. Yoshida, "3D brain MRI reconstruction based
on 2D super-resolution technology," in 2020 IEEE international conference on
systems, man, and cybernetics (SMC), 2020: IEEE, pp. 18-23.
[5] X. Dai et al., "Multimodal MRI synthesis using unified generative adversarial
networks," Medical physics, vol. 47, no. 12, pp. 6343-6354, 2020.
[6] O. Ronneberger, P. Fischer, and T. Brox, "U-net: Convolutional networks for
biomedical image segmentation," in Medical Image Computing and
Computer-Assisted Intervention–MICCAI 2015: 18th International Conference,
Munich, Germany, October 5-9, 2015, Proceedings, Part III 18, 2015: Springer,
pp. 234-241.
[7] B. Xu, N. Wang, T. Chen, and M. Li, "Empirical evaluation of rectified
activations in convolutional network," arXiv preprint arXiv:1505.00853, 2015.
[8] D. P. Kingma and J. Ba, "Adam: A method for stochastic optimization," arXiv
preprint arXiv:1412.6980, 2014.
[9] I. Goodfellow et al., "Generative adversarial networks," Communications of
the ACM, vol. 63, no. 11, pp. 139-144, 2020.
[10] Q. Mao, H.-Y. Lee, H.-Y. Tseng, S. Ma, and M.-H. Yang, "Mode seeking
generative adversarial networks for diverse image synthesis," in Proceedings
of the IEEE/CVF conference on computer vision and pattern recognition, 2019,
pp. 1429-1437.
[11] X. Wang et al., "Esrgan: Enhanced super-resolution generative adversarial
41
networks," in Proceedings of the European conference on computer vision
(ECCV) workshops, 2018, pp. 0-0.
[12] K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image
recognition," in Proceedings of the IEEE conference on computer vision and
pattern recognition, 2016, pp. 770-778.
[13] C. Ledig et al., "Photo-realistic single image super-resolution using a
generative adversarial network," in Proceedings of the IEEE conference on
computer vision and pattern recognition, 2017, pp. 4681-4690.