近年來，深度學習技術在醫學領域發展逐漸被廣泛地應用；它能協助醫師提升診斷的準確率以及減少他們在影像判讀上的負擔。然而，針對特定任務從頭訓練一個深度學習模型往往需要大量的訓練資料。然而，蒐集醫學資料比起一般資料還要困難需多，也降低深度學習技術在許多臨床應用的可行性。為了克服這個挑戰，我們使用元學習(meta-learning)來降低深度學習模型對於訓練資料量的要求。在本篇論文中，我們在醫學影像切割的問題上使用一種稱作model-agnostic meta-learning (MAML) 的元學習演算法。由於元學習演算法需要經由不同的資料集(即任務)來學習，我們使用了五個公開的醫學影像資料集(包含了磁振造影及電腦斷層掃描的資料集)來評估，在不同條件下MAML演算法的表現。最後，在使用MAML的情況下，我們在第一節腰椎的電腦斷層影像的影像切割任務中，只用了60張訓練影像，便得到了dice coefficient (DC) 90.6%的成果。然而，如果沒有使用MAML，在同樣使用60張訓練影像時，傳統深度學習模型只能得到DC 77.9%的結果。與電腦斷層掃描影像分割結果相似，在使用MAML的情況下，在第一節腰椎的磁振造影的影像切割任務中，我們只用了60張訓練影像就達成DC 89.1%的成果；而沒有使用MAML則是只有DC 67.8%的準確率。總結來說，本研究展示了在醫學影像分割問題中使用元學習技術，將能有效降低對於訓練資料數量的要求，從而提高人工智慧技術在許多臨床應用的實用性。

In recent years, deep learning techniques are widely applied in medical applications. These techniques allow physicians to increase diagnostic accuracy and reduce their burden. However, training a deep learning model for a specific task from scratch frequently requires plenty of training data. However, collection of medical data is more difficult than general data, which lowers practicality of deep learning techniques in many clinical applications. To overcome this challenge, we propose to use meta-learning techniques to reduce amount of required training data. In our study, a meta-learning method, named as model-agnostic meta-learning (MAML), is applied to tackle problems of medical image segmentation. Because a meta-learning algorithm learns from various datasets (i.e., tasks), we use 5 public medical image datasets acquired with two imaging modalities (MRI and CT) to evaluate our method under different conditions. Finally, after using MAML, we achieve Dice coefficient of 90.6% in a CT Lumbar vertebra 1 (CT_L1) segmentation task with 60 training images. In contrast, we get a dice coefficient (DC) of 77.9% under the same experimental condition using a conventional deep learning model. Similarly, we achieve a DC of 89.1 % in a MRI Lumbar vertebra 1 (MRI_L1) segmentation task with 60 training images, and only achieve a DC of 67.8% with a plain deep learning model. In conclusion, our study demonstrates that applying meta-learning techniques for medical image segmentation effectively reduces requirement of large training datasets, which greatly improves practicality of artificial intelligence techniques in clinical applications.

1. He, K., et al. Deep residual learning for image recognition. in Proceedings of the IEEE conference on computer vision and pattern recognition. 2016.
2. Simonyan, K. and A. Zisserman, Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556, 2014.
3. Lin, T.-Y., et al. Focal loss for dense object detection. in Proceedings of the IEEE international conference on computer vision. 2017.
4. 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.
5. Chen, L.-C., et al., Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs. IEEE transactions on pattern analysis and machine intelligence, 2017. 40(4): p. 834-848.
6. Dai, J., K. He, and J. Sun. Boxsup: Exploiting bounding boxes to supervise convolutional networks for semantic segmentation. in Proceedings of the IEEE international conference on computer vision. 2015.
7. Lin, D., et al. Scribblesup: Scribble-supervised convolutional networks for semantic segmentation. in Proceedings of the IEEE conference on computer vision and pattern recognition. 2016.
8. Koch, G., R. Zemel, and R. Salakhutdinov. Siamese neural networks for one-shot image recognition. in ICML deep learning workshop. 2015. Lille.
9. Wang, T., et al. Few-shot adaptive faster r-cnn. in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019.
10. Zhang, C., et al. Canet: Class-agnostic segmentation networks with iterative refinement and attentive few-shot learning. in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019.
11. Siam, M., B.N. Oreshkin, and M. Jagersand. Amp: Adaptive masked proxies for few-shot segmentation. in Proceedings of the IEEE/CVF International Conference on Computer Vision. 2019.
12. Janssens, R., G. Zeng, and G. Zheng. Fully automatic segmentation of lumbar vertebrae from CT images using cascaded 3D fully convolutional networks. in 2018 IEEE 15th International Symposium on Biomedical Imaging (ISBI 2018). 2018. IEEE.
13. Lessmann, N., et al., Iterative fully convolutional neural networks for automatic vertebra segmentation and identification. Medical image analysis, 2019. 53: p. 142-155.
14. Finn, C., P. Abbeel, and S. Levine. Model-agnostic meta-learning for fast adaptation of deep networks. in International Conference on Machine Learning. 2017. PMLR.
15. Zhao, H., et al. Pyramid scene parsing network. in Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.
16. Howard, A.G., et al., Mobilenets: Efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861, 2017.
17. Tan, M. and Q. Le. Efficientnet: Rethinking model scaling for convolutional neural networks. in International Conference on Machine Learning. 2019. PMLR.
18. Ronneberger, O., P. Fischer, and T. Brox. U-net: Convolutional networks for biomedical image segmentation. in International Conference on Medical image computing and computer-assisted intervention. 2015. Springer.
19. Shyam, P., S. Gupta, and A. Dukkipati. Attentive recurrent comparators. in International conference on machine learning. 2017. PMLR.
20. Shaban, A., et al., One-shot learning for semantic segmentation. arXiv preprint arXiv:1709.03410, 2017.
21. Rakelly, K., et al., Conditional networks for few-shot semantic segmentation. 2018.
22. Boudiaf, M., et al., Transductive information maximization for few-shot learning. arXiv preprint arXiv:2008.11297, 2020.
23. Boudiaf, M., et al. Few-Shot Segmentation Without Meta-Learning: A Good Transductive Inference Is All You Need? in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2021.
24. Li, Z., et al., Meta-sgd: Learning to learn quickly for few-shot learning. arXiv preprint arXiv:1707.09835, 2017.
25. Fu, K., et al., Meta-SSD: Towards fast adaptation for few-shot object detection with meta-learning. IEEE Access, 2019. 7: p. 77597-77606.
26. Andrychowicz, M., et al., Learning to learn by gradient descent by gradient descent. arXiv preprint arXiv:1606.04474, 2016.
27. Ravi, S. and H. Larochelle, Optimization as a model for few-shot learning. 2016.
28. Santoro, A., et al. Meta-learning with memory-augmented neural networks. in International conference on machine learning. 2016. PMLR.
29. Wang, Y.-X. and M. Hebert. Learning to learn: Model regression networks for easy small sample learning. in European Conference on Computer Vision. 2016. Springer.
30. Sung, F., et al. Learning to compare: Relation network for few-shot learning. in Proceedings of the IEEE conference on computer vision and pattern recognition. 2018.
31. Mishra, N., et al., A simple neural attentive meta-learner. arXiv preprint arXiv:1707.03141, 2017.
32. Ibragimov, B., et al., Shape representation for efficient landmark-based segmentation in 3-D. IEEE transactions on medical imaging, 2014. 33(4): p. 861-874.
33. Korez, R., et al., A framework for automated spine and vertebrae interpolation-based detection and model-based segmentation. IEEE transactions on medical imaging, 2015. 34(8): p. 1649-1662.
34. Chu, C., et al., Annotated T2-weighted MR images of the Lower Spine. 2015.
35. Chu, C., et al., Fully automatic localization and segmentation of 3D vertebral bodies from CT/MR images via a learning-based method. PloS one, 2015. 10(11): p. e0143327.
36. Bakas, S., et al., Advancing the cancer genome atlas glioma MRI collections with expert segmentation labels and radiomic features. Scientific data, 2017. 4(1): p. 1-13.
37. Menze, B.H., et al., The multimodal brain tumor image segmentation benchmark (BRATS). IEEE transactions on medical imaging, 2014. 34(10): p. 1993-2024.
38. Bakas, S., et al., Identifying the best machine learning algorithms for brain tumor segmentation, progression assessment, and overall survival prediction in the BRATS challenge. arXiv preprint arXiv:1811.02629, 2018.
39. Bakas, S., et al., Segmentation Labels for the Pre-operative Scans of the TCGA-GBM collection. 2017.
40. Bakas, S., et al., Segmentation Labels for the Pre-operative Scans of the TCGA-LGG collection. 2017.
41. Petitjean, C., et al., Right ventricle segmentation from cardiac MRI: a collation study. Medical image analysis, 2015. 19(1): p. 187-202.
42. Russakovsky, O., et al., Imagenet large scale visual recognition challenge. International journal of computer vision, 2015. 115(3): p. 211-252.
43. Tustison, N.J., et al., N4ITK: improved N3 bias correction. IEEE transactions on medical imaging, 2010. 29(6): p. 1310-1320.
44. Dai, J., et al., R-fcn: Object detection via region-based fully convolutional networks. arXiv preprint arXiv:1605.06409, 2016.
45. Srivastava, N., et al., Dropout: a simple way to prevent neural networks from overfitting. The journal of machine learning research, 2014. 15(1): p. 1929-1958.
46. Hendryx, S.M., et al., Meta-learning initializations for image segmentation. arXiv preprint arXiv:1912.06290, 2019.