在台灣，肺癌(Lung Cancer)的罹患率是十大癌症中的第二名，並且連續十二年佔據癌症死亡率之首。肺癌初期不易察覺，但經醫學證明，肺癌篩檢可有效降低死亡率及提高存活率。為了協助醫師在電腦斷層掃瞄影像(Low-Dose Computed Tomography, LDCT)中找出肺結節，本論文改善YoloV4，以完成偵測肺結節(Lung Nodule)的任務。由骨幹(Backbone)網路－CSPDarkNet53進行不同大小的特徵擷取；接著經過頸部(Neck)架構－FPN、PAN更進一步擷取特徵同時與骨幹網路做資訊融合，保留經過多卷積層而失去的原始資訊；最後在頭部(Head)架構－Detector根據特徵資訊找出肺結節。另外，我們的貢獻在於，因應醫學影像的特性，如: 解析度較高、目標較小等，且更精細的資訊對於醫學影像極為重要，因此，我們在原本架構上增加了一條連結，連通最底層的C1、C2保留原始資訊，並加上空間金字塔池化模組(Spatial Pyramid Pooling Module, SPP Module)增強原始資訊。完成偵測網路後，我們還提出了使用形態學(Morphological Processing)影像處理基於冠狀切面(Coronal-View)影像去除水平切面(Axial-View)非肺部影像，來減少肺外誤判(False Positive)，以及基於3D資訊減少血管誤判。最後，針對主要偵測目標，召回率(Recall)為81.0%和準確率(Precision)為87.4%。

In Taiwan, the probability of having lung cancer ranks second among the top ten cancers, and the death rate is the first for 12 consecutive years. Lung cancer is not easy to detect at the beginning, but we can do lung cancer examination to reduce death rate and increase survival rate. To help doctors to find lung nodules in Low-Dose Computed Tomography (LDCT) images, we improve YoloV4 to complete the task of detecting lung nodules. We use CSPDarkNet53 as the backbone to extract features in different sizes. Then, using FPN and PAN as the neck to extract more features and do the feature fusion from the backbone to keep the original information which lost in multiple convolutional layers. Finally, we detect lung nodules at the head. Due to the characteristics of medical images, such as higher resolution and smaller targets, we add a connection from lower layers (C1 and C2) to keep the original information and use a spatial pyramid pooling module to enhance the information. After the detection network, we remove axial-view non-lung slices based on accumulated coronal-view slices using morphological processing to reduce the false positive at outside of lung and reduce false positive based on blood vessel remove in 3D. Finally, for the main detection target, we got recall rate: 81.0% and precision rate: 87.4%.

[1] A. Bochkovskiy, C.Y. Wang, and H.Y.M. Liao, “YOLOv4: Optimal Speed and Accuracy of Object Detection,” In arXiv, vol. 2004.10934, 2020.
[2] K. Chen, J. Pang, J. Wang, Y. Xiong, X. Li, S. Sun, W. Feng, Z. Liu, J. Shi, W. Ouyang, C.C. Loy, and D. Lin, “Hybrid Task Cascade for Instance Segmentation,” In IEEE Conference on Computer Vision and Pattern Recognition, pp. 4974-4983, 2019.
[3] S.H. Gao, M.M. Cheng, K. Zhao, X.Y. Zhang, M.H. Yang, and P. Torr, “Res2Net: A New Multi-scale Backbone Architecture,” In IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 43, No. 2, pp. 652-662, 2021.
[4] K. He, X. Zhang, S. Ren, and J. Sun, “Deep Residual Learning for Image Recognition,” In Conference on Computer Vision and Pattern Recognition, Vol. 1, pp. 770-778, 2016.
[5] K. He, X. Zhang, S. Ren, and J. Sun, “Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition,” In IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 37, No. 9, pp. 1904-1916, 2015.
[6] J. Hu, L. Shen, S. Albanie, G. Sun, and E. Wu, “Squeeze-and-Excitation Networks,” In Conference on Computer Vision and Pattern Recognition, pp. 7132-7141, 2018.
[7] P.F. Jaeger, S.A.A. Kohl, S. Bickelhaupt, F. Isensee, T.A. Kuder, H. Schlemmer, and K.H. Maier-Hein, “Retina U-Net: Embarrassingly Simple Exploitation of Segmentation Supervision for Medical Object Detection,” In arXiv, vol. 1811.08661, 2018.
[8] Y.L. Li, and S.J. Wang, “HAR-Net: Joint Learning of Hybrid Attention for Single-stage Object Detection,” In arXiv, vol. 1904.11141, 2019.
[9] T.Y. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan, and S. Belongie, “Feature Pyramid Networks for Object Detection,” In Conference on Computer Vision and Pattern Recognition, Vol. 1, pp. 936-944, 2017.
[10] T.Y. Lin, P. Goyal, R. Girshick, K. He, and P. Dollár, “Focal Loss for Dense Object Detection,” In IEEE International Conference on Computer Vision, Vol. 1, pp. 2999-3007, 2017.
[11] M. Liu, Y. Tan, and L. Chen, “Pneumonia detection based on deep neural network Retinanet,” In International Conference on Image, Video Processing and Artificial Intelligence, Vol. 11321, pp. 113210F, 2019.
[12] S. Liu, L. Qi, H.F. Qin, J.P. Shi, and J.Y. Jia, “Path Aggregation Network for Instance Segmentation,” In IEEE Conference on Computer Vision and Pattern Recognition, pp. 8759-8768, 2018.
[13] J. Pang, K. Chen, J. Shi, H. Feng, W. Ouyang, and D. Lin, “Libra R-CNN: Towards Balanced Learning for Object Detection,” In IEEE Conference on Computer Vision and Pattern Recognition, pp. 821-830, 2019.
[14] J. Redmon, and A. Farhadi, “YOLOv3: An Incremental Improvement,” In arXiv, vol. 1804.02767, 2018.
[15] S.Q. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks,” In International Conference on Neural Information Processing Systems, Vol. 1, pp. 91-99, 2015.
[16] O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional Networks for Biomedical Image Segmentation,” In International Conference on Medical Image Computing and Computer-Assisted Intervention, pp. 234-241, 2015.
[17] J. Schlemper, O. Oktay, M. Schaap, M. Heinrich, B. Kainz, B. Glocker, and D. Rueckert, “Attention Gated Networks: Learning to Leverage Salient Regions in Medical Images,” In arXiv, vol. 1808.08114, 2018.
[18] S.Y. Sun, J.M. Pang, J.P. Shi, S. Yi, and W.L. Ouyang, “FishNet: A Versatile Backbone for Image, Region, and Pixel Level Prediction,” In International Conference on Neural Information Processing Systems, pp. 762-772, 2018.
[19] M. Tan, and Q.V. Le, “MixConv: Mixed Depthwise Convolutional Kernels,” In British Machine Vision Conference, pp. 1-13, 2019.
[20] W. Zhu, C. Liu, W. Fan, and X. Xie, “DeepLung: Deep 3D Dual Path Nets for Automated Pulmonary Nodule Detection and Classification,” In IEEE Workshop on Applications of Computer Vision, pp. 673-681, 2018.
[21] M. Zlocha, Q. Dou, and B. Glocker, “Improving RetinaNet for CT Lesion Detection with Dense Masks from Weak RECIST Labels,” In International Conference on Medical Image Computing and Computer-Assisted Intervention, pp. 402-410, 2019.