圖像分割是分析冠狀動脈造影 (CAG) X 光圖像的重要步驟。然而，最近基於深度學習的分割模型通常過於複雜，無法在邊緣設備上高效執行。在本文中，我們提出了一種稱為 ASPSeg(自適應空間切割)的新方法，以加速冠狀動脈造影序列的分割過程。
ASPSeg 是一種自上而下的自適應算法，根據影片內容動態地將圖像分割成塊，確保計算資源集中在感興趣的區域。此外，我們通過利用影像序列的時空資訊來提高 ASPSeg 的性能和準確性。
我們在邊緣節點上評估了 ASPSeg 的性能。實驗結果表明，ASPSeg 可以將CPU 上的計算負載減少近 50%，延遲減少 60%，且分割的準確度下降可以忽略不計。該技術與修剪和量化等其他模型壓縮的方法互補，提供了一種正交的方法來提高模型推理的效率。

Image segmentation is an essential step in analyzing coronary angiography (CAG) X-ray images. However, the recent deep learning-based segmentation models are often too complex to be efficiently executed on edge devices. In this paper, we propose a novel method, called ASPSeg (Adaptive Space-Pruning Segmentation), to accelerate the segmentation process of Coronary Angiography sequences. ASPSeg is a top-down adaptive algorithm that dynamically splits images into patches based on the video content, ensuring that the computational resources are focused on the regions of interest. In addition, we enhance the performance and accuracy of ASPSeg by utilizing spatial-temporal information from video sequences. We evaluated the performance of ASPSeg on edge nodes. The experimental results show that ASPSeg can reduce the computational load by nearly 50% and the latency by 60% on CPU, with negligible accuracy degradation. This technique is complementary to other model reduction methods, such as pruning and quantization, providing an orthogonal approach to improve the efficiency of model inference.

[1] Olaf Ronneberger, Philipp Fischer, and Thomas Brox. “U-Net: Convolutional Networks for Biomedical Image Segmentation”. Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015: 18th International Confer- ence, Munich, Germany, October 5-9, 2015, Proceedings, Part III. Ed. by Nassir Navab et al. Cham: Springer International Publishing, 2015, pp. 234–241. isbn: 978-3-319-24574-4.
[2] Jieneng Chen et al. TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation. 2021. arXiv: 2102.04306 [cs.CV].
[3] Hu Cao et al. Swin-Unet: Unet-like Pure Transformer for Medical Image Seg- mentation. 2021. arXiv: 2105.05537 [eess.IV].
[4] Yann LeCun, John Denker, and Sara Solla. “Optimal Brain Damage”. Advances in Neural Information Processing Systems. Ed. by D. Touretzky. Vol. 2. Morgan- Kaufmann, 1989.
[5] B. Hassibi, D.G. Stork, and G.J. Wolff. “Optimal Brain Surgeon and general network pruning”. IEEE International Conference on Neural Networks. 1993, 293–299 vol.1.
[6] Max Jaderberg, Andrea Vedaldi, and Andrew Zisserman. Speeding up Convo- lutional Neural Networks with Low Rank Expansions. 2014. arXiv: 1405.3866 [cs.CV].
[7] Jian Xue, Jinyu Li, and Yifan Gong. “Restructuring of deep neural network acoustic models with singular value decomposition”. Proc. Interspeech 2013. 2013, pp. 2365–2369.
[8] Yufei Ma et al. “Scalable and modularized RTL compilation of Convolutional Neural Networks onto FPGA”. 2016 26th International Conference on Field Programmable Logic and Applications (FPL). 2016, pp. 1–8.
[9] Vincent Vanhoucke, Andrew Senior, and Mark Z. Mao. “Improving the speed of neural networks on CPUs”. Deep Learning and Unsupervised Feature Learning Workshop, NIPS 2011. 2011.
[10] Benoit Jacob et al. “Quantization and Training of Neural Networks for Effi- cient Integer-Arithmetic-Only Inference”. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). June 2018.
[11] Owais Ali et al. “Implementation of a Modified U-Net for Medical Image Seg- mentation on Edge Devices”. IEEE Transactions on Circuits and Systems II: Express Briefs 69.11 (2022), pp. 4593–4597.
[12] Qingliang Liu, Shuai Zhou, and Jinmei Lai. “EdgeMedNet: Lightweight and Ac- curate U-Net for Implementing Efficient Medical Image Segmentation on Edge Devices”. IEEE Transactions on Circuits and Systems II: Express Briefs 70.12 (2023), pp. 4329–4333.
[13] Xingxiang Tao et al. “A Lightweight Network for Accurate Coronary Artery Segmentation Using X-Ray Angiograms”. en. Front Public Health 10 (May 2022), p. 892418.
[14] Fernando Cervantes-Sanchez et al. “Automatic Segmentation of Coronary Ar- teries in X-ray Angiograms using Multiscale Analysis and Artificial Neural Net- works”. Applied Sciences 9.24 (2019). issn: 2076-3417.
[15] Kaiming He et al. “Deep Residual Learning for Image Recognition”. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). June 2016.
[16] Song Han, Huizi Mao, and William J. Dally. Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Cod- ing. 2016. arXiv: 1510.00149 [cs.CV].
[17] Hao Li et al. “Pruning Filters for Efficient ConvNets”. International Conference on Learning Representations. 2017.
[18] Zhuang Liu et al. “Learning Efficient Convolutional Networks through Network Slimming”. 2017 IEEE International Conference on Computer Vision (ICCV). 2017, pp. 2755–2763.
[19] Ji Lin et al. “Runtime Neural Pruning”. Advances in Neural Information Pro- cessing Systems. Ed. by I. Guyon et al. Vol. 30. Curran Associates, Inc., 2017.
[20] Tara N. Sainath et al. “Low-rank matrix factorization for Deep Neural Net- work training with high-dimensional output targets”. 2013 IEEE International Conference on Acoustics, Speech and Signal Processing. 2013, pp. 6655–6659.
[21] Sridhar Swaminathan et al. “Sparse low rank factorization for deep neural net- work compression”. Neurocomputing 398 (2020), pp. 185–196. issn: 0925-2312.
[22] Emily L Denton et al. “Exploiting Linear Structure Within Convolutional Net- works for Efficient Evaluation”. Advances in Neural Information Processing Sys- tems. Ed. by Z. Ghahramani et al. Vol. 27. Curran Associates, Inc., 2014.
[23] Itay Hubara et al. “Binarized Neural Networks”. Advances in Neural Information Processing Systems. Ed. by D. Lee et al. Vol. 29. Curran Associates, Inc., 2016.
[24] Chenzhuo Zhu et al. “Trained Ternary Quantization”. International Conference on Learning Representations. 2017.
[25] Kuan Wang et al. “HAQ: Hardware-Aware Automated Quantization With Mixed Precision”. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). June 2019.
[26] Yu-Shuo Wang et al. “EFFCA: Enhanced Frangi Filter for Coronary Angiog- raphy Segmentation on Mobile Edge Devices”. 2023 IEEE International Confer- ence on E-health Networking, Application & Services (Healthcom). 2023, pp. 238– 243.
[27] Alejandro F. Frangi et al. “Multiscale vessel enhancement filtering”. Medical Image Computing and Computer-Assisted Intervention — MICCAI’98. Ed. by William M. Wells, Alan Colchester, and Scott Delp. Berlin, Heidelberg: Springer Berlin Heidelberg, 1998, pp. 130–137. isbn: 978-3-540-49563-5.
[28] Andrew G. Howard et al. MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications. 2017. arXiv: 1704.04861 [cs.CV].
[29] Andrew Howard et al. “Searching for MobileNetV3”. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). Oct. 2019.
[30] Mark Sandler et al. “MobileNetV2: Inverted Residuals and Linear Bottlenecks”.
Proceedings of the IEEE Conference on Computer Vision and Pattern Recogni- tion (CVPR). June 2018.
[31] Jason D. Browning and Steven L. Tanimoto. “Segmentation of pictures into regions with a tile-by-tile method”. Pattern Recognition 15.1 (1982), pp. 1–10. issn: 0031-3203.
[32] M.-H. Chen and T. Pavlidis. “Image seaming for segmentation on parallel ar- chitecture”. IEEE Transactions on Pattern Analysis and Machine Intelligence 12.6 (1990), pp. 588–594.
[33] Sang Sun Lee and H.T. Tanaka. “Parallel image segmentation with adaptive mesh”. Proceedings 15th International Conference on Pattern Recognition. ICPR- 2000. Vol. 1. 2000, 635–639 vol.1.
[34] H.T. Tanaka and F. Kishino. “Adaptive mesh generation for surface reconstruc- tion: parallel hierarchical triangulation without discontinuities”. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. 1993, pp. 88– 94.
[35] Hiromi T. Tanaka. “Accuracy-Based Sampling and Reconstruction with Adap- tive Meshes for Parallel Hierarchical Triangulation”. Computer Vision and Image Understanding 61.3 (1995), pp. 335–350. issn: 1077-3142.
[36] Daniel Gritzner and Jörn Ostermann. SegForestNet: Spatial-Partitioning-Based Aerial Image Segmentation. 2024. arXiv: 2302.01585 [cs.CV].
[37] Daniel Gritzner and Jörn Ostermann. “Semantic Segmentation of Aerial Images Using Binary Space Partitioning”. KI 2021: Advances in Artificial Intelligence. Ed. by Stefan Edelkamp, Ralf Möller, and Elmar Rueckert. Cham: Springer International Publishing, 2021, pp. 116–134. isbn: 978-3-030-87626-5.
[38] Mingxin Jin et al. “Low-rank and sparse decomposition with spatially adaptive filtering for sequential segmentation of 2D+t vessels”. Physics in Medicine & Biology 63.17 (Aug. 2018), 17LT01.
[39] Mingxin Jin et al. “Extracting contrast-filled vessels in X-ray angiography by graduated RPCA with motion coherency constraint”. Pattern Recognition 63 (2017), pp. 653–666. issn: 0031-3203.
[40] Hua Ma et al. “Automatic online layer separation for vessel enhancement in X- ray angiograms for percutaneous coronary interventions”. Medical Image Anal- ysis 39 (2017), pp. 145–161. issn: 1361-8415.
[41] Binjie Qin et al. “Accurate vessel extraction via tensor completion of background layer in X-ray coronary angiograms”. Pattern Recognition 87 (2019), pp. 38–54. issn: 0031-3203.
[42] Binjie Qin et al. “Robust PCA Unrolling Network for Super-Resolution Vessel Extraction in X-Ray Coronary Angiography”. IEEE Transactions on Medical Imaging 41.11 (2022), pp. 3087–3098.
[43] Shuang Song et al. “Spatio-Temporal Constrained Online Layer Separation for Vascular Enhancement in X-Ray Angiographic Image Sequence”. IEEE Trans- actions on Circuits and Systems for Video Technology 30.10 (2020), pp. 3558– 3570.
[44] Shaoyan Xia et al. “Vessel Segmentation of X-Ray Coronary Angiographic Image Sequence”. IEEE Transactions on Biomedical Engineering 67.5 (2020), pp. 1338– 1348.
[45] Emmanuel J. Candes et al. Robust Principal Component Analysis? 2009. arXiv: 0912.3599 [cs.IT].
[46] Lu Wang et al. “Coronary artery segmentation in angiographic videos utilizing spatial-temporal information”. BMC Medical Imaging 20.1 (Sept. 2020), p. 110.
[47] National Taiwan University Medical Imaging. CAG biomedical-images. https: //scidm.nchc.org.tw/dataset/cag. 2021.
[48] Ilya Loshchilov and Frank Hutter. “Decoupled Weight Decay Regularization”. International Conference on Learning Representations. 2019.
[49] Diederik P. Kingma and Jimmy Ba. “Adam: A Method for Stochastic Optimiza- tion.” ICLR (Poster). Ed. by Yoshua Bengio and Yann LeCun. 2015.
[50] Lucas Liebenwein et al. “Compressing Neural Networks: Towards Determining the Optimal Layer-wise Decomposition”. Advances in Neural Information Pro- cessing Systems. Ed. by A. Beygelzimer et al. 2021.
[51] Xiangyu Zhang et al. “Accelerating Very Deep Convolutional Networks for Clas- sification and Detection”. IEEE Transactions on Pattern Analysis and Machine Intelligence 38.10 (2016), pp. 1943–1955.
[52] Cenk Baykal et al. “Sensitivity-Informed Provable Pruning of Neural Networks”. SIAM Journal on Mathematics of Data Science 4.1 (2022), pp. 26–45. eprint: https://doi.org/10.1137/20M1383239.
[53] Lucas Liebenwein et al. “Provable Filter Pruning for Efficient Neural Networks”. International Conference on Learning Representations. 2020.
[54] Zongwei Zhou et al. “UNet++: A Nested U-Net Architecture for Medical Image Segmentation”. Deep Learning in Medical Image Analysis and Multimodal Learn- ing for Clinical Decision Support. Lecture notes in computer science. Cham: Springer International Publishing, 2018, pp. 3–11.