基于深度学习的冠状动脉IVOCT图像中易损斑块识别算法研究进展

Advances in deep learning-based algorithms for recognizing vulnerable plaque in coronary intravascular optical coherence tomography image

介绍了冠状动脉血管内光学相干断层成像(intravascular optical coherence tomography,IVOCT)图像中易损斑块的智能识别算法的研究现状,分析了传统机器学习算法在易损斑块识别领域的不足及深度学习算法的优势。从斑块分类、检测、分割3个方面阐述了基于深度学习的斑块识别网络的结构特征及针对性的结构改进方法,指出了现有基于深度学习的识别算法的局限性,提出了提升识别精度、理解易损斑块破裂机理是该领域的未来发展方向。

The development status of intelligent recognition algorithms of the vulnerable plaque in the coronary intravascular optical coherence tomography(IVOCT)image was introduced firstly.The shortcomings of traditional machine learning-based algorithms in the field of vulnerable plaque recognition and the advantages of deep learning-based algorithm were explained.The structural characteristics and targeted structural improvement of plaque recognition network were described under classification,detection and segmentation tasks.The limitations of the existing recognition algorithms and the development directions of improving accuracy and understanding the mechanism of vulnerable plaque rupture were discussed.