光学相干断层扫描技术（OCT）已成为诊断冠状动脉狭窄的关键技术，因其可识别影像中的斑块及易损斑块，所以该技术对辅助诊断冠心病具有十分重要的意义。但当前研究领域内尚缺乏对冠脉 OCT 图像全自动、多区域、高精度的分割算法。因此本文提出了一种基于中智学理论的冠脉 OCT 图像的多区域、全自动的分割算法，以期实现对纤维斑块和脂质区的高精度分割。本文基于隶属度函数重新定义了 OCT 图像转换至中智学领域 T 图的方法，进而达到提高纤维斑块的分割精度的目的。针对脂质区的分割，本算法加入同态滤波增强图像，使用中智学将 OCT 图像转换至中智学领域的 I 图，进一步使用形态学方法，实现高精度的分割。本文对 9 位患者、40 组具有典型斑块的 OCT 图像进行分析，并与医生手动分割结果进行比较，实验结果证明，本文算法避免了传统中智学的过分割及欠分割问题，准确地分割出斑块区域，且算法具有较好的鲁棒性，因此本文工作或可有效提高医生分割斑块的准确率，期望可以辅助临床医生对冠心病的诊断与治疗。
Optical coherence tomography (OCT) has become a key technique in the diagnosis of coronary artery stenosis, which can identify plaques and vulnerable plaques in the image. Therefore, this technique is of great significance for the diagnosis of coronary heart disease. However, there is still a lack of automatic, multi-region, high-precision segmentation algorithms for coronary OCT images in the current research field. Therefore, this paper proposes a multi-zone, fully automated segmentation algorithm for coronary OCT images based on neutrosophic theory, which achieves high-precision segmentation of fibrous plaques and lipid regions. In this paper, the method of transforming OCT images into T in the area of neutrosophics is redefined based on the membership function, and the segmentation accuracy of fiber plaques is improved. For the segmentation of lipid regions, the algorithm adds homomorphic filter enhancement images, and uses OCT to transform OCT images into I in the field of neutrosophics, and further uses morphological methods to achieve high-precision segmentation. In this paper, 40 OCT images from 9 patients with typical plaques were analyzed and compared with the results of manual segmentation by doctors. Experiments show that the proposed algorithm avoids the over-segmentation and under-segmentation problems of the traditional neutrosophic theory method, and accurately segment the patch area. Therefore, the work of this paper can effectively improve the accuracy of segmentation of plaque for doctors, and assist clinicians in the diagnosis and treatment of coronary heart disease.