基于人工智能分析颈内动脉颅外段迂曲特征及对称性的应用性评价

The tortuosity and symmetry characteristics of extracranial segment of cervical carotid arteries: evaluation of CT angiography by deep learning-based artificial intelligence

目的验证人工智能技术提取的血管形态参数在颈内动脉颅外段(C1段)形态评价中的价值。方法收集2022年3月至11月于首都医科大学朝阳医院急诊进行头颈联合CT血管造影(CTA)检查的200例患者的全部血管资料。首先,通过人工智能深度学习的CTA影像资料,获取C1段定量解剖参数,包括实际长度、相对长度、距离因子度量、拐点计数指标、扭率、迂曲指数、弯曲度、曲率、角度和角度度量值指标10项血管相关参数;其次,由影像医师对C1段的对称性进行视觉评估并依据视觉评估结果将患者分为C1对称组及C1非对称组。通过配对t检验分析C1对称组和C1非对称组左右侧血管形态参数的差异;随后依据Metz分型将C1段分为直型、C型、S型、卷曲、折曲5种类型,使用ANOVA单因素分析不同迂曲类型血管形态之间存在的差异,验证上述10项血管形态学参数在不同迂曲类型C1段间存在的差异。结果200例患者的双侧C1段共400例血管被纳入分析,76例为女性。对C1段的对称性视觉评估结果显示,87例患者为C1对称组,113例患者为C1非对称组。C1对称组中10项血管形态指标左右侧差异均无统计学意义(P均>0.05),C1非对称组中左右侧C1段的相对长度、拐点计数指标和距离因子度量差异具有统计学意义(P<0.05),而其余指标在双侧间差异不存在统计学意义(P>0.05)。对C1段迂曲类型视觉评估结果显示:24例为直型,126例为S型,182例为C型,18例为卷曲,50例为折曲。对400例C1段血管进行迂曲类型分类后,应用人工智能技术提取的相对长度、距离因子度量、拐点计数指标、迂曲指数、弯曲度、曲率、角度和角度度量指标8项血管参数在不同迂曲类型间差异存在统计学意义(P均<0.05),实际长度、扭率在不同迂曲类型间差异无统计学意义(P>0.05)。结论血管形态学参数可以作为辅助评估血管形态的重要指标,这些定量指标可以帮助准确识别复杂的颈内动脉CTA三维图像上血管的走行、形态以及双侧血管对称性。

Objective To explore the value of vascular morphological parameters extracted by artificial intelligence technology in the morphological evaluation of the internal neck arteries (C1) morphological evaluation.MethodsVascular data of 200 patients who underwent emergency head-neck CT angiography (CTA) examination at Beijing Chaoyang Hospital, Capital Medical University from March 2022 to November 2022 were collected. artificial intelligence deep learning was utilized to analyze CTA image data. Quantitative anatomical parameters of the C1 segment of the internal carotid artery were extracted, including the actual length, relative length, distance factor metric, sum of angle metrics, tortuosity index, inflection count metrics, bending length, angle, curvature, torsion. Ten vascular-related parameters in total Secon, a radiologist conducted visual assessment of the symmetry of C1 and categorized the vascular morphology into the symmetry and the asymmetric group according to the visual assessment result. There are differences in the form. Subsequently C1 is divided into five types: "straight", "C-type", "S-type", "curling", and "folding" based on the METZ type. The sample non-parameter test analysis was performed to examine the differences in vascular morphology between the various types and to verify the difference in vascular morphology parameters among the different METZ types.ResultsA total of 400 vessels of bilateral C1 segments from 200 patients were included in the analysis, including 76 females. Visual assessment results of C1 symmetry showed that 87 patients had symmetrical segments, while 113 patients had asymmetrical segments. There were no statistically significant differences in the 10 parameters in the symmetrical C1 group, while statistically significant differences were found in relative length, inflection point count index, and distance factor measurement in the asymmetrical C1 group (P<0.05), and no statistically significant differences were found in the remaining parameters between the bilateral sides (P>0.05). Visual assessment results of C1 tortuosity types showed 24 "straight", 126 "S-shaped", 182 "C-shaped", 18 "coiled", and 50 "folded" cases. After classifying the 400 cases of C1 according to tortuosity types, there were statistically significant differences in the 8 vascular parameters[relative length, distance factor metric, sum of angle metrics, tortuosity index, inflection count metrics, bending length, angle, curvature] extracted by artificial intelligence technology among different tortuosity types, with the exception of actual length or torsion.ConclusionVascular morphological parameters are crucial indicators for auxiliary evaluation of vascular morphology. These quantitative measures facilitate the accurate identification of the course, morphology, and bilateral symmetry of vessels on complex three-dimensional CTA images of cervical carotid arteries.