基于多模态超声组学对颈动脉内膜剥脱术前评估与术后脑缺血风险预测机制研究

Internal carotid artery stenosis and occlusion is one of the major causes of ischemic stroke. Carotid endarterectomy (CEA) can significantly reduce the risk of ischemic stroke in patients with carotid stenosis. However, the potential cerebral ischemia after CEA greatly reduces the therapeutic benefits of patients. Therefore, elucidating the mechanism of cerebral ischemia after CEA and early screening of high risk patients who may develop cerebral ischemia after CEA can assist clinicians to take preventive measures as early as possible and finally, improve the prognosis of patients. Present research regarding the risk factors of cerebral ischemia after CEA mainly focuses on clinical risk factors, and there are few reports on carotid artery structure, hemodynamics, and the atherosclerotic plaques. In recent years, medical imaging has made a great breakthrough in clinical application by adopting artificial intelligence to mine disease information and quantitatively extract medical image features. Grey-scale ultrasound radiomics has developed into to contrast-enhanced ultrasound radiomics, and single-modal has developed into multi-modal radiomics. Based on the current clinical demand and the rapid development of imaging technology, we aim to quantify the characteristics of multimodal ultrasound images of patients using the deep learning technology based on convolution neural network, and finally, to construct a model for predicting the risk of cerebral ischemia after CEA based on multi-modal ultrasound radiomics.

颈内动脉狭窄、闭塞是导致缺血性脑卒中的重要原因之一。颈动脉内膜剥脱术(carotid endarterectomy,CEA)可显著降低颈动脉狭窄患者缺血性卒中风险，但CEA后可能发生的脑缺血事件大大降低患者的治疗获益。因此，阐明CEA后脑缺血发生机制、早期筛查CEA后可能发生脑缺血的高危患者，可辅助临床医生尽早采取预防手段、改善患者预后。目前CEA后脑缺血风险研究主要集中于临床风险对术后脑缺血的影响，对颈动脉结构、血流动力学、斑块参与CEA后脑缺血的机制及预后价值报道较少。近年来，影像组学利用人工智能方法挖掘疾病信息、定量提取医学影像特征，临床应用取得重大突破。超声组学由灰阶超声组学发展至超声造影组学；单模态发展至多模态超声组学。基于目前临床需求及影像组学技术的迅速发展，拟利用基于卷积神经网络的深度学习技术将患者多模态超声图像特征定量化，构建基于多模态超声组学的CEA后脑缺血风险预测模型。

背景 本研究拟探索颈动脉斑块超声特征在预测颈动脉内膜切除（carotid endarterectomy, CEA）术后新发脑缺血病变中的应用价值。.主要研究内容 连续纳入1061例在首都医科大学宣武医院神经外科行颈CEA患者。所有患者在CEA术前行颈动脉彩色多普勒超声检查、脑磁共振弥散加权成像(diffusion-weighted imaging, DWI)，以评估颈动脉血管结构、血流动力学特征、颈动脉斑块特征、术前有无脑梗死。所有患者CEA术后30天内行DWI检查明确有无新发脑缺血灶。采用Adobe Photoshop CS5软件对颈动脉斑块灰阶中位值（gray-scale median, GSM）进行量化。根据患者CEA术后脑DWI成像将患者分为两组：新发脑缺血组、无新发脑缺血组。构建基于患者临床特征、血管结构、血流动力学特征、斑块特征的CEA术后新发脑缺血预测模型。.重要结果及关键数据 共169例患者CEA术后30天内出现新发脑缺血灶（15.9%）。与无新发脑缺血患者相比，新发脑缺血患者平均年龄大(64±7岁比 63±8岁，P=0.022)、症状性颈动脉狭窄患者占比高(79.3% 比 71.9%，P=0.046)、术前脑梗死发生率高(72.8% 比 62.6%，P=0.011) 、斑块GSM低（33.9±22.0 比 65.4±21.5，P<0.001)、溃疡性斑块检出率高(33.1% vs. 22.2%，P=0.002)。责任斑块GSM诊断CEA术后新发脑缺血的ROC曲线下面积为0.837。约登指数最大时对应的GSM临界值为30.5，此时GSM诊断CEA术后新发脑缺血的敏感度为68.6%，特异度为93.2%。与GSM＞30.5的脑缺血患者相比，GSM≤30.5的脑缺血患者CEA术后多发性脑缺血病变的发生率更高(59.5% 比 41.5%，P=0.030)。单因素和多因素分析证实，GSM≤30.5，溃疡斑块、症状性狭窄为CEA术后新发脑缺血灶的危险因素。 .科学意义 基于本研究构建的预测模型可辅助临床医生在CEA术前明确术后新发脑缺血高风险患者，为患者选择最佳治疗策略，制定个体化的治疗及随访方案。