Kakkos S K, Stevens J M, Nicolaides A N, Kyriacou E, Pattichis C S, Geroulakos G, Thomas D
Department of Vascular Surgery, Imperial College, London, UK.
Eur J Vasc Endovasc Surg. 2007 Apr;33(4):422-9. doi: 10.1016/j.ejvs.2006.10.018. Epub 2006 Dec 11.
The aim of our study was to determine the association between objective, computerised texture analysis of carotid plaque ultrasonic images and embolic CT-brain infarction in patients presenting with hemispheric neurological symptoms.
Cross-sectional study in patients with 50%-99% (ECST) carotid stenosis.
Carotid plaque ultrasonic images (n=54, 26 with TIAs and 28 with stroke) obtained during carotid ultrasound were normalised and standardised for resolution and subsequently assessed visually for the presence of discrete echogenic or juxtaluminal echolucent components and overall echogenicity (plaque type). Using computer software, 51 histogram/textural features of the plaque outlines were calculated. Factor analysis was subsequently applied to eliminate redundant variables. Small cortical, large cortical and discrete subcortical infarcts on CT-brain scan were considered as being embolic.
Twenty-five cases (46%) had embolic infarcts. On logistic regression, grey-scale median (GSM), a measure of echolucency, spatial grey level dependence matrices (SGLDM) correlation and SGLDM information measure of correlation-1, measures of homogeneity were significant (p<0.05), but not grey level runlength statistics (RUNL) Run Percentage (RP), stenosis severity, type of symptoms or echolucent juxtaluminal components. Using ROC curves methodology, SGLDM information measure of correlation-1 improved the value of GSM in distinguishing embolic from non-embolic CT-brain infarction.
Computerised texture analysis of ultrasonic images of symptomatic carotid plaques can identify those that are associated with brain infarction, improving the results achieved by GSM alone. This methodology could be applied to prospective natural history studies of symptomatic patients not operated on or randomised trials of patients undergoing carotid angioplasty and stenting in order to identify high-risk subgroups for cerebral infarction.
我们研究的目的是确定出现半球性神经症状的患者中，颈动脉斑块超声图像的客观计算机纹理分析与CT脑栓塞性梗死之间的关联。
对颈动脉狭窄程度为50%-99%（欧洲颈动脉外科试验标准）的患者进行横断面研究。
对在颈动脉超声检查期间获得的颈动脉斑块超声图像（n=54，26例有短暂性脑缺血发作，28例有中风）进行分辨率归一化和标准化处理，随后对离散的回声增强或管腔旁无回声成分的存在情况以及整体回声性（斑块类型）进行视觉评估。使用计算机软件计算斑块轮廓的51个直方图/纹理特征。随后应用因子分析来消除冗余变量。CT脑扫描上的小皮质梗死、大皮质梗死和离散的皮质下梗死被视为栓塞性梗死。
25例（46%）有栓塞性梗死。在逻辑回归分析中，作为无回声性度量的灰度中位数（GSM）、空间灰度级依赖矩阵（SGLDM）相关性以及SGLDM相关性信息度量-1（均为同质性度量）具有显著意义（p<0.05），但灰度级游程长度统计（RUNL）的游程百分比（RP）、狭窄严重程度、症状类型或管腔旁无回声成分则不然。使用ROC曲线方法，SGLDM相关性信息度量-1提高了GSM在区分栓塞性与非栓塞性CT脑梗死方面的价值。
有症状颈动脉斑块超声图像的计算机纹理分析能够识别出与脑梗死相关的斑块，改进仅使用GSM所获得的结果。这种方法可应用于未接受手术的有症状患者的前瞻性自然史研究，或应用于接受颈动脉血管成形术和支架置入术患者的随机试验，以识别脑梗死的高危亚组。
Zotero 插件