Roberts Gemma, Kane Joseph P M, Lloyd Jim J, Petrides George S, Howe Kim, O'Brien John T, Thomas Alan J
Institute of Neuroscience, Newcastle University.
Nuclear Medicine Department, Royal Victoria Infirmary, Newcastle upon Tyne.
Nucl Med Commun. 2019 Jul;40(7):734-743. doi: 10.1097/MNM.0000000000001024.
Cardiac I-MIBG imaging is an established technique for the diagnosis of dementia with Lewy bodies but various analysis methods are reported in the literature. We assessed different methods in the same cohort of patients to inform best practice.
Seventeen patients with dementia with Lewy bodies, 15 with Alzheimer's disease and 16 controls were included. Planar images were acquired 20 min and 4 h after injection. Nine operators produced heart-to-mediastinum ratios (HMRs) using freehand and 6, 7 and 8 cm diameter circular cardiac regions. Interoperator variation was measured using the coefficient of variation. HMR differences between methods were assessed using analysis of variance. Seven raters assessed the images visually. Accuracy was compared using receiver operating characteristic analysis.
There were significant differences in HMR between region methods (P=0.006). However, with optimised cut-offs there was no significant difference in accuracy (P=0.2-1.0). The sensitivity was 65-71% and specificity 100% for all HMR methods. Variation was lower with fixed regions than freehand (P<0.001). Visual rating sensitivity and specificity were 65 and 77% on early images and 76 and 71% on delayed images. There was no significant difference in HMR between early and delayed images (P=0.4-0.7) although a greater separation between means was seen on delayed images (0.73 vs. 0.95).
HMR analysis using a suitable cut-off is more accurate than visual rating. Accuracy is similar for all methods, but freehand regions are more variable and 6 cm circles easiest to place. We recommend calculating HMR using a 6 cm circular cardiac region of interest on delayed images.
心肌碘代苄胍（I-MIBG）显像术是诊断路易体痴呆的一项成熟技术，但文献报道了多种分析方法。我们在同一组患者中评估了不同方法，以指导最佳实践。
纳入17例路易体痴呆患者、15例阿尔茨海默病患者和16名对照者。注射后20分钟和4小时采集平面图像。9名操作人员使用徒手画以及直径为6、7和8厘米的圆形心脏感兴趣区得出心纵隔比（HMR）。使用变异系数测量操作者间的差异。采用方差分析评估不同方法之间的HMR差异。7名评估者对图像进行视觉评估。使用受试者操作特征分析比较准确性。
不同感兴趣区方法之间的HMR存在显著差异（P = 0.006）。然而，采用优化的临界值时，准确性无显著差异（P = 0.2 - 1.0）。所有HMR方法的敏感性为65% - 71%，特异性为100%。固定感兴趣区的变异低于徒手画（P < 0.001）。早期图像的视觉评估敏感性和特异性分别为65%和77%，延迟图像分别为76%和71%。早期和延迟图像之间的HMR无显著差异（P = 0.4 - 0.7），尽管延迟图像上均值之间的分离更大（0.73对0.95）。
使用合适的临界值进行HMR分析比视觉评估更准确。所有方法的准确性相似，但徒手画感兴趣区的变异性更大，6厘米的圆最容易放置。我们建议在延迟图像上使用直径6厘米的圆形心脏感兴趣区计算HMR。
