当前位置: X-MOL 学术J Nucl. Med. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Deep Learning–Based Attenuation Correction Improves Diagnostic Accuracy of Cardiac SPECT
The Journal of Nuclear Medicine ( IF 9.3 ) Pub Date : 2023-03-01 , DOI: 10.2967/jnumed.122.264429
Aakash D Shanbhag 1 , Robert J H Miller 2 , Konrad Pieszko 3 , Mark Lemley 1 , Paul Kavanagh 1 , Attila Feher 4 , Edward J Miller 4 , Albert J Sinusas 4 , Philipp A Kaufmann 5 , Donghee Han 1 , Cathleen Huang 1 , Joanna X Liang 1 , Daniel S Berman 1 , Damini Dey 1 , Piotr J Slomka 6
Affiliation  

To improve diagnostic accuracy, myocardial perfusion imaging (MPI) SPECT studies can use CT-based attenuation correction (AC). However, CT-based AC is not available for most SPECT systems in clinical use, increases radiation exposure, and is impacted by misregistration. We developed and externally validated a deep-learning model to generate simulated AC images directly from non-AC (NC) SPECT, without the need for CT. Methods: SPECT myocardial perfusion imaging was performed using 99mTc-sestamibi or 99mTc-tetrofosmin on contemporary scanners with solid-state detectors. We developed a conditional generative adversarial neural network that applies a deep learning model (DeepAC) to generate simulated AC SPECT images. The model was trained with short-axis NC and AC images performed at 1 site (n = 4,886) and was tested on patients from 2 separate external sites (n = 604). We assessed the diagnostic accuracy of the stress total perfusion deficit (TPD) obtained from NC, AC, and DeepAC images for obstructive coronary artery disease (CAD) with area under the receiver-operating-characteristic curve. We also quantified the direct count change among AC, NC, and DeepAC images on a per-voxel basis. Results: DeepAC could be obtained in less than 1 s from NC images; area under the receiver-operating-characteristic curve for obstructive CAD was higher for DeepAC TPD (0.79; 95% CI, 0.72–0.85) than for NC TPD (0.70; 95% CI, 0.63–0.78; P < 0.001) and similar to AC TPD (0.81; 95% CI, 0.75–0.87; P = 0.196). The normalcy rate in the low-likelihood-of-coronary-disease population was higher for DeepAC TPD (70.4%) and AC TPD (75.0%) than for NC TPD (54.6%, P < 0.001 for both). The positive count change (increase in counts) was significantly higher for AC versus NC (median, 9.4; interquartile range, 6.0–14.2; P < 0.001) than for AC versus DeepAC (median, 2.4; interquartile range, 1.3–4.2). Conclusion: In an independent external dataset, DeepAC provided improved diagnostic accuracy for obstructive CAD, as compared with NC images, and this accuracy was similar to that of actual AC. DeepAC simplifies the task of artifact identification for physicians, avoids misregistration artifacts, and can be performed rapidly without the need for CT hardware and additional acquisitions.



中文翻译:

基于深度学习的衰减校正提高了心脏 SPECT 的诊断准确性

为了提高诊断准确性,心肌灌注成像 (MPI) SPECT 研究可以使用基于 CT 的衰减校正 (AC)。然而,基于 CT 的 AC 不适用于临床使用的大多数 SPECT 系统,会增加辐射暴露,并受到配准不当的影响。我们开发并外部验证了一种深度学习模型,可直接从非 AC (NC) SPECT 生成模拟 AC 图像,无需 CT。方法:使用99m Tc-sestamibi 或99m Tc-tetrofosmin 在带有固态探测器的现代扫描仪上进行 SPECT 心肌灌注成像。我们开发了一种条件生成对抗神经网络,该网络应用深度学习模型 (DeepAC) 来生成模拟 AC SPECT 图像。该模型使用在 1 个站点 ( n = 4,886)执行的短轴 NC 和 AC 图像进行训练,并在来自 2 个独立外部站点的患者 ( n = 604) 上进行测试。我们评估了从 NC、AC 和 DeepAC 图像获得的应激性总灌注不足 (TPD) 对阻塞性冠状动脉疾病 (CAD) 的诊断准确性,以及接受者操作特征曲线下的面积。我们还按体素量化了 AC、NC 和 DeepAC 图像之间的直接计数变化。结果:可以在不到1秒的时间内从NC图像中获得DeepAC;DeepAC TPD 的阻塞性 CAD 受试者工作特征曲线下面积 (0.79; 95% CI, 0.72–0.85) 高于 NC TPD (0.70; 95% CI, 0.63–0.78; P < 0.001),并且与AC TPD(0.81;95% CI,0.75–0.87;P = 0.196)。DeepAC TPD (70.4%) 和 AC TPD (75.0%) 的冠状动脉疾病低可能性人群的正常率高于 NC TPD (54.6%,两者 P < 0.001 ) AC 与 NC 的阳性计数变化(计数增加)(中位数,9.4;四分位距,6.0-14.2;P < 0.001)显着高于 AC 与 DeepAC(中位数,2.4;四分位距,1.3-4.2)。结论:在独立的外部数据集中,与 NC 图像相比,DeepAC 提高了阻塞性 CAD 的诊断准确性,并且该准确性与实际 AC 的诊断准确性相似。DeepAC 简化了医生的伪影识别任务,避免了配准错误的伪影,并且可以快速执行,无需 CT 硬件和额外的采集。

更新日期:2023-03-02
down
wechat
bug