Deep-Learning Generated Synthetic Double Inversion Recovery Images Improve Multiple Sclerosis Lesion Detection
Deep-Learning Generated Synthetic Double Inversion Recovery Images Improve Multiple Sclerosis Lesion Detection
复制标题
DOI：
10.1097/rli.0000000000000640
复制
发表时间：
2020-05-01
影响因子：
6.7
通讯作者：
Wiestler, Benedikt
中科院分区：
文献类型：
--
作者：
Finck, Tom;Li, Hongwei;Wiestler, Benedikt
该研究的目的是实施一种深度学习工具来生成合成双反转恢复（双反转恢复）图像，并将其诊断性能与多发性硬化症（MS）患者的常规序列进行比较。材料和方法对于这项回顾性分析，从2014年至2016年的前瞻性观察队列中随机选择了100例MS患者（65例女性，37 [22-68]岁）。在50名患者的子集中，训练人工神经网络（DiamondGAN），以根据标准采集（T1、T2和液体衰减反转恢复[FLAIR]）生成合成回波（回波）。利用所得到的网络，为其余50名受试者生成了样本量。这些图像以及常规采集的MRI（真MRI）和FLAIR图像由2名对MRI图像来源不知情的独立阅片者评估MS病变。使用Wilcoxon符号秩检验比较不同模式下的病变计数，并进行评分者间分析。对比噪声比的客观图像质量进行了比较。结果与FLAIR图像相比，MRI能更好地检出病灶（31.4 ± 20.7vs22.8 ± 12.7，P < 0.001）。这种改善主要归因于对皮质病变的显示改善（12.3 +/- 10.8 vs 7.2 +/- 5.6，P < 0.001）。FLAIR 0.92评分者间信度良好（95%置信区间[CI]，0.85-0.95），置信度0.93（95% CI，0.87-0.96），真实可信度0.95（95%CI，0.85-0.98），MRI的对比噪声比优于FLAIR（22.0 +/- 6.4 vs 16.7 +/- 3.6，P = 0.009）;与truealone相比无显著差异（22.0 +/- 6.4 vs 22.4 +/- 7.9，P = 0.87）。结论：与使用标准模式相比，计算生成的MRI图像改善了病变显示。这种方法展示了人工智能如何帮助改善特定病理的成像。
ObjectivesThe aim of the study was to implement a deep-learning tool to produce synthetic double inversion recovery (synthDIR) images and compare their diagnostic performance to conventional sequences in patients with multiple sclerosis (MS). Materials and MethodsFor this retrospective analysis, 100 MS patients (65 female, 37 [22-68] years) were randomly selected from a prospective observational cohort between 2014 and 2016. In a subset of 50 patients, an artificial neural network (DiamondGAN) was trained to generate a synthetic DIR (synthDIR) from standard acquisitions (T1, T2, and fluid-attenuated inversion recovery [FLAIR]). With the resulting network, synthDIR was generated for the remaining 50 subjects. These images as well as conventionally acquired DIR (trueDIR) and FLAIR images were assessed for MS lesions by 2 independent readers, blinded to the source of the DIR image. Lesion counts in the different modalities were compared using a Wilcoxon signed-rank test, and interrater analysis was performed. Contrast-to-noise ratios were compared for objective image quality. ResultsUtilization of synthDIR allowed to detect significantly more lesions compared with the use of FLAIR images (31.4 +/- 20.7 vs 22.8 +/- 12.7, P < 0.001). This improvement was mainly attributable to an improved depiction of juxtacortical lesions (12.3 +/- 10.8 vs 7.2 +/- 5.6, P < 0.001). Interrater reliability was excellent in FLAIR 0.92 (95% confidence interval [CI], 0.85-0.95), synthDIR 0.93 (95% CI, 0.87-0.96), and trueDIR 0.95 (95% CI, 0.85-0.98).Contrast-to-noise ratio in synthDIR exceeded that of FLAIR (22.0 +/- 6.4 vs 16.7 +/- 3.6, P = 0.009); no significant difference was seen in comparison to trueDIR (22.0 +/- 6.4 vs 22.4 +/- 7.9, P = 0.87). ConclusionsComputationally generated DIR images improve lesion depiction compared with the use of standard modalities. This method demonstrates how artificial intelligence can help improving imaging in specific pathologies.