Deep Generative Medical Image Harmonization for Improving Cross-Site Generalization in Deep Learning Predictors.
Deep Generative Medical Image Harmonization for Improving Cross-Site Generalization in Deep Learning Predictors.
复制标题
DOI：
10.1002/jmri.27908
复制
发表时间：
2022-03
期刊：
影响因子：
--
通讯作者：
iSTAGING and PHENOM consortia
中科院分区：
文献类型：
--
作者：
Bashyam VM;Doshi J;Erus G;Srinivasan D;Abdulkadir A;Singh A;Habes M;Fan Y;Masters CL;Maruff P;Zhuo C;Völzke H;Johnson SC;Fripp J;Koutsouleris N;Satterthwaite TD;Wolf DH;Gur RE;Gur RC;Morris JC;Albert MS;Grabe HJ;Resnick SM;Bryan NR;Wittfeld K;Bülow R;Wolk DA;Shou H;Nasrallah IM;Davatzikos C;iSTAGING and PHENOM consortia
在医学成像领域，基于深度学习的方法尚未得到广泛的临床应用，部分原因是不同成像设备和采集协议的泛化性能有限。估计的大脑年龄和生物年龄之间的偏差是大脑健康的既定生物标志物，并且此类模型可能受益于增加的跨站点概括性。开发和评估基于深度学习的图像协调方法，以提高深度学习年龄预测的跨站点通用性。来自6家临床试验机构的回顾性8876例受试者。使用所有受试者训练协调模型。年龄预测模型使用来自单个研究中心的2739名受试者进行训练，并使用来自其他研究中心的其余6137名受试者进行测试。在1.5 T和3T StarGAN v2条件下，使用磁化准备梯度回波（MPWS）或扰相梯度回波序列（SPGR）快速采集的脑成像执行从不同数据集到参考域的规范映射，以减少基于部位的变化，同时保留语义信息。深度学习年龄预测的泛化性能使用协调、直方图匹配和非协调数据进行评估。平均绝对误差和皮尔逊相关估计年龄和生物年龄量化的年龄预测模型的性能。我们的结果表明，样本外数据的年龄预测有了实质性的改善，总体平均绝对误差从15.81（±0.21）岁提高到11.86（±0.11）岁，直方图匹配到7.21（±0.22）岁，基于GAN的协调。在多站点的情况下，在5个样本站点中，平均绝对误差从9.78（±6.69）年改善到7.74（±3.03）年，直方图归一化为5.32（±4.07）年，基于GAN的协调。虽然还需要进一步的研究，但基于GAN的医学图像协调似乎是改善跨站点深度学习泛化的一个有前途的工具。
In the medical imaging domain, deep learning based methods have yet to see widespread clinical adoption, in part due to limited generalization performance across different imaging devices and acquisition protocols. The deviation between estimated brain age and biological age is an established biomarker of brain health and such models may benefit from increased cross-site generalizability. To develop and evaluate a deep learning based image harmonization method to improve cross-site generalizability of deep learning age prediction. Retrospective 8876 subjects from 6 sites. Harmonization models were trained using all subjects. Age prediction models were trained using 2739 subjects from a single site and tested using the remaining 6137 subjects from various other sites. Brain imaging with magnetization prepared rapid acquisition with gradient echo (MPRAGE) or spoiled gradient echo sequences (SPGR) at 1.5 and 3T StarGAN v2, was used to perform a canonical mapping from diverse datasets to a reference domain to reduce site-based variation while preserving semantic information. Generalization performance of deep learning age prediction was evaluated using harmonized, histogram matched, and unharmonized data. Mean absolute error and Pearson correlation between estimated age and biological age quantified the performance of the age prediction model. Our results indicated a substantial improvement in age prediction in out-of-sample data, with the overall mean absolute error improving from 15.81 (±0.21) years to 11.86 (±0.11) with histogram matching to 7.21 (±0.22) years with GAN-based harmonization. In the multisite case, across the 5 out of sample sites, mean absolute error improved from 9.78 (±6.69) years to 7.74 (±3.03) years with histogram normalization to 5.32 (±4.07) years with GAN-based harmonization. While further research is needed, GAN-based medical image harmonization appears to be a promising tool for improving cross-site deep learning generalization.
登录
查看更多内容
影响因子：
5.7
作者：
Jenkinson, M;Bannister, P;Smith, S
通讯作者：
Smith, S
影响因子：
10.6
作者：
Gao, Yuan;Liu, Yingchao;Yu, Jinhua
通讯作者：
Yu, Jinhua
影响因子：
4.8
作者：
Doshi, Jimit;Erus, Guray;Ou, Yangming;Gaonkar, Bilwaj;Davatzikos, Christos
通讯作者：
Davatzikos, Christos
影响因子：
5.7
作者：
Pomponio, Raymond;Erus, Guray;Davatzikos, Christos
通讯作者：
Davatzikos, Christos
影响因子：
14.5
作者：
Bashyam, Vishnu M.;Erus, Guray;Davatzikos, Christos
通讯作者：
Davatzikos, Christos