Data augmentation with Mixup: Enhancing performance of a functional neuroimaging-based prognostic deep learning classifier in recent onset psychosis.
Data augmentation with Mixup: Enhancing performance of a functional neuroimaging-based prognostic deep learning classifier in recent onset psychosis.
复制标题
DOI：
10.1016/j.nicl.2022.103214
复制
发表时间：
2022
影响因子：
4.2
通讯作者：
Davidson, Ian
中科院分区：
文献类型：
--
作者：
Smucny, Jason;Shi, Ge;Lesh, Tyler A.;Carter, Cameron S.;Davidson, Ian
在fMRI数据上使用深度学习（DL）时，小样本量是一个挑战。我们研究了一种名为Mixup的数据增强程序对fMRI DL的影响。mixup使用fMRI数据改善了基于dl的精神病临床结局预测。虽然深度学习作为精神病学的预后工具有很大的希望，但该方法的局限性在于它需要大量的训练样本才能实现可复制的准确性。这对于fMRI数据集来说是有问题的，因为它们通常很小，因为获得它们所需的时间，成本和资源相当多。最近开发的一种名为Mixup的自监督学习方法可能有助于克服这一挑战。在Mixup中，学习器组合成对的训练实例以产生虚拟的第三个实例，这是两个实例及其标签的线性组合。该过程也非常适合于通常在fMRI数据集中发现的共配准图像。在这里，我们比较了基于任务fMRI的深度学习者在Mixup与没有Mixup的情况下预测近期发作精神病治疗反应的表现。从82例近期发作精神病患者的认知控制任务中提取全脑fMRI时间序列数据，并用于预测"改善者"（n = 47）与"非改善者"（n = 35）状态，其中改善者定义为治疗1年后简明精神病评定量表总评分降低20%。Mixup显著改善了性能（无Mixup的准确度：76.5%[95%CI：75.9 - 77.1%];有Mixup的准确度：80.1%[95%CI：79.4 - 80.8%]）。消融显示改善是由于改善者和非改善者的改善。这些结果表明，使用Mixup可以显着提高性能并减少基于fMRI的预后深度学习器的过拟合，还可以帮助克服许多神经成像数据集固有的小样本挑战。
Small sample sizes are a challenge when using deep learning (DL) on fMRI data. We examined effects of a data augmentation procedure called Mixup on fMRI DL. Mixup improved dl-based prediction of clinical outcome in psychosis using fMRI data. Although deep learning holds great promise as a prognostic tool in psychiatry, a limitation of the method is that it requires large training sample sizes to achieve replicable accuracy. This is problematic for fMRI datasets as they are typically small due to the considerable time, cost, and resources necessary to obtain them. A recently developed self-supervised learning method called Mixup may help overcome this challenge. In Mixup, the learner combines pairs of training instances to produce a virtual third instance that is a linear combination of the two instances and their labels. This procedure is also well-suited to the coregistered images typically found in fMRI datasets. Here we compared performance of a task fMRI-based deep learner with Mixup vs without Mixup on predicting response to treatment in recent onset psychosis. Whole brain fMRI time series data were extracted from a cognitive control task in 82 patients with recent onset psychosis and used to predict “Improver” (n = 47) vs “Non-Improver” (n = 35) status, with Improver defined as showing a 20 % reduction in total Brief Psychiatric Rating Scale score after 1 year of treatment. Mixup significantly improved performance (accuracy without Mixup: 76.5 % [95 % CI: 75.9–77.1 %]; accuracy with Mixup: 80.1 % [95 % CI: 79.4–80.8 %]). Ablation showed the improvement was due to improvement in both Improvers and Non-Improvers. These results suggest that using Mixup may significantly improve performance and reduce overfitting of fMRI-based prognostic deep learners and may also help overcome the small sample size challenge inherent to many neuroimaging datasets.
登录
查看更多内容
影响因子：
25.8
作者：
Horga, Guillermo;Wengler, Kenneth;Cassidy, Clifford M.
通讯作者：
Cassidy, Clifford M.
影响因子：
6.6
作者：
Blessing, Esther M.;Murty, Vishnu P.;Goff, Donald C.
通讯作者：
Goff, Donald C.
影响因子：
25.8
作者：
Lesh TA;Tanase C;Geib BR;Niendam TA;Yoon JH;Minzenberg MJ;Ragland JD;Solomon M;Carter CS
通讯作者：
Carter CS
影响因子：
2.9
作者：
Phillips RC;Salo T;Carter CS
通讯作者：
Carter CS
影响因子：
8.2
作者：
Elliott, Maxwell L.;Knodt, Annchen R.;Hariri, Ahmad R.
通讯作者：
Hariri, Ahmad R.