On the robustness of deep learning-based lung-nodule classification for CT images with respect to image noise.
On the robustness of deep learning-based lung-nodule classification for CT images with respect to image noise.
复制标题
关于基于深度学习的CT图像肺结节分类对图像噪声的鲁棒性。
DOI：
10.1088/1361-6560/abc812
复制
发表时间：
2020-12-22
影响因子：
3.5
通讯作者：
Jia X
中科院分区：
文献类型：
--
作者：
Shen C;Tsai MY;Chen L;Li S;Nguyen D;Wang J;Jiang SB;Jia X
稳健性是评价医学图像分析方法的一个重要方面。在这项研究中，我们研究了基于深度学习的CT图像肺结节分类模型对噪声扰动的鲁棒性。建立了一个深度神经网络（DNN），用于将肺结节的3D CT图像分类为恶性或良性组。所建立的DNN能够基于CT图像预测肺结节的恶性率，与放射科医生的预测相比，在10倍交叉验证中，测试数据集的曲线下面积（AUC）为0.91。然后，我们评估其对噪声扰动的鲁棒性。我们向输入CT图像中添加随机生成的噪声信号或通过优化方案生成的噪声信号，该优化方案使用基于给定mAs水平的噪声功率谱的现实噪声模型，并监控DNN的输出。结果表明，CT噪声能够影响所建立的DNN模型的预测结果。在100 mAs的随机噪声扰动下，DNN对11.2%的训练数据和17.4%的测试数据的预测至少成功改变了一次。对于基于优化的扰动，百分比分别增加到23.4%和34.3%。我们进一步评估了具有不同架构、参数、输出标签数量等的模型的鲁棒性，这些模型都存在不同程度的鲁棒性问题。为了提高模型的鲁棒性，我们根据经验提出了一种自适应训练方案。它通过在训练数据集中包含扰动来微调DNN模型，这些扰动成功地改变了DNN的扰动。重复执行自适应方案以逐渐提高DNN的鲁棒性。在两次迭代后，通过自适应训练方案，影响DNN预测的100 mAs扰动数量减少到10.8%，测试减少到21.1%。我们的研究表明，鲁棒性可能是CT图像的示例性基于深度学习的肺结节分类模型的一个潜在问题，这表明在开发类似模型时需要评估和确保模型的鲁棒性。所提出的自适应训练方案可能能够提高模型的鲁棒性。
Robustness is an important aspect when evaluating a method of medical image analysis. In this study, we investigated robustness of a deep learning-based lung nodule classification model for CT images with respect to noise perturbations. A deep neural network (DNN) was established to classify 3D CT images of lung nodules into malignant or benign groups. The established DNN was able to predict malignancy rate of lung nodules based on CT images, achieving the area under the curve (AUC) of 0.91 for the testing dataset in a 10-fold cross validation as compared to radiologists’ prediction. We then evaluated its robustness against noise perturbations. We added to the input CT images noise signals generated randomly or via an optimization scheme using a realistic noise model based on a noise power spectrum for a given mAs level, and monitored the DNN’s output. The results showed that the CT noise was able to affect the prediction results of the established DNN model. With random noise perturbations at 100 mAs, DNN’s predictions for 11.2% of training data and 17.4% of testing data were successfully altered by at least once. The percentage increased to 23.4% and 34.3%, respectively, for optimization-based perturbations. We further evaluated robustness of models with different architectures, parameters, number of output labels etc., and robustness concern was found in these models to different degrees. To improve model robustness, we empirically proposed an adaptive training scheme. It fine-tuned the DNN model by including perturbations in the training dataset that successfully altered the DNN’s perturbations. The adaptive scheme was repeatedly performed to gradually improve DNN’s robustness. The numbers of perturbations at 100 mAs affecting DNN’s predictions were reduced to 10.8% for training and 21.1% for testing by the adaptive training scheme after two iterations. Our study illustrated that robustness may potentially be a concern for an exemplary deep learning-based lung nodule classification model for CT images, indicating the needs for evaluating and ensuring model robustness when developing similar models. The proposed adaptive training scheme may be able to improve model robustness.
登录
查看更多内容
影响因子：
2.1
作者：
Dolly, Steven;Chen, Hsin-Chen;Li, Hua
通讯作者：
Li, Hua
影响因子：
4.4
作者：
Clark, Kenneth;Vendt, Bruce;Prior, Fred
通讯作者：
Prior, Fred
影响因子：
0.6
作者：
Bhalla, Ashu Seith;Das, Abanti;Goyal, Ankur
通讯作者：
Goyal, Ankur
影响因子：
4
作者：
Hua KL;Hsu CH;Hidayati SC;Cheng WH;Chen YJ
通讯作者：
Chen YJ
影响因子：
64.8
作者：
Mnih, Volodymyr;Kavukcuoglu, Koray;Hassabis, Demis
通讯作者：
Hassabis, Demis