Unsupervised abnormality detection through mixed structure regularization (MSR) in deep sparse autoencoders
Unsupervised abnormality detection through mixed structure regularization (MSR) in deep sparse autoencoders
复制标题
DOI：
10.1002/mp.13464
复制
发表时间：
2019-05-01
期刊：
影响因子：
3.8
通讯作者：
Goshen, Liran
中科院分区：
文献类型：
--
作者：
Freiman, Moti;Manjeshwar, Ravindra;Goshen, Liran
目的本研究的目的是介绍和评估混合结构正则化（MSR）方法的深度稀疏自动编码器，旨在无监督的异常检测医学图像。基于使用深度稀疏自编码器识别离群值的无监督异常检测对于计算机辅助检测系统来说是一种非常有吸引力的方法，因为它只需要健康的数据进行训练，而不是专家注释的异常。然而，需要正则化来避免网络对训练数据的过度拟合。方法我们使用了90名受试者的冠状动脉计算机断层扫描血管造影（CCTA）数据集，并附有专家注释的中心线。我们使用自动算法分割冠状动脉腔和壁，并在需要时进行手动校正。我们将正常冠状动脉横截面定义为管腔与壁面积之比大于0.8的横截面。我们将数据集分为训练，验证和测试组，采用十重交叉验证方案。我们训练了一个深度稀疏过完备自动编码器模型，用于具有随机结构和噪声增强的正态建模。我们评估了我们的深度稀疏自动编码器与无去噪MSR（SAE-MSR）和去噪（SDAE-MSR）的性能，并与深度稀疏自动编码器（SAE）和深度稀疏去噪自动编码器（SDAE）模型进行比较，以从测试组的CCTA数据中检测冠状动脉疾病。结果与SAE和SDAE相比，SDAE-MSR的累积曲线下面积（AUC）提高了20%;平均精密度（AP）提高了30%（AUC：0.78至0.94，AP：0.66至0.86）区分轻度狭窄的冠状动脉横截面（狭窄分级< 0.3）和重度狭窄（狭窄分级> 0.7）的冠状动脉横截面。改善具有统计学显著性（Mann-Whitney U检验，P < 0.001）。同样，SDAE-MSR达到了最佳汇总AUC（AP），SAE和SDAE改善18%（18%）（AUC：0.71 - 0.84，AP：0.68 - 0.80）。改善具有统计学显著性（Mann-Whitney U检验，P < 0.05）。结论与普通深度自编码器相比，除了显式稀疏正则化项和具有高斯噪声的输入数据的随机损坏之外，具有MSR的深度稀疏自编码器具有使用深度学习改进无监督异常检测的潜力。
Purpose The purpose of this study is to introduce and evaluate the mixed structure regularization (MSR) approach for a deep sparse autoencoder aimed at unsupervised abnormality detection in medical images. Unsupervised abnormality detection based on identifying outliers using deep sparse autoencoders is a very appealing approach for computer-aided detection systems as it requires only healthy data for training rather than expert annotated abnormality. However, regularization is required to avoid overfitting of the network to the training data. Methods We used coronary computed tomography angiography (CCTA) datasets of 90 subjects with expert annotated centerlines. We segmented coronary lumen and wall using an automatic algorithm with manual corrections where required. We defined normal coronary cross section as cross sections with a ratio between lumen and wall areas larger than 0.8. We divided the datasets into training, validation, and testing groups in a tenfold cross-validation scheme. We trained a deep sparse overcomplete autoencoder model for normality modeling with random structure and noise augmentation. We assessed the performance of our deep sparse autoencoder with MSR without denoising (SAE-MSR) and with denoising (SDAE-MSR) in comparison to deep sparse autoencoder (SAE), and deep sparse denoising autoencoder (SDAE) models in the task of detecting coronary artery disease from CCTA data on the test group. Results The SDAE-MSR achieved the best aggregated area under the curve (AUC) with a 20% improvement and the best aggregated Average Precision (AP) with a 30% improvement upon the SAE and SDAE (AUC: 0.78 to 0.94, AP: 0.66 to 0.86) in distinguishing between coronary cross sections with mild stenosis (stenosis grade < 0.3) and coronary cross sections with severe stenosis (stenosis grade > 0.7). The improvements were statistically significant (Mann-Whitney U-test, P < 0.001). Similarly, The SDAE-MSR achieved the best aggregated AUC (AP) with an 18% (18%) improvement upon the SAE and SDAE (AUC: 0.71 to 0.84, AP: 0.68 to 0.80). The improvements were statistically significant (Mann-Whitney U-test, P < 0.05). Conclusion Deep sparse autoencoders with MSR in addition to explicit sparsity regularization term and stochastic corruption of the input data with Gaussian noise have the potential to improve unsupervised abnormality detection using deep-learning compared to common deep autoencoders.