Automated pectoral muscle identification on MLO-view mammograms: Comparison of deep neural network to conventional computer vision.
Automated pectoral muscle identification on MLO-view mammograms: Comparison of deep neural network to conventional computer vision.
复制标题
DOI：
10.1002/mp.13451
复制
发表时间：
2019-03
期刊：
影响因子：
3.8
通讯作者：
Xiangyuan Ma;Jun Wei;Chuan Zhou;M. Helvie;H. Chan;Lubomir M. Hadjiiski;Yao Lu
中科院分区：
文献类型：
--
作者：
Xiangyuan Ma;Jun Wei;Chuan Zhou;M. Helvie;H. Chan;Lubomir M. Hadjiiski;Yao Lu
本研究的目的是开发一种完全自动化的深度学习方法，用于识别内外斜位（MLO）乳腺X线照片上的胸肌，并与我们之前开发的使用传统图像特征分析的纹理场方向（TFO）方法进行比较，评估其性能。胸肌分割是自动图像分析的重要步骤，如乳腺密度或实质模式分类，病变检测和多视图相关。材料和方法在数据收集前获得机构审查委员会（IRB）的批准。来自我们先前研究的729个MLO视图乳房X线照片的数据集，包括637个数字化胶片乳房X线照片（DFM）和92个数字乳房X线照片（DM），用于训练和验证我们的深度卷积神经网络（DCNN）分割方法。此外，我们从131名患者中收集了一组独立的203个DM进行测试。用Lumiscan数字化仪将胶片乳腺X线照片以50 μm × 50 μm的像素大小进行数字化。所有DM均采用GE系统采集，像素大小为100 μm × 100 μm。经验丰富的MQSA放射科医生在每张乳房X线照片上手动绘制胸肌边界作为参考标准。我们训练DCNN来估计乳房X光片上胸肌区域的概率图。DCNN由用于捕获多分辨率图像上下文的收缩路径和用于预测胸肌区域的对称扩展路径组成。比较了三种DCNN结构用于胸肌的自动识别。在DCNN的训练中使用了十倍交叉验证。训练后，我们在交叉验证期间将十个训练好的模型应用于独立的DM测试集。通过对来自十个模型的概率图的集合求平均，获得每个测试DM的预测胸肌区域作为平均概率图。通过相对于参考标准的三个性能指标评价DCNN分割的胸肌：（a）胸肌区域的重叠面积百分比（POA），（B）Hausdorff距离（Hdist）和（c）边界之间的平均欧几里得距离（AvgDist）。将结果与TFO方法获得的结果进行比较，作为我们的基线。进行双尾配对t检验以检查DCNN与基线之间差异的显著性。结果在交叉验证集的10个测试分区中，DCNN实现的平均POA为96.5 ± 2.9%，平均Hdist为2.26 ± 1.31 mm，平均AvgDist为0.78 ± 0.58 mm，而基线方法的相应测量值分别为94.2 ± 4.8%，3.69 ± 2.48 mm和1.30 ± 1.22 mm，分别对于独立DM测试集，DCNN实现的平均POA为93.7% ± 6.9%，平均Hdist为3.80 ± 3.21 mm，平均AvgDist为1.49 ± 1.62 mm，而基线方法分别为86.9% ± 16.0%、7.18 ± 14.22 mm和3.98 ± 14.13 mm。结论与TFO方法相比，DCNN方法显著提高了乳腺X线片上胸肌识别的准确性（P < 0.05）。
OBJECTIVES The aim of this study was to develop a fully automated deep learning approach for identification of the pectoral muscle on mediolateral oblique (MLO) view mammograms and evaluate its performance in comparison to our previously developed texture-field orientation (TFO) method using conventional image feature analysis. Pectoral muscle segmentation is an important step for automated image analyses such as breast density or parenchymal pattern classification, lesion detection, and multiview correlation. MATERIALS AND METHODS Institutional Review Board (IRB) approval was obtained before data collection. A dataset of 729 MLO-view mammograms including 637 digitized film mammograms (DFM) and 92 digital mammograms (DM) from our previous study was used for the training and validation of our deep convolutional neural network (DCNN) segmentation method. In addition, we collected an independent set of 203 DMs from 131 patients for testing. The film mammograms were digitized at a pixel size of 50 μm × 50 μm with a Lumiscan digitizer. All DMs were acquired with GE systems at a pixel size of 100 μm × 100 μm. An experienced MQSA radiologist manually drew the pectoral muscle boundary on each mammogram as the reference standard. We trained the DCNN to estimate a probability map of the pectoral muscle region on mammograms. The DCNN consisted of a contracting path to capture multiresolution image context and a symmetric expanding path for prediction of the pectoral muscle region. Three DCNN structures were compared for automated identification of pectoral muscles. Tenfold cross-validation was used in training of the DCNNs. After training, we applied the ten trained models during cross-validation to the independent DM test set. The predicted pectoral muscle region of each test DM was obtained as the mean probability map by averaging the ensemble of probability maps from the ten models. The DCNN-segmented pectoral muscle was evaluated by three performance measures relative to the reference standard: (a) the percent overlap area (POA) of the pectoral muscle regions, (b) the Hausdorff distance (Hdist), and (c) the average Euclidean distance (AvgDist) between the boundaries. The results were compared to those obtained with the TFO method, used as our baseline. A two-tailed paired t test was performed to examine the significance in the differences between the DCNN and the baseline. RESULTS In the ten test partitions of the cross-validation set, the DCNN achieved a mean POA of 96.5 ± 2.9%, a mean Hdist of 2.26 ± 1.31 mm, and a mean AvgDist of 0.78 ± 0.58 mm, while the corresponding measures by the baseline method were 94.2 ± 4.8%, 3.69 ± 2.48 mm, and 1.30 ± 1.22 mm, respectively. For the independent DM test set, the DCNN achieved a mean POA of 93.7% ± 6.9%, a mean Hdist of 3.80 ± 3.21 mm, and a mean AvgDist of 1.49 ± 1.62 mm comparing to 86.9% ± 16.0%, 7.18 ± 14.22 mm, and 3.98 ± 14.13 mm, respectively, by the baseline method. CONCLUSION In comparison to the TFO method, DCNN significantly improved the accuracy of pectoral muscle identification on mammograms (P < 0.05).