{"title":"Fully Automated Bone Age Assessment on Large-Scale Hand X-Ray Dataset.","authors":"Xiaoying Pan, Yizhe Zhao, Hao Chen, De Wei, Chen Zhao, Zhi Wei","doi":"10.1155/2020/8460493","DOIUrl":null,"url":null,"abstract":"<p><p>Bone age assessment (BAA) is an essential topic in the clinical practice of evaluating the biological maturity of children. Because the manual method is time-consuming and prone to observer variability, it is attractive to develop computer-aided and automated methods for BAA. In this paper, we present a fully automatic BAA method. To eliminate noise in a raw X-ray image, we start with using U-Net to precisely segment hand mask image from a raw X-ray image. Even though U-Net can perform the segmentation with high precision, it needs a bigger annotated dataset. To alleviate the annotation burden, we propose to use deep active learning (AL) to select unlabeled data samples with sufficient information intentionally. These samples are given to Oracle for annotation. After that, they are then used for subsequential training. In the beginning, only 300 data are manually annotated and then the improved U-Net within the AL framework can robustly segment all the 12611 images in RSNA dataset. The AL segmentation model achieved a Dice score at 0.95 in the annotated testing set. To optimize the learning process, we employ six off-the-shell deep Convolutional Neural Networks (CNNs) with pretrained weights on ImageNet. We use them to extract features of preprocessed hand images with a transfer learning technique. In the end, a variety of ensemble regression algorithms are applied to perform BAA. Besides, we choose a specific CNN to extract features and explain why we select that CNN. Experimental results show that the proposed approach achieved discrepancy between manual and predicted bone age of about 6.96 and 7.35 months for male and female cohorts, respectively, on the RSNA dataset. These accuracies are comparable to state-of-the-art performance.</p>","PeriodicalId":47063,"journal":{"name":"International Journal of Biomedical Imaging","volume":null,"pages":null},"PeriodicalIF":3.3000,"publicationDate":"2020-03-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1155/2020/8460493","citationCount":"25","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"International Journal of Biomedical Imaging","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1155/2020/8460493","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"2020/1/1 0:00:00","PubModel":"eCollection","JCR":"Q2","JCRName":"ENGINEERING, BIOMEDICAL","Score":null,"Total":0}
引用次数: 25
Abstract
Bone age assessment (BAA) is an essential topic in the clinical practice of evaluating the biological maturity of children. Because the manual method is time-consuming and prone to observer variability, it is attractive to develop computer-aided and automated methods for BAA. In this paper, we present a fully automatic BAA method. To eliminate noise in a raw X-ray image, we start with using U-Net to precisely segment hand mask image from a raw X-ray image. Even though U-Net can perform the segmentation with high precision, it needs a bigger annotated dataset. To alleviate the annotation burden, we propose to use deep active learning (AL) to select unlabeled data samples with sufficient information intentionally. These samples are given to Oracle for annotation. After that, they are then used for subsequential training. In the beginning, only 300 data are manually annotated and then the improved U-Net within the AL framework can robustly segment all the 12611 images in RSNA dataset. The AL segmentation model achieved a Dice score at 0.95 in the annotated testing set. To optimize the learning process, we employ six off-the-shell deep Convolutional Neural Networks (CNNs) with pretrained weights on ImageNet. We use them to extract features of preprocessed hand images with a transfer learning technique. In the end, a variety of ensemble regression algorithms are applied to perform BAA. Besides, we choose a specific CNN to extract features and explain why we select that CNN. Experimental results show that the proposed approach achieved discrepancy between manual and predicted bone age of about 6.96 and 7.35 months for male and female cohorts, respectively, on the RSNA dataset. These accuracies are comparable to state-of-the-art performance.
骨龄评估(BAA)是评估儿童生物学成熟度的重要课题。由于手工方法耗时长,且易受观测者变化的影响,因此开发BAA的计算机辅助和自动化方法是很有吸引力的。本文提出了一种全自动BAA方法。为了消除原始x射线图像中的噪声,我们首先使用U-Net从原始x射线图像中精确分割手掩膜图像。尽管U-Net可以实现高精度的分割,但它需要更大的标注数据集。为了减轻标注负担,我们建议使用深度主动学习(deep active learning, AL)来有意地选择具有足够信息的未标记数据样本。这些示例提供给Oracle进行注释。之后,它们被用于后续的训练。最初,只有300张数据需要手工标注,然后在人工智能框架下改进的U-Net可以鲁棒分割RSNA数据集中的所有12611张图像。人工智能分割模型在标注测试集中的Dice得分为0.95。为了优化学习过程,我们在ImageNet上使用了六个具有预训练权值的现成深度卷积神经网络(cnn)。我们使用迁移学习技术提取预处理手图像的特征。最后,应用了多种集成回归算法来执行BAA。此外,我们选择一个特定的CNN来提取特征,并解释为什么我们选择该CNN。实验结果表明,该方法在RSNA数据集上实现了男性和女性队列的人工骨龄和预测骨龄分别约为6.96个月和7.35个月的差异。这些精度可与最先进的性能相媲美。
期刊介绍:
The International Journal of Biomedical Imaging is managed by a board of editors comprising internationally renowned active researchers. The journal is freely accessible online and also offered for purchase in print format. It employs a web-based review system to ensure swift turnaround times while maintaining high standards. In addition to regular issues, special issues are organized by guest editors. The subject areas covered include (but are not limited to):
Digital radiography and tomosynthesis
X-ray computed tomography (CT)
Magnetic resonance imaging (MRI)
Single photon emission computed tomography (SPECT)
Positron emission tomography (PET)
Ultrasound imaging
Diffuse optical tomography, coherence, fluorescence, bioluminescence tomography, impedance tomography
Neutron imaging for biomedical applications
Magnetic and optical spectroscopy, and optical biopsy
Optical, electron, scanning tunneling/atomic force microscopy
Small animal imaging
Functional, cellular, and molecular imaging
Imaging assays for screening and molecular analysis
Microarray image analysis and bioinformatics
Emerging biomedical imaging techniques
Imaging modality fusion
Biomedical imaging instrumentation
Biomedical image processing, pattern recognition, and analysis
Biomedical image visualization, compression, transmission, and storage
Imaging and modeling related to systems biology and systems biomedicine
Applied mathematics, applied physics, and chemistry related to biomedical imaging
Grid-enabling technology for biomedical imaging and informatics