AbstractThis study constructed deep learning models using plain skull radiograph images to predict the accurate postnatal age of infants under 12 months. Utilizing the results of the trained deep learning models, it aimed to evaluate the feasibility of employing major changes visible in skull X-ray images for assessing postnatal cranial development through gradient-weighted class activation mapping.

摘要本研究利用颅骨X线平片图像构建深度学习模型，以预测12个月以下婴儿的准确出生年龄。利用经过训练的深度学习模型的结果，旨在评估通过梯度加权类激活映射来评估颅骨X射线图像中可见的主要变化评估产后颅骨发育的可行性。

We developed DenseNet-121 and EfficientNet-v2-M convolutional neural network models to analyze 4933 skull X-ray images collected from 1343 infants. Notably, allowing for a ± 1 month error margin, DenseNet-121 reached a maximum corrected accuracy of 79.4% for anteroposterior (AP) views (average: 78.0 ± 1.5%) and 84.2% for lateral views (average: 81.1 ± 2.9%).

我们开发了DenseNet-121和EfficientNet-v2-M卷积神经网络模型，以分析从1343名婴儿收集的4933张颅骨X射线图像。值得注意的是，考虑到±1个月的误差，DenseNet-121的前后（AP）视图的最大校正精度为79.4%（平均值：78.0±1.5%），侧视图的最大校正精度为84.2%（平均值：81.1±2.9%）。

EfficientNet-v2-M reached a maximum corrected accuracy 79.1% for AP views (average: 77.0 ± 2.3%) and 87.3% for lateral views (average: 85.1 ± 2.5%). Saliency maps identified critical discriminative areas in skull radiographs, including the coronal, sagittal, and metopic sutures in AP skull X-ray images, and the lambdoid suture and cortical bone density in lateral images, marking them as indicators for evaluating cranial development.

EfficientNet-v2-M对AP视图的最大校正精度达到79.1%（平均值：77.0±2.3%），对侧视图的最大校正精度达到87.3%（平均值：85.1±2.5%）。显着性图确定了颅骨X线照片中的关键区分区域，包括AP颅骨X射线图像中的冠状，矢状和异位缝合线，以及侧面图像中的lambdoid缝合线和皮质骨密度，将其标记为评估颅骨发育的指标。

These findings highlight the precision of deep learning in estimating infant age through non-invasive methods, offering the progress for clinical diagnostics and developmental assessment tools..

这些发现突出了深度学习在通过非侵入性方法估计婴儿年龄方面的准确性，为临床诊断和发育评估工具提供了进展。。

IntroductionAs neonates grow, the cranium undergoes significant changes, such as the narrowing of cranial sutures and the closure of fontanels, which are critical markers of normal cranial development1,2. Given that craniosynostosis—a condition characterized by the premature fusion of cranial sutures, affecting approximately 1 out of every 2500 live births—can lead to severe neurodevelopmental impairments if left untreated, the need for accurate diagnostic tools for its early detection is crucial3,4,5,6.

引言随着新生儿的成长，颅骨会发生重大变化，例如颅骨缝线变窄和font门闭合，这是正常颅骨发育的关键标志1,2。鉴于颅缝早闭症是一种以颅骨缝线过早融合为特征的疾病，如果不及时治疗，每2500例活产中约有1例会导致严重的神经发育障碍，因此对其早期发现的准确诊断工具的需求至关重要3,4,5,6。

Furthermore, in developing diagnostic tools capable of distinguishing such pathological conditions, it is imperative first to have tools that provide information serving as criteria for indicating normal cranial development. However, the lack of comprehensive research on these developmental milestones has led to a scarcity of reference data, making it challenging to assess whether an infant's cranial growth aligns with normal chronological changes.Plain skull X-ray imaging, with its low radiation exposure and non-invasiveness, offers a valuable resource for evaluating cranial development, yet its potential has been underutilized in the context of age estimation.With the advent of convolutional neural network (CNN) models, deep learning has revolutionized the ability to classify and interpret medical images with precision surpassing traditional methods7,8.

此外，在开发能够区分此类病理状况的诊断工具时，必须首先使用提供信息的工具作为指示正常颅骨发育的标准。然而，由于缺乏对这些发展里程碑的全面研究，导致缺乏参考数据，因此难以评估婴儿的颅骨生长是否符合正常的时间变化。普通颅骨X射线成像具有低辐射暴露和无创性，为评估颅骨发育提供了宝贵的资源，但在年龄估计方面其潜力尚未得到充分利用。随着卷积神经网络（CNN）模型的出现，深度学习已经彻底改变了对医学图像进行分类和解释的能力，其精度超过了传统方法7,8。

This technological advancement opens new avenues for the application of plain skull X-rays in the precise estimation of postnatal age, facilitating the early detection of cranial anomalies such as craniosynostosis.Therefore, this study aims to assess the utility of a deep learning model developed to predict the postnatal age of infants using plain skull X-ray images, evaluating its applicability in clinical medicine an.

这项技术进步为应用颅骨X射线精确估计出生后年龄开辟了新途径，有助于早期发现颅骨异常，如颅缝早闭。因此，本研究旨在评估深度学习模型的实用性，该模型用于使用颅骨X射线图像预测婴儿的出生后年龄，评估其在临床医学和临床医学中的适用性。

1.

1.

on the AP or Town’s skull X-ray image, they encompassed the upper margin of the supraorbital rim and the lower margin of the mandible (Fig. 1A)

在AP或Town的头骨X射线图像上，它们包括眶上缘的上缘和下颌骨的下缘（图1A）

2.

2.

on the lateral skull X-ray image, they included the supraorbital rim, the foremost part of the mandible, and the posterior margin of the cervical spinous process (Fig. 1B).

在侧颅骨X射线图像上，它们包括眶上缘，下颌骨的最前端和颈椎棘突的后缘（图1B）。

Figure 1The defined region of exclusion (ROE) in the skull X-ray for image tailoring. (A) Anteroposterior (AP) or Town’s view skull X-ray showing the defined ROE. The borders of the ROE extend from the upper margin of the supraorbital rim to the lower margin of the mandible. (B) Lateral skull X-ray with the ROE including the supraorbital rim, the foremost part of the mandible, and the posterior margin of the cervical spinous process.

图1用于图像裁剪的颅骨X射线中定义的排除区域（ROE）。（A）前后（AP）或城镇视野颅骨X射线显示定义的ROE。ROE的边界从眶上缘的上缘延伸到下颌骨的下缘。（B） 侧颅骨X射线，ROE包括眶上缘，下颌骨的最前端和颈椎棘突的后缘。

(C) Post-processed AP or Town’s view skull X-ray. The region below the upper margin of the ROE has been removed. (D) Post-processed lateral skull X-ray. A square box, defined by the upper and right margins of the ROE, has been removed.Full size imageThe defined area on each of 293 skull X-ray images was labeled as ROE by a neurosurgery expert (H.S.L).

（C） 。ROE上边缘以下的区域已被删除。（D） 后处理侧颅骨X射线。由ROE的上边缘和右边缘定义的方形框已被删除。全尺寸图像神经外科专家（H.S.L）将293张颅骨X射线图像中的每个定义区域标记为ROE。

The entire ROE dataset was divided into training, validation, and test datasets through random sampling with a ratio of 8:1:1. The MobileNetV3 model was trained for object detection of the labeled ROE. Regarding training parameters, the Adam optimizer was used with an initial learning rate of 1e − 3 and batch size of 16.

通过以8:1:1的比例随机抽样，将整个ROE数据集分为训练，验证和测试数据集。训练MobileNetV3模型用于标记ROE的对象检测。关于训练参数，使用Adam优化器，初始学习率为1e-3，批量大小为16。

Subsequently, post-processing was performed on all images to eliminate the detected ROEs based on the following criteria:.

随后，根据以下标准对所有图像进行后处理，以消除检测到的ROE：。

1)

1)

on AP or Town’s skull X-ray, the region below the upper margin of the ROE was removed (Fig. 1C)

在AP或Town的头骨X光检查中，鱼子上缘下方的区域被移除（图1C）

2)

2)

on the lateral skull X-ray, the square box defined by the upper margin of the ROE and right margin of the ROE was removed (Fig. 1D).

在侧颅骨X射线上，删除了由ROE上边缘和ROE右边缘定义的方形框（图1D）。

All tailored images were then reviewed by a neurosurgeon (H.S.L.) and adjusted for any misprocessing. After tailoring the region of interest (ROI) in the images, all images were center-symmetrically zero-padded into square shapes to match the longer side of the width and height. Bi-linear interpolation was applied to the transformed square images of different sizes to resize them to a uniform size of 1024 × 1024 pixels.

然后由神经外科医生（H.S.L.）检查所有定制的图像，并针对任何错误处理进行调整。在裁剪图像中的感兴趣区域（ROI）后，将所有图像中心对称地零填充成方形，以匹配宽度和高度的较长一侧。将双线性插值应用于不同大小的变换后的方形图像，以将其调整为1024×1024像素的均匀大小。

Min–max normalization was applied to normalize all images.Training CNN modelsTo construct deep-learning models, two different CNN architectures, DenseNet-121 and EfficientNet-V2-M, were adopted. DenseNet-121 has an improved algorithm for feature representation and learning efficiency and has been effective at medical image classification10, and EfficientNet-V2-M, which has been relatively recently introduced and has shown higher performance in general image classification tasks with low computational cost11,12.

应用最小-最大归一化来归一化所有图像。训练CNN模型为了构建深度学习模型，采用了两种不同的CNN架构，DenseNet-121和EfficientNet-V2-M。DenseNet-121具有改进的特征表示和学习效率算法，并且在医学图像分类10和EfficientNet-V2-M方面是有效的，EfficientNet-V2-M是最近引入的，在一般图像分类任务中表现出更高的性能，计算成本低11,12。

In brief, DenseNet consists of dense blocks linking the feature map of previous layers together, while the EfficientNet-V2-M model searches for the most effective CNN architecture using neural architecture search, similar to EfficientNet. DenseNet-121 and EfficientNet-V2-M had previously been trained with the ImageNet dataset and were fine-tuned by unboxing the weights11,12,13.

简而言之，DenseNet由将前几层的特征图连接在一起的密集块组成，而EfficientNet-V2-M模型使用类似于EfficientNet的神经架构搜索来搜索最有效的CNN架构。DenseNet-121和EfficientNet-V2-M之前已经用ImageNet数据集进行了训练，并通过取消权重11,12,13进行了微调。

All layers were unfreezed, allowing fine-tuning of every layer in the network.The batch size was set at 8 for DenseNet-121 and 4 for EfficientNet-V2-M, the maximum capacity that the GPU memory of our hardware could handle with each architecture. Categorical cross-entropy was used as the loss function, and the Adam optimizer was applied14.

。使用分类交叉熵作为损失函数，并应用Adam优化器14。

The initial learning rate was set to 0.0001 and was reduced by a factor of 0.1 every 10 epochs. Early stopping was employed after the 20t.

初始学习率设置为0.0001，每10个时期降低0.1倍。20t后采用提前停车。

Data availability

数据可用性

The authors confirm that the meta-data supporting the results of the deep learning is provided within the supplementary information files.

作者确认，补充信息文件中提供了支持深度学习结果的元数据。

ReferencesSwischuk, L. E. The normal pediatric skull. Variations and artefacts. Radiol. Clin. N. Am. 10(2), 277–290 (1972) (published Online First: 1972/08/01).Article

参考文献Wischuk，L.E。正常的儿科头骨。变异和人工制品。放射性。临床。N、 上午10（2），277–290（1972）（首次在线发布：1972年8月1日）。文章

CAS

中科院

PubMed

PubMed

Google Scholar

谷歌学者

Swischuk, L. E. The growing skull. Semin. Roentgenol. 9(2), 115–124. https://doi.org/10.1016/0037-198x(74)90027-3 (1974) (published Online First: 1974/04/01).Article

Swischuk，L.E。生长中的头骨。塞米。伦琴。9（2），115-124。https://doi.org/10.1016/0037-198x（74）90027-3（1974）（首次在线发布：1974/04/01）。文章

CAS

中科院

PubMed

PubMed

Google Scholar

谷歌学者

Speltz, M. L. et al. Neurodevelopment of infants with single-suture craniosynostosis: Presurgery comparisons with case-matched controls. Plast. Reconstr. Surg. 119(6), 1874–1881. https://doi.org/10.1097/01.prs.0000259184.88265.3f (2007) (published Online First: 2007/04/19).Article

Speltz，M.L.等人。单缝颅缝早闭婴儿的神经发育：与病例匹配对照的术前比较。塑料。重建。Surg.119（6），1874-1881年。https://doi.org/10.1097/01.prs.0000259184.88265.3f（2007）（首次在线发布：2007年4月19日）。文章

CAS

中科院

PubMed

PubMed

PubMed Central

公共医学中心

Google Scholar

谷歌学者

Lekovic, G. P., Bristol, R. E. & Rekate, H. L. Cognitive impact of craniosynostosis. Semin. Pediatr. Neurol. 11(4), 305–310. https://doi.org/10.1016/j.spen.2004.12.001 (2004) (published Online First: 2005/04/15).Article

Lekovic，G.P.，Bristol，R.E.＆Rekate，H.L。颅缝早闭症的认知影响。塞米。儿科。神经病学。11（4），305-310。https://doi.org/10.1016/j.spen.2004.12.001（2004年）（首次在线发布：2005年4月15日）。文章

PubMed

PubMed

Google Scholar

谷歌学者

Shim, K. W., Park, E. K., Kim, J. S., Kim, Y. O. & Kim, D. S. Neurodevelopmental problems in non-syndromic craniosynostosis. J. Korean Neurosurg. Soc. 59(3), 242–246. https://doi.org/10.3340/jkns.2016.59.3.242 (2016) (published Online First: 2016/05/27).Article

Shim，K.W.，Park，E.K.，Kim，J.S.，Kim，Y.O。＆Kim，D.S。非综合征性颅缝早闭症的神经发育问题。J、 韩国神经外科。Soc。59（3），242-246。https://doi.org/10.3340/jkns.2016.59.3.242（2016）（首次在线发布：2016年5月27日）。文章

CAS

中科院

PubMed

PubMed

PubMed Central

公共医学中心

Google Scholar

谷歌学者

Proctor, M. R. & Meara, J. G. A review of the management of single-suture craniosynostosis, past, present, and future. J. Neurosurg. Pediatr. 24(6), 622–631. https://doi.org/10.3171/2019.7.Peds18585 (2019) (published Online First: 2019/12/02).Article

Proctor，M.R。＆Meara，J.G。回顾单缝线颅缝早闭的管理，过去，现在和未来。J、 神经外科。儿科。24（6），622-631。https://doi.org/10.3171/2019.7.Peds18585。文章

PubMed

PubMed

Google Scholar

谷歌学者

Byeon, S. J., Park, J., Cho, Y. A. & Cho, B. J. Automated histological classification for digital pathology images of colonoscopy specimen via deep learning. Sci. Rep. 12(1), 12804. https://doi.org/10.1038/s41598-022-16885-x (2022) (published Online First: 20220727).Article

Byeon，S.J.，Park，J.，Cho，Y.A。＆Cho，B.J。通过深度学习对结肠镜检查标本的数字病理图像进行自动组织学分类。科学。代表12（1），12804。https://doi.org/10.1038/s41598-022-16885-x（2022）（首次在线发布：20220727）。文章

ADS

广告

CAS

中科院

PubMed

PubMed

PubMed Central

公共医学中心

Google Scholar

谷歌学者

Poplin, R. et al. Prediction of cardiovascular risk factors from retinal fundus photographs via deep learning. Nat. Biomed. Eng. 2(3), 158–164. https://doi.org/10.1038/s41551-018-0195-0 (2018) (published Online First: 20180219).Article

Poplin，R.等人。通过深度学习从视网膜眼底照片预测心血管危险因素。。工程2（3），158-164。https://doi.org/10.1038/s41551-018-0195-0（2018）（首次在线发布：20180219）。文章

PubMed

PubMed

Google Scholar

谷歌学者

Kim, D. K., Cho, B. J., Lee, M. J. & Kim, J. H. Prediction of age and sex from paranasal sinus images using a deep learning network. Medicine (Baltimore) 100(7), e24756. https://doi.org/10.1097/MD.0000000000024756 (2021) (published Online First: 2021/02/21).Article

Kim，D.K.，Cho，B.J.，Lee，M.J。和Kim，J.H。使用深度学习网络从鼻窦图像预测年龄和性别。医学（巴尔的摩）100（7），e24756。https://doi.org/10.1097/MD.0000000000024756（2021年）（首次在线发布：2021年2月21日）。文章

PubMed

PubMed

Google Scholar

谷歌学者

Hou, Y., Wu, Z., Cai, X. & Zhu, T. The application of improved densenet algorithm in accurate image recognition. Sci. Rep. 14(1), 8645. https://doi.org/10.1038/s41598-024-58421-z (2024) (published Online First: 2024/04/15).Article

Hou，Y.，Wu，Z.，Cai，X。＆Zhu，T。改进的densenet算法在精确图像识别中的应用。科学。代表14（1），8645。https://doi.org/10.1038/s41598-024-58421-z（2024年）（首次在线发布：2024年4月15日）。文章

ADS

广告

CAS

中科院

PubMed

PubMed

PubMed Central

公共医学中心

Google Scholar

谷歌学者

Gao Huang, Z.L. Laurens van der Maaten and Kilian Weinberger. Densely Connected Convolutional Networks (DenseNets). CVPR 2017.Mingxing Tan QVL. EfficientNetV2: Smaller Models and Faster Training. International Conference on Machine Learning, 2021.Mingxing Tan, Q.V.L. EfficientNet: Rethinking model scaling for convolutional neural networks international conference on machine learning.

。密集连接的卷积网络（DenseNets）。CVPR 2017。Tan Mingxing QVL。EfficientNetV2：更小的模型和更快的训练。国际机器学习会议，2021。Tan Mingxing Tan，Q.V.L。EfficientNet：卷积神经网络的重新思考模型缩放国际机器学习会议。

2019. 11 [published Online First: 24 May 2019].Diederik, P., Kingma, J.B. Adam: A method for stochastic optimization. International Conference on Learning Representations, 2014.Aditya Chattopadhyay, A.S., Prantik Howlader, V. Balasubramanian. Grad-CAM++: Improved Visual Explanations for Deep Convolutional Networks.

2019年11月[首次在线发布：2019年5月24日]。Diederik，P.，Kingma，J.B.Adam：一种随机优化方法。国际学习表征会议，2014年。Aditya Chattopadhyay，A.S.，Prantik Howlader，V.Balasubramanian。Grad CAM++：改进了深度卷积网络的视觉解释。

IEEE Workshop/Winter Conference on Applications of Computer Vision, 2017.Halabi, S. S. et al. The RSNA pediatric bone age machine learning challenge. Radiology 290(2), 498–503. https://doi.org/10.1148/radiol.2018180736 (2019) (published Online First: 2018/11/27).Article .

IEEE计算机视觉应用研讨会/冬季会议，2017年。Halabi，S.S.等人。RSNA儿科骨龄机器学习挑战。放射学290（2），498-503。https://doi.org/10.1148/radiol.2018180736（2019）（首次在线发布：2018年11月27日）。。

PubMed

PubMed

Google Scholar

谷歌学者

LeCun, Y., Bengio, Y. & Hinton, G. Deep learning. Nature 521(7553), 436–444. https://doi.org/10.1038/nature14539 (2015) (published Online First: 2015/05/29).Article

LeCun，Y.，Bengio，Y。＆Hinton，G。深度学习。自然521（7553），436-444。https://doi.org/10.1038/nature14539（2015）（首次在线发布：2015年5月29日）。文章

ADS

广告

CAS

中科院

PubMed

PubMed

Google Scholar

谷歌学者

Quitadamo, L. R. et al. Support vector machines to detect physiological patterns for EEG and EMG-based human–computer interaction: A review. J. Neural Eng. 14(1), 011001. https://doi.org/10.1088/1741-2552/14/1/011001 (2017) (published Online First: 2017/01/10).Article

Quitadamo，L.R.等人。基于EEG和EMG的人机交互生理模式检测的支持向量机：综述。J、 神经工程14（1），011001。https://doi.org/10.1088/1741-2552/14/1/011001（2017）（首次在线发布：2017/01/10）。文章

ADS

广告

CAS

中科院

PubMed

PubMed

Google Scholar

谷歌学者

Ichikawa, K. & Morishita, S. A simple but powerful heuristic method for accelerating k-means clustering of large-scale data in life science. IEEE/ACM Trans. Comput. Biol. Bioinform. 11(4), 681–692. https://doi.org/10.1109/TCBB.2014.2306200 (2014) (published Online First: 2014/07/01).Article .

Ichikawa，K。＆Morishita，S。一种简单但功能强大的启发式方法，用于加速生命科学中大规模数据的K均值聚类。IEEE/ACM Trans。计算机。生物。生物信息。。https://doi.org/10.1109/TCBB.2014.2306200（2014）（首次在线发布：2014年7月1日）。。

PubMed

PubMed

Google Scholar

谷歌学者

Murtaza, S. S., Kolpak, P., Bener, A. & Jha, P. Automated verbal autopsy classification: Using one-against-all ensemble method and Naive Bayes classifier. Gates Open Res. 2, 63. https://doi.org/10.12688/gatesopenres.12891.2 (2018) (published Online First: 2019/05/28).Article

Murtaza，S.S.，Kolpak，P.，Bener，A。＆Jha，P。自动口头尸检分类：使用一对所有集合方法和朴素贝叶斯分类器。大门打开第2、63号决议。https://doi.org/10.12688/gatesopenres.12891.2（2018）（首次在线发布：2019/05/28）。文章

PubMed

PubMed

Google Scholar

谷歌学者

Soffer, S. et al. Convolutional neural networks for radiologic images: A radiologist’s guide. Radiology 290(3), 590–606. https://doi.org/10.1148/radiol.2018180547 (2019) (published Online First: 2019/01/30).Article

Soffer，S.等人，《放射学图像的卷积神经网络：放射科医师指南》。放射学290（3），590-606。https://doi.org/10.1148/radiol.2018180547（2019）（首次在线发布：2019/01/30）。文章

PubMed

PubMed

Google Scholar

谷歌学者

Wang, Y., Zhu, F., Boushey, C. J. & Delp, E. J. Weakly supervised food image segmentation using class activation maps. Proc. Int. Conf. Image Proc. 2017, 1277–1281. https://doi.org/10.1109/ICIP.2017.8296487 (2017) (published Online First: 2017/09/01).Article

Wang，Y.，Zhu，F.，Boushey，C.J。＆Delp，E.J。使用类激活图的弱监督食品图像分割。程序。内部形态图像处理。20171277-1281年。https://doi.org/10.1109/ICIP.2017.8296487（2017）（首次在线发布：2017/09/01）。文章

PubMed

PubMed

Google Scholar

谷歌学者

Momose, K. J. Developmental approach in the analysis of roentgenograms of the pediatric skull. Radiol. Clin. N. Am. 9(1), 99–116 (1971) (published Online First: 1971/04/01).Article

Momose，K.J。儿童颅骨X线照片分析中的发育方法。放射性。临床。N、 上午9（1），99-116（1971）（首次在线发布：1971年4月1日）。文章

CAS

中科院

PubMed

PubMed

Google Scholar

谷歌学者

Slater, B. J. et al. Cranial sutures: A brief review. Plast. Reconstr. Surg. 121(4), 170e-e178. https://doi.org/10.1097/01.prs.0000304441.99483.97 (2008) (published Online First: 2008/03/20).Article

Slater，B.J.等人，《颅缝：简要回顾》。塑料。重建。附件121（4），170e-e178。https://doi.org/10.1097/01.prs.0000304441.99483.97（2008）（首次在线发布：2008年3月20日）。文章

CAS

中科院

PubMed

PubMed

Google Scholar

谷歌学者

Download referencesAcknowledgementsThe authors are thankful to all staff members of the Neurosurgical Department and the patients of the study group whose contributions made this work possible.FundingThis research was supported by the Bio & Medical Technology Development Program of the National Research Foundation (NRF) and funded by the Korean government (MSIT) (No.

下载参考文献致谢作者感谢神经外科的所有工作人员和研究组的患者，他们的贡献使这项工作成为可能。资助这项研究得到了国家研究基金会（NRF）生物与医学技术发展计划的支持，并由韩国政府（MSIT）资助（No。

NRF-2022R1C1C1010643).Author informationAuthor notesThese authors contributed equally: Heui Seung Lee and Jaewoong Kang.Authors and AffiliationsDepartment of Neurosurgery, College of Medicine, Hallym University Sacred Heart Hospital, Hallym University, 22, Gwanpyeong-Ro 170Beon-Gil, Dongan-Gu, Anyang-Si, Gyeonggi-Do, 14068, Republic of KoreaHeui Seung Lee & Ji Hee KimInterdisciplinary Program for Bioinformatics, Graduate School, Seoul National University, Seoul, Republic of KoreaHeui Seung LeeMedical Artificial Intelligence Center, Hallym University Medical Center, Anyang, Republic of KoreaJaewoong Kang, So Eui Kim & Bum-Joo ChoDepartment of Ophthalmology, College of Medicine, Hallym University Sacred Heart Hospital, Hallym University, 22, Gwanpyeong-Ro 170Beon-Gil, Dongan-Gu, Anyang-Si, Gyeonggi-Do, 14068, Republic of KoreaBum-Joo ChoAuthorsHeui Seung LeeView author publicationsYou can also search for this author in.

NRF-2022R1C1C1010643）。作者信息作者注意到这些作者做出了同样的贡献：Heui Seung Lee和Jaewong Kang。作者和附属机构哈利姆大学圣心医院医学院神经外科，哈利姆大学，22岁，关平路170Beon Gil，东安谷，安阳市，京畿道，14068，大韩民国，首尔国立大学研究生院，生物信息学跨学科计划，首尔，大韩民国，大韩民国医学人工智能中心，大韩民国，安阳哈利姆大学，22岁，关平路170Beon Gil，东安谷，安阳寺，京畿道，14068，韩国共和国Joo ChoAuthorsHeui Seung LeeView作者出版物您也可以在中搜索这位作者。

PubMed Google ScholarJaewoong KangView author publicationsYou can also search for this author in

PubMed Google ScholarJaewoong KangView作者出版物您也可以在

PubMed Google ScholarSo Eui KimView author publicationsYou can also search for this author in

PubMed Google ScholarSo Eui KimView作者出版物您也可以在

PubMed Google ScholarJi Hee KimView author publicationsYou can also search for this author in

PubMed Google ScholarJi Hee KimView作者出版物您也可以在

PubMed Google ScholarBum-Joo ChoView author publicationsYou can also search for this author in

PubMed Google ScholarBum Joo ChoView作者出版物您也可以在

PubMed Google ScholarContributionsConceptualization and Study Design: BJ Choi, HS Lee. Methodology: BJ CHO. Formal analyses: HS Lee, BJ Cho, SE Kim, JW Kang. Data curation: JH Kim, HS Lee. Writing Original Draft: HS Lee, BJ Cho, JW Kang. Editing: BJ Cho, HS Lee, JW Kang. All authors reviewed and approved manuscript.Corresponding authorsCorrespondence to.

PubMed谷歌学术贡献概念和研究设计：BJ Choi，HS Lee。方法：BJ CHO。正式分析：HS Lee，BJ Cho，SE Kim，JW Kang。数据管理：JH Kim，HS Lee。撰写原稿：HS Lee，BJ Cho，JW Kang。编辑：BJ Cho，HS Lee，JW Kang。所有作者都审查并批准了手稿。通讯作者通讯。

Heui Seung Lee or Bum-Joo Cho.Ethics declarations

李惠承或赵文珠。道德宣言

Competing interests

相互竞争的利益

The authors declare no competing interests.

作者声明没有利益冲突。

Additional informationPublisher's noteSpringer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Supplementary InformationSupplementary Information.Rights and permissions

Additional informationPublisher的noteSpringer Nature在已发布地图和机构隶属关系中的管辖权主张方面保持中立。补充信息补充信息。权限和权限

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.

开放获取本文是根据知识共享署名4.0国际许可证授权的，该许可证允许以任何媒体或格式使用，共享，改编，分发和复制，只要您对原始作者和来源给予适当的信任，提供知识共享许可证的链接，并指出是否进行了更改。

The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

。如果材料未包含在文章的知识共享许可中，并且您的预期用途不受法律法规的许可或超出许可用途，则您需要直接获得版权所有者的许可。

To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/..

要查看此许可证的副本，请访问http://creativecommons.org/licenses/by/4.0/..

Reprints and permissionsAbout this articleCite this articleLee, H.S., Kang, J., Kim, S.E. et al. Estimating infant age from skull X-ray images using deep learning.

转载和许可本文引用本文Lee，H.S.，Kang，J.，Kim，S.E.等人使用深度学习从颅骨X射线图像估计婴儿年龄。

Sci Rep 14, 16600 (2024). https://doi.org/10.1038/s41598-024-64489-4Download citationReceived: 08 February 2024Accepted: 10 June 2024Published: 18 July 2024DOI: https://doi.org/10.1038/s41598-024-64489-4Share this articleAnyone you share the following link with will be able to read this content:Get shareable linkSorry, a shareable link is not currently available for this article.Copy to clipboard.

Sci Rep 1416600（2024）。https://doi.org/10.1038/s41598-024-64489-4Download引文接收日期：2024年2月8日接受日期：2024年6月10日发布日期：2024年7月18日OI：https://doi.org/10.1038/s41598-024-64489-4Share本文与您共享以下链接的任何人都可以阅读此内容：获取可共享链接对不起，本文目前没有可共享的链接。复制到剪贴板。

Provided by the Springer Nature SharedIt content-sharing initiative

由Springer Nature SharedIt内容共享计划提供

KeywordsInfantile skullSkull sutureInfant ageCraniosynostosisX-ray

关键词婴儿颅骨颅骨缝合婴儿颅骨X射线

CommentsBy submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

评论通过提交评论，您同意遵守我们的条款和社区指南。如果您发现有虐待行为或不符合我们的条款或准则，请将其标记为不合适。