Automated feature detection in dental periapical radiographs by using deep learning
Automated feature detection in dental periapical radiographs by using deep learning
复制标题
DOI：
10.1016/j.oooo.2020.08.024
复制
发表时间：
2021-06-04
影响因子：
2.9
通讯作者：
Khurram, Syed Ali
中科院分区：
文献类型：
--
作者：
Khan, Hassan Aqeel;Haider, Muhammad Ali;Khurram, Syed Ali
目标。本研究的目的是通过使用基于深度学习（DL）的计算机视觉技术来研究根尖周x线片（pr）中常见发现的自动特征检测、分割和量化。研究设计。由3名专家（2名口腔病理学家和1名牙髓科医生）在206张数字pr上标记龋齿、牙槽骨退缩和根间放射率。pr被分为“训练和验证”和“测试”数据集，分别由176个和30个pr组成。在训练过程中，将图像数据的多次变换作为深度神经网络的输入。将现有的和专门构建的深度学习架构的结果进行比较，以确定最适合自动化分析的架构。U-Net架构及其变体在所有指标上都明显优于Xnet和SegNet。在验证数据集上，整体表现最好的架构是“U-Net+Densenet121”（平均交集/联合[mIoU] = 0.501； Dice系数= 0.569）。在“测试”数据集上，所有架构的性能都下降了；“U-Net”表现最佳（mIoU = 0.402； Dice系数= 0.453）。根间放射率最难分割。深度学习具有自动分析pr的潜力，但需要进一步研究。在现有的现成架构中，U-Net及其变体提供了最好的性能。通过专门构建的体系结构和更大的多中心队列，可以获得进一步的性能提升。（口腔外科口腔医学口腔病理口腔放射学2021；131:711-720）
Objective. The aim of this study was to investigate automated feature detection, segmentation, and quantification of common findings in periapical radiographs (PRs) by using deep learning (DL)-based computer vision techniques.Study Design. Caries, alveolar bone recession, and interradicular radiolucencies were labeled on 206 digital PRs by 3 specialists (2 oral pathologists and 1 endodontist). The PRs were divided into "Training and Validation" and "Test" data sets consisting of 176 and 30 PRs, respectively. Multiple transformations of image data were used as input to deep neural networks during training. Outcomes of existing and purpose-built DL architectures were compared to identify the most suitable architecture for automated analysis.Results. The U-Net architecture and its variant significantly outperformed Xnet and SegNet in all metrics. The overall best performing architecture on the validation data set was "U-Net+Densenet121" (mean intersection over union [mIoU] = 0.501; Dice coefficient = 0.569). Performance of all architectures degraded on the "Test" data set; "U-Net" delivered the best performance (mIoU = 0.402; Dice coefficient = 0.453). Interradicular radiolucencies were the most difficult to segment.Conclusions. DL has potential for automated analysis of PRs but warrants further research. Among existing off-the-shelf architectures, U-Net and its variants delivered the best performance. Further performance gains can be obtained via purpose-built architectures and a larger multicentric cohort. (Oral Surg Oral Med Oral Pathol Oral Radiol 2021;131:711-720)