首页> 外文会议>SPIE Medical Imaging Conference >Reduction of Unnecessary Thyroid Biopsies using Deep Learning
【24h】

Reduction of Unnecessary Thyroid Biopsies using Deep Learning

机译:使用深度学习减少不必要的甲状腺活组织检查

获取原文

摘要

Thyroid nodules are extremely common lesions and highly detectable by ultrasound (US). Severalstudies have shown that the overall incidence of papillary thyroid cancer in patients with nodulesselected for biopsy is only about 10%. Therefore, there is a clinical need for a dramatic reduction ofthyroid biopsies. In this study, we present a guided classification system using deep learning thatpredicts malignancy of nodules from B-mode US. We retrospectively collected transverse andlongitudinal images of 150 benign and 150 malignant thyroid nodules with biopsy proven results.We divided our dataset into training (n=460), validation(n=40), and test (n=100) datasets. Wemanually segmented nodules from B-mode US images and provided the nodule mask as a secondinput channel to the convolutional neural network (CNN) for increasing the attention of noduleregions in images. We evaluated the classification performance of different CNN architectures suchas Inception and Resnet50 CNN architectures with different input images. The InceptionV3 modelshowed the best performance on the test dataset: 86% (sensitivity), 90% (specificity), and 90%precision when the threshold was set for highest accuracy. When the threshold was set formaximum sensitivity (0 missed cancers), the ROC curve suggests the number of biopsies may bereduced by 52% without missing patients with malignant thyroid nodules. We anticipate that thisperformance can be further improved with including more patients and the information from otherultrasound modalities.
机译:甲状腺结节是极常见的病变,通过超声(US)高度可检测的病变。一些研究表明,结节患者乳头状甲状腺癌的总发病率选择活组织检查仅为10%。因此,临床需要急剧减少甲状腺活组织检查。在这项研究中,我们展示了一个使用深度学习的引导分类系统从B模式预测Nodules的恶性肿瘤。我们回顾性地收集横向和纵向图像150良性和150个恶性甲状腺结节,活组织检查成熟结果。我们将DataSet划分为训练(n = 460),验证(n = 40),测试(n = 100)数据集。我们手动分割来自B模式美国图像的结节,并提供了Nodule掩码作为秒卷积神经网络(CNN)的输入通道,用于增加结节注意力图像中的区域。我们评估了不同CNN架构的分类性能作为具有不同输入图像的开始和Reset50 CNN架构。 Inceptionv3模型在测试数据集中展示了最佳性能:86%(灵敏度),90%(特异性)和90%精度阈值设置为最高精度时。当阈值设置为最大敏感性(0个错过癌症),ROC曲线表明活检的数量可能是减少52%而不缺少恶性甲状腺结节患者。我们预料到这一点可以通过包括更多患者和来自其他人的信息进一步改善性能超声方式。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号