Improving Sensitivity on Identification and Delineation of Intracranial Hemorrhage Lesion Using Cascaded Deep Learning Models
Improving Sensitivity on Identification and Delineation of Intracranial Hemorrhage Lesion Using Cascaded Deep Learning Models
复制标题
DOI：
10.1007/s10278-018-00172-1
复制
发表时间：
2019-06-01
影响因子：
4.4
通讯作者：
Park, Sinyoul
中科院分区：
文献类型：
--
作者：
Cho, Junghwan;Park, Ki-Su;Park, Sinyoul
在急诊室中，高度准确地检测颅内出血是诊断决策和治疗的关键临床问题。在诊断准确性研究的背景下，敏感性和特异性之间存在权衡。为了在保持特异性的同时提高灵敏度，我们提出了一种级联深度学习模型，该模型使用两个卷积神经网络（CNN）和双全卷积网络（FCN）构建。建立级联CNN模型用于识别出血;此后，双重FCN用于检测颅内出血的五种不同亚型并描绘其病变。使用总共135，974张CT图像，包括33，391张标记为出血的图像，每个CNN/FCN模型分别在通过两种不同的窗位/宽度设置预处理的图像数据上进行训练。一个是默认窗口（50/100[水平/宽度]），另一个是笔划窗口设置（40/40）。通过将它们结合起来，与单个CNN和FCN模型相比，我们在出血性病变的二进制分类和分割方面获得了更好的结果。在确定是否出血时，灵敏度提高了约1%（97.91% [+/- 0.47]），同时保持了特异性（98.76% [+/- 0.10]）。对于出血病灶的描绘，我们获得了80.19%的准确率和82.15%的召回率的整体分割性能，与使用单个FCN模型相比提高了3.44%。
Highly accurate detection of the intracranial hemorrhage without delay is a critical clinical issue for the diagnostic decision and treatment in an emergency room. In the context of a study on diagnostic accuracy, there is a tradeoff between sensitivity and specificity. In order to improve sensitivity while preserving specificity, we propose a cascade deep learning model constructed using two convolutional neural networks (CNNs) and dual fully convolutional networks (FCNs). The cascade CNN model is built for identifying bleeding; hereafter the dual FCN is to detect five different subtypes of intracranial hemorrhage and to delineate their lesions. Using a total of 135,974 CT images including 33,391 images labeled as bleeding, each of CNN/FCN models was trained separately on image data preprocessed by two different settings of window level/width. One is a default window (50/100[level/width]) and the other is a stroke window setting (40/40). By combining them, we obtained a better outcome on both binary classification and segmentation of hemorrhagic lesions compared to a single CNN and FCN model. In determining whether it is bleeding or not, there was around 1% improvement in sensitivity (97.91% [+/- 0.47]) while retaining specificity (98.76% [+/- 0.10]). For delineation of bleeding lesions, we obtained overall segmentation performance at 80.19% in precision and 82.15% in recall which is 3.44% improvement compared to using a single FCN model.