Automatic Segmentation Method of Phalange Regions Based on Residual U-Net and MSGVF Snakes
Automatic Segmentation Method of Phalange Regions Based on Residual U-Net and MSGVF Snakes
复制标题
DOI：
10.23919/iccas47443.2019.8971740
复制
发表时间：
2019-10
期刊：
影响因子：
--
通讯作者：
Kohei Kawagoe;Kazuhiro Hatano;S. Murakami;Huimin Lu-;Hyoungseop Kim;T. Aoki
中科院分区：
文献类型：
--
作者：
Kohei Kawagoe;Kazuhiro Hatano;S. Murakami;Huimin Lu-;Hyoungseop Kim;T. Aoki
骨病包括类风湿关节炎和骨质疏松症。虽然使用计算机x线摄影（CR）图像进行视觉筛查是诊断骨质疏松症的有效方法，但也有一些类似的疾病表现为低骨量状态。为此，我们的目标是开发一个计算机辅助诊断（CAD）系统，以支持骨质疏松症的CR图像自动诊断。在本文中，我们使用卷积神经网络（CNN）和多尺度梯度向量流蛇（MSGVF snake）算法从CR图像中分割出每个指骨区域。将该方法应用于15例检测，得到的真阳性率为92.95%，假阳性率为2.21%，假阴性率为7.05 %。
Bone diseases include rheumatoid arthritis and osteoporosis. Although visual screening using computed radiography (CR) images is an effective method for diagnosing osteoporosis, there are some similar diseases that exhibit low bone mass status. To this end, we aim to develop a computer-aided diagnostic (CAD) system to support the automatic diagnosis of osteoporosis from CR images. In this paper, we use convolutional neural network (CNN) and multiscale gradient vector flow snakes (MSGVF Snakes) algorithms to segment each finger bone regions from the CR image. The proposed method is applied to 15 cases, 92.95 [%] of the true positive rates, 2.21 [%] of the false positive rates, 7.05 [%] of the false negative rates are obtained respectively.