Hough Forest-based Corner Detection for Cervical Spine Radiographs
Hough Forest-based Corner Detection for Cervical Spine Radiographs
复制标题
DOI：
--
复制
发表时间：
2015
期刊：
影响因子：
--
通讯作者：
S. M. R. Al-Arif;Muhammad Asad;K. Knapp;M. Gundry;G. Slabaugh
中科院分区：
文献类型：
--
作者：
S. M. R. Al-Arif;Muhammad Asad;K. Knapp;M. Gundry;G. Slabaugh
颈椎(颈部)对创伤相关损伤高度敏感，急诊医生必须仔细分析。在这项工作中，我们提出了一种基于霍夫森林的颈椎X线片角点检测方法，作为迈向计算机辅助诊断工具的第一步。我们提出了一种新的基于两阶段监督学习(分类和回归)的面片模型来估计颈椎体的角点。我们的方法是使用106张颈部X光图像进行评估的，这些图像包括530个椎骨和2120个角，这些图像已经由专家放射科医生手动标定。结果表明，该算法具有较好的性能，最低中值误差为1.98 mm
The cervical spine (neck region) is highly sensitive to trauma related injuries, which must be analysed carefully by emergency physicians. In this work, we propose a Hough Forest-based corner detection method for cervical spine radiographs, as a first step towards a computer-aided diagnostic tool. We propose a novel patch-based model based on two-stage supervised learning (classification and regression) to estimate the corners of cervical vertebral bodies. Our method is evaluated using 106 cervical x-ray images consisting of 530 vertebrae and 2120 corners, which have been demarcated manually by an expert radiographer. The results show promising performance of the proposed algorithm, with a lowest median error of 1.98 mm