Douglas T S, Vaughan C L, Wynne S M
MRC/UCT Medical Imaging Research Unit, Department of Human Biology, University of Cape Town, South Africa.
Med Biol Eng Comput. 2004 Jan;42(1):37-43. doi: 10.1007/BF02351009.
A stereo-photogrammetric method for three-dimensional reconstruction of points in low-dose digital X-ray images obtained using a scanner with similar imaging geometry to that of computed tomography scan projection radiography, was analysed. A calibration frame containing 25 radio-opaque markers with known three-dimensional locations was scanned, and the accuracy of reconstruction of the marker positions under varying control point configurations and separation angles was assessed. Errors of less than 1 mm were obtained when nine test points were reconstructed, with 16, 11 and 7 control points at a 90 degrees separation angle, and with 16 and 11 control points at 75 degrees and 60 degrees separation angles. The optimum reconstruction, with a resultant error of 0.68 mm, was found to occur at a separation angle of 90 degrees, with the largest number of control points (16) used to calculate the parameters of the transformation. Extrapolation in the scanning direction beyond the space defined by the control points gave errors of less than 2 mm. This method should be suitable for three-dimensional point reconstruction in applications such as cephalometry, brachytherapy planning and assessment of spinal shape.
分析了一种立体摄影测量方法,用于对使用与计算机断层扫描投影射线照相成像几何结构相似的扫描仪获得的低剂量数字X射线图像中的点进行三维重建。扫描了一个包含25个具有已知三维位置的不透射线标记的校准框架,并评估了在不同控制点配置和分离角度下标记位置重建的准确性。当重建九个测试点时,在90度分离角下使用16、11和7个控制点,以及在75度和60度分离角下使用16和11个控制点时,获得了小于1毫米的误差。发现最佳重建结果是在90度分离角下,使用最大数量的控制点(16个)来计算变换参数,此时合成误差为0.68毫米。在扫描方向上超出控制点定义的空间进行外推时,误差小于2毫米。该方法应适用于诸如头影测量、近距离放射治疗计划和脊柱形状评估等应用中的三维点重建。