双探头投影偏差所致伪影的识别与校正

Artifacts caused by misalignment of projection in dual head SPECT: effect, detection and correction

目的探讨双探头ＳＰＥＣＴ仪采集中，由于投影位置偏差使２组图像连接处出现偏移，造成图像伪影与偏差的识别和校正。方法①模型采集：２只探头置成９０°（Ｌ模式）或１８０°（Ｈ模式）夹角，内插球形冷区的Ｊａｓｚｃｚａｋ模型和心肌灌注模型采集。用单探头采集的方法模拟投影偏差，关闭１只探头，在采集全过程的１／２处移动模型。②偏差探测：方法一是２只探头在同一角度采集点源图像，计算２幅点源矩心的偏差；方法二是在２只探头相邻的角度采集点源图像，计算２幅点源矩心的偏差。在１台ＳＰＥＣＴ仪上将这２种方法与移动探测软件进行比较测试。③偏差校正：对模型采集的所有偏差图像，采用移动校正软件进行校正。结果在Ｈ模式中，随着偏差增大，图像逐渐模糊和失真。Ｌ模式中靶心图显示，Ｙ方向偏差造成心尖部缺损伪影，Ｘ方向偏差造成多位置缺损伪影。２种偏差探测方法得到相同结果（Ｙ方向偏差＝３４５ｍｍ，Ｘ方向＝０），与移动探测软件结果一致。模型采集的所有偏差图像经过移动校正软件校正，重建后的图像或靶心图恢复正常，与没有偏差时相同。结论投影偏差导致图像伪影，其中Ｌ采集模式Ｘ方向大于６９ｍｍ偏差，Ｌ模式Ｙ方向和Ｈ模式大于１３８ｍｍ的偏差，将可能造成假阳性或假?

Objective To study the effect, detection and correction of artifacts due to misalignment of projection matrix, which occur in jointed projections acquired respectively by two heads in dual head SPECT Methods ①Phantom acquisition: setting the included angle of two heads in 180° (H mode), a Jaszczak phantom with six spheroids inserted around it was imaged; and setting the included angle of two heads in 90° (L mode), a cardiac phantom was imaged To simulate the misalignment, a single head acquisition was used and the phantom was moved 1 pixel (6 9 mm), 2 pixels (13 8 mm) and 3 pixels (20 7 mm) forward at half of acquisition process ②Misalignment detection and correction: two heads acquired a point source in same angle or in adjacent angle, then the misalignment of two point images was worked out In the separated angle method, a motion caused by a rotation of head was compensated for, because the principle of projection misalignment is similar to patient motion, a patient motion detection and correction program could be employed to detect and correct the misalignment Results In H mode, when the misalignment was increased, images became blurry and distortional In L mode, polar maps of cardiac phantom showed that Y axial misalignment resulted in an apex defect and X axial misalignment resulted in multiple defects Two misalignment detection methods, in same angle and adjacent angle, could get results equal to patient motion detection and with a patient motion detection program all artifacts were removed on above phantom images Conclusions Projection misalignment causes artifacts in SPECT More than 13 8 mm X and Y axial misalignment in H mode, more than 6 9 mm X axial and more than 13 8 mm Y axial misalignment in L mode may lead to false positive and negative results Two simple methods are effective to detect the misalignment Patient motion correction or center of rotation correction could be used to remove the artifacts