CAJ | 학술논문

设计并实现了一种单光子发射断层成像（ SPECT ）心肌重建图像的左心室长轴自动定位算法。首先通过迭代阈值方法实现原始图像的分割去噪；然后基于连续区块理论，通过先验知识与约束条件自动提取三维感兴趣区（ ROI），排除其他器官与噪声的干扰；下一步对数据进行重采样操作，改善数据质量，并应用形态学中的骨架化算法得到足够的、能够描述心肌轮廓的采样点；在确定合适的、能够在XY方向完整描述心肌形状的平面（α平面）后，通过二次曲线拟合得到XY方向偏角（α角）；随后通过插值方法沿α角与Z轴方向获取新的截面（β平面），并在β平面内二次曲线拟合得到仰角β角，α角与β角便给出了左心室长轴的定位。最后，本算法在原始模体数据、重建数据以及含噪声重建数据各49组测试中，全部定位成功，且平均绝对误差小于5°，并对真实临床SPECT心肌图像数据实现了成功自动定位。
설계병실현료일충단광자발사단층성상（ SPECT ）심기중건도상적좌심실장축자동정위산법。수선통과질대역치방법실현원시도상적분할거조；연후기우련속구괴이론，통과선험지식여약속조건자동제취삼유감흥취구（ ROI），배제기타기관여조성적간우；하일보대수거진행중채양조작，개선수거질량，병응용형태학중적골가화산법득도족구적、능구묘술심기륜곽적채양점；재학정합괄적、능구재XY방향완정묘술심기형상적평면（α평면）후，통과이차곡선의합득도XY방향편각（α각）；수후통과삽치방법연α각여Z축방향획취신적절면（β평면），병재β평면내이차곡선의합득도앙각β각，α각여β각편급출료좌심실장축적정위。최후，본산법재원시모체수거、중건수거이급함조성중건수거각49조측시중，전부정위성공，차평균절대오차소우5°，병대진실림상SPECT심기도상수거실현료성공자동정위。
A new method was developed to automatically determine the orientation of left ventricular long axis from reconstructed myocardial SPECT images .The method consists following steps:firstly, image segmentation and noise reduction are performed using an iterative thresholding method .Secondly, the 3D cardiac region of interest ( ROI) is automatically extracted with a new serial box -region method , which is enhanced by knowl-edge and constraint conditions .Thirdly, the input volume is resampled in order to achieve enough sampling points for data-fitting.In the final data-fitting step,α-plane which well presents the shape of left ventricu-lar is generated, and the azimuth angle αis computed by quadratic curve fitting on α-plane.Similarly, aβ-plane is then generated by interpolating the volume data along the azimuth αand the axis Z , and the elevationβis computed by quadratic curve fitting on the β-plane, and the orientation of the long axis is defined by (α,β) .The method was applied to three groups of testing images from original phantom images , noiseless recon-structed images and noisy reconstructed images .In each group , 49 images with different (α,β) values were tested .A success rate of 100%was achieved , and the mean absolute estimation error was less than 5 degrees in all cases .The method was successfully applied to real clinical SPECT cardiac image data .We conclude that the proposed method is feasible for calculating (α,β) for clinical SPECT application .