| 摘要: | 在核子醫學影像技術的臨床前小動物試驗中，若想使體積小的動物器官重建影像具備與人類器官重建影像同等的精確度，則必須使用比人類器官影像重建更高空間解析度的影像擷取設備。關於投影影像的擷取設備，此論文採用單光子放射顯微鏡系統(Single Photon Emission Microscope,SPEM)，該系統為單光子放射電腦斷層掃描系統(SPECT)的高空間解析度分支，所用設備包含了電子增益電荷耦合元件(EMCCD)、高質量光影像縮倍管、摻鉈碘化銫晶體 CsI(Tl)及七針孔式準直儀。
在電腦斷層掃描技術領域中為求得高品質的重建物體影像，需具備物空間與像空間的轉換關係矩陣，即高解析度且精確的影像系統矩陣。而本研究為建立更精確的影像系統矩陣，因此改善了本實驗室舊有的成像模型計算方法。
過去我們的實驗室致力於獲取 SPEM 的旋轉載物平台的空間關係幾何參數，為此進行了兩項實驗：幾何校正實驗和格點掃描實驗。經過兩項實驗後獲得了物空間範圍內各個位置點源的投影影像，通過分析這些點源的投影影像，我們可以獲得原物體空間的點響應函數，並將其簡化為二維高斯函數，其函數包括六項參數，六項參數分別為光通量、投影點的 x、y 質心座
標、投影點橢圓的主軸方向以及長軸與短軸的長度。將前述兩項實驗投影點vi的 x、y 座標用於幾何參數的擬合，取得空間座標的轉換關係參數，利用此參數進行全域優化，獲得全域座標系。
接著使用此座標系並結合前述之六項高斯參數、格點全域座標，首先進行針孔軸的擬合計算，於取得針孔軸向量之後，建立並改善過去實驗室的成像模型。建立了成像模型後利用此改良後模型建立不同格點間距的影像系統矩陣，並將其用於序列子集之期望值最大化演算法來進行影像重建以得到物體影像，並且為了提高影像重建的解析度及對比度，本研究於序列子集之期望值最大化演算法重建中及重建後分別對於物體重建的結果，皆應用移變點響應函數且使用反卷積演算法來進行去模糊演算，並且比較於影像重建演算法中及後進行去模糊演算的重建結果差異。
;In preclinical small animal experiments of nuclear medicine imaging technology, it is necessary to use imaging equipment with higher spatial resolution than that used for human organ imaging. The goal is to make the reconstructed images of small animal organs have the same precision as the reconstructed images of human organs. This thesis utilizes the Single Photon Emission Microscope (SPEM) as the imaging system to acquire projection images. SPEM is a specialized branch of the Single Photon Emission Computed Tomography (SPECT) with high spatial resolution. It contains a seven-pinhole collimator, a thallium-doped cesium iodide crystal CsI (Tl), an electrostatic de-magnifier tube, and an Electron-Multiplying Charge-Coupled Device (EMCCD).
To achieve high-quality reconstruction of object images in the field of computed tomographic technology, it requires a matrix that establishes the transformation relationship between the object space and the image space, that is, a high-resolution and accurate imaging system matrix. This study aims to establish a more accurate imaging system matrix by improving and optimizing the steps and calculation methods used in our laboratory′s previous imaging system
matrix.
In the past, our laboratory had focused on obtaining the geometric parameters between the detector, the translation stages, and the rotating platform for SPEM. For this purpose, two experiments were conducted: a Geometric Calibration experiment and a Grid-Scan experiment. After these two experiments, we obtained the projection images of point sources at various positions within the object space. By analyzing the projection images of these point sources, we got the point response functions of the object space and simplified them into two- dimensional Gaussian functions with six parameters. These six parameters are flux, x and y centroids of the point source projection, the principal axis angle of the ellipse-like projection image, and the length parameters of the major and minor axes. The x and y centroid coordinates were used to fit the geometric parameters between the detector, the translation stages, and the rotating platform for subsequently obtaining a global coordinate system.
Using this coordinate system in combination with the previously mentioned six Gaussian parameters and the global coordinates of the grid, we first calculated the pinhole-axis fitting. After obtained the pinhole-axis vectors, this study established and improved the previous imaging model from our laboratory. We then used this improved model to create imaging system matrices of different voxel spacings. These different matrices will be used in the Ordered-Subset
Expectation Maximization (OSEM) algorithm to reconstruct the object image. To enhance the resolution and contrast of the reconstructed object images, we applied the shift-variant point response functions and used the deconvolution algorithm during and after the OSEM reconstruction to perform deblurring calculations. Additionally, we compared the differences in reconstruction results between the original OSEM reconstruction, and the reconstructions with the deblurring algorithm during and after the OSEM reconstructions. |
