單光子射出電腦斷層(SPECT)利用適當核醫示蹤劑，所重建的真實橫軸斷層影像已成為診斷上的重要利器[5]。傳統的SPECT進行影像重建多利用濾波反投影法(FBP)，其優點為速度快，但因多角度的投影至切片，卻也造成了不需要的假影和雜訊的出現。近年來由於電腦運算速度的加快，再加上疊代法運算是根據光子射出的機率統計特性和物理分佈而發展，它減少了雜訊，改善影像品質，也降低了假影的產生，因此會比濾波反投影法有更佳的結果。本文在利用濾波反投影法(Filtered Back Projection, FBP)和疊代運演算法中的序列子集均值與最佳化演算法(ordered subsets expectation maximization, OS-EM)比較兩種運演算法的影像品質。探討1999年8月－2000年8月於臺北市立萬芳醫院核子醫學科做之骨掃描(Bone SPECT)病患，從中選出30人，使用儀器為SIEMENS E. cam雙頭安格閃爍攝影機，裝配低能量，高解晰度平行孔準直儀，注射藥劑為25 mCi之Tc-99m MDP (Methylene-diphosphonate)。造影於注射後三到四小時開始收取影像，取像時問為20秒／幀(frame)，128×128矩陣大小，共記錄128張投影資料。先依機器內建之濾波反投影法（重組濾波器為Butterworth filter）及應用chang's方法衰減校正程式(chang's attenuation correction; pixel width: 4.7952mm, pixel height: 4.7952mm)作影像重組成冠狀切面，矢狀切面和橫切面。另外，再用疊代法將同組資料進行影像重組，疊代法並未進行衰減校正。濾波反投影法和疊代法依其操作難易度，產生假影，操作時間，影像品質來做比較。傳統利用濾波皮投影法來重建影像在核子醫學診斷上佔有重要的地位，但其最大的缺點在於星狀假影的產生以及雜訊的處理效果較差。利用疊代法重建骨掃描影像，是根據光子的統計特性所發展出來，因此，影像品質和對比比濾波反投影法好，假影雜訊也較少，但是其需要大量的資料運算，因此操作時間較長，大約是濾波反投影法的8-9倍。

The goal of clinical nuclear medicine imaging is to aid diagnosis, prognosis and treatment planning and monitoring. In order to do so, nuclear medicine image must be of high diagnostic and quantitative accuracy. Single photon emission computed tomography (SPECT) is generally used today to refer to true trsnsaxial tomography with stand nuclear medicine radiotracers. Filter back-projection (FBP) is most commonly used reconstruction algorithm in SPECT, but produces a streak or star artifacts that result from data being laid onto the slice image from various projections. Because of increase in computer power and improvement in reconstruction algorithms, iterative reconstruction has become clinically available as an alternative to FBP, since they can explicitly model the underlying Poisson statistics of photon emissions as well as the physical properties of a SPECT scanner. Iterative reconstruction has improved noise properties over FBP, and give improved image quality with a reduction in streak artifacts. The purpose of this study is to compare image quality of Tc-99m MDP (Methylene-diphosphonate) bone SPECT images reconstructed by FBP and iterative method based on ordered-subsets expectation maximization (OSEM).