Program in Health and Welfare, Graduate School of Comprehensive Scientific Research, Prefectural University of Hiroshima, Mihara, Hiroshima, Japan.
Ann Nucl Med. 2012 Jan;26(1):16-25. doi: 10.1007/s12149-011-0534-7. Epub 2011 Sep 10.
A normal perfusion database (NDB) is imperative for the statistical imaging of brain function. This study validates a novel NDB created under the same injection dose and acquisition conditions for three gamma camera systems and evaluates optimal pre-filter parameters for three-dimensional stereotactic surface projections (3D-SSPs).
We compared a novel NDB that matched the databases in each of three vendor gamma camera systems with a conventionally constructed NDB (conventional NDB) and a NDB constructed in-house for 3D-SSP. We generated hypoperfused regions where pre-specified volumes were simulated for various areas in SPECT images. The properties of each NDB were evaluated based on the distribution of the standard deviation (SD). Abnormal accumulation regions were validated using Z, extent, and artifactual scores. Detection error was used to evaluate the optimal Butterworth pre-filter cutoff frequency with the perfusion defect rate (PDR) in 3D-SSP.
The SD distribution was the same in the novel NDB and in the NDB constructed in-house, and the SD of the peak distribution was 0.08-0.07. The Z and extent scores of the novel DB and the NDB constructed in-house were similar, but increased along with the artifactual scores when using the conventional NDB. Many artifacts appeared in the Z score map when using the conventional NDB. The detection error deviated from the actual value by -1.3% at a cutoff frequency of 0.58 cycles/cm and a PDR of 30%, which was the lowest. The cutoff frequency became lower or higher, and the low-perfusion defect rate increased according to the increasing detection error. The optimal cutoff frequency was between 0.52 and 0.58 cycles/cm.
We generated a novel NDB according to the individual devices and compared it with a conventional and a NDB constructed in-house. The Z and extent scores were essentially equal when using the novel DB and the NDB constructed in-house, but considerably differed when using the conventional NDB. The optimal cutoff frequency of the Butterworth filter evaluated from the detection error was in the range of 0.52-0.58 cycles/cm. The detection error increased the perfusion defect rate by <15% and this was undetectable in 3D-SSP. The next step will be to improve the accuracy of the extent of abnormal regions and the sensitivity of the Z score using a novel NDB constructed according to the individual devices.
正常灌注数据库（NDB）对于脑功能的统计成像至关重要。本研究验证了在相同注射剂量和采集条件下为三种伽马相机系统创建的新型 NDB，并评估了用于三维立体定位表面投影（3D-SSP）的最佳预滤波器参数。
我们比较了一种新型 NDB，该数据库与三种供应商伽马相机系统的数据库相匹配，以及为 3D-SSP 构建的常规 NDB 和内部构建的 NDB。我们在 SPECT 图像的各个区域模拟了指定体积的低灌注区域。根据标准偏差（SD）的分布评估每个 NDB 的特性。使用 Z、程度和人工评分验证异常积聚区域。使用 3D-SSP 中的灌注缺陷率（PDR）评估最佳巴特沃斯预滤波器截止频率的检测误差。
新型 NDB 和内部构建的 NDB 的 SD 分布相同，峰值分布的 SD 为 0.08-0.07。新型 DB 和内部构建的 NDB 的 Z 和程度评分相似，但使用常规 NDB 时会随着人工评分的增加而增加。使用常规 NDB 时，Z 评分图中出现了许多伪影。当截止频率为 0.58 个周期/cm 且 PDR 为 30%时，检测误差偏离实际值-1.3%，这是最低的。随着检测误差的增加，截止频率变低或变高，低灌注缺陷率增加。最佳截止频率在 0.52 到 0.58 个周期/cm 之间。
我们根据单个设备生成了一个新型 NDB，并将其与常规 NDB 和内部构建的 NDB 进行了比较。使用新型 DB 和内部构建的 NDB 时，Z 和程度评分基本相同，但使用常规 NDB 时则有很大差异。从检测误差评估的巴特沃斯滤波器的最佳截止频率在 0.52-0.58 个周期/cm 之间。检测误差将灌注缺陷率增加了<15%，在 3D-SSP 中无法检测到。下一步将根据单个设备构建新型 NDB，以提高异常区域程度的准确性和 Z 评分的敏感性。
