基于系统仿真的数字乳腺层析成像性能分析

目的 基于系统仿真的数字乳腺层析成像（DBT）进行性能分析研究。方法 基于临床DBT系统获取不同辐射剂量下物理体模和常规剂量水平下临床患者的原始测量数据，并进行低剂量仿真和使用3种重建算法进行重建，即Feldkamp-Davis-Kress（FDK）算法、联合代数重建技术（SART）、具有全变分约束的自适应最速下降凸集投影（ASDPOCS-TV）算法。采用信噪比（SNR），峰值信噪比（PSNR），噪声功率谱（NPS），伪影扩散函数（ASF）以及 ASF的全宽半高值（FWHM）指标对辐射剂量水平和重建算法两因素对图像质量的影响进行统计对比分析。结果 低剂量DBT仿真策略的有效性经评估实验得以保证；在合适的剂量水平范围内，增加剂量可促进减少高频噪声成分和显著提高信噪比（P<0.05），当曝光剂量低于40 mAs，不同剂量水平下的图像具有相似表征；不同重建算法在不同解剖结构区域处的表现性能有差异，ASDPOCS-TV算法的结果包含更少的层间伪影和更轻微的噪声，优于SART 算法和FDK算法。结论 辐射剂量和重建算法的选择对DBT成像质量有显著影响，应当平衡各影响因素、整体图像质量、临床诊断需求之间的关系，以达到临床任务中最优成像性能的目标。

Assessment of imaging performance of digital breast tomosynthesis based on systematic simulation

Objective To assess the imaging performance of digital breast tomography (DBT) based on systematic simulation. Methods The raw measurements of physical phantoms at a variety of radiation dose levels and clinical patients at the normal radiation dose were acquired from a clinical DBT system for low-dose simulation and reconstruction using 3 reconstruction algorithms, namely Feldkamp-Davis-Kress (FDK), simultaneous algebraic reconstruction technique (SART) and adaptive steepest-descent projection onto convex sets with total-variation constraint (ASDPOCS-TV) algorithms. The image quality was compared across different radiation dose levels and reconstruction algorithms in terms of signal- to-noise ratio (SNR), peak signal-to- noise ratio (PSNR), noise-power spectrum (NPS), artifact spread function (ASF) and full width at half maximum (FWHM) of ASF indexes. Results The reliability of low-dose DBT simulation strategy was verified by the experiment. Within a suitable range of dose levels, increasing the doses resulted in reduced high-frequency noise component and significantly increased SNR (P<0.05). But when the value of exposure was below 40 mAs, the images acquired at different dose levels had similar representation. The performance of the 3 reconstruction algorithms varied for different anatomical structures, and the image quality of ASDPOCS-TV algorithm was generally superior to SART and FDK algorithms with less through-plane artifacts and noise. Conclusion The quality of DBT images is significantly affected by both radiation dose and reconstruction algorithms. A tradeoff of the parameters, the overall image quality and the clinical needs for diagnostic purposes should be considered to achieve the optimal imaging performance on a given clinical task.