在常规CT成像系统中, 发出的X射线是连续能谱的, 导致重建图像出现硬化伪影, 影响了材料组分区分, 无法进行定量表征。 解决这一问题的关键在于实现多能谱CT成像, 利用多个窄能谱段或单能量CT图像, 提高组分与图像灰度的对应性。 相对于传统CT, 能谱CT具有更强的组分区分能力, 有利于实现物体组分的定量分析。 现有的基于光子计数探测器多谱CT在成像时间分辨率和空间分辨率存在局限, 基于能谱滤波的多谱CT能谱区分度受限。 而基于变电压多谱投影序列盲分离的多谱CT, 通过分解连续能谱投影, 获取窄能谱投影, 进而实现能谱CT成像, 确保物质组分与重建图像灰度值的对应性, 实用性较强。 但是由于X射线能谱和物体组分的未知, 在盲分离过程中, 衰减系数未知, 并且能谱划分是不确定的, 导致窄能谱CT重建图像的能量指向性不强, 对应能量值与参考能量偏差偏大, 影响组分定量分析。 因此, 针对盲分离中能谱划分不确定性和重建图像能量指向性问题的开展研究。 利用衰减系数的光电效应和康普顿效应分解, 构建能量约束, 消除能量的不确定性, 降低分解所得投影重建图像的能量与参考能量值的偏差。 在基于以残差的局部方差和最小为优化目标的分解模型中, 将分解模型中的衰减系数按光电效应和康普顿效应分解为能量项和材料项, 利用能量项的可预知性, 依据预先划分的窄能谱段设置其值, 固定各分解投影对应的窄能谱段, 作为对能量的约束条件。 求解所得各分解投影为能量已知投影, 对其重建可得到能量确定的各窄能谱段的图像。 选择衰减系数相近的硅铝材质构成外硅内铝圆柱体进行实验验证, 在有能量约束的求解结果中, 硅铝衰减系数与参考值偏差小于无能量约束, 所得重建图像中硅铝变化率与理论值趋势较一致, 能量指向性强, 与参考能量偏差降低。 结果表明, 所提方法解决了基于变电压序列盲分离多谱CT成像的能量指向问题, 能谱分辨率更高, 组分表征更准确。

In X-ray CT imaging system, the polychromatic X-ray results in beam hardening artifacts, which influence the material composition distinction. It can’t realize quantitative characterization. The multi-spectrum CT can reach the higher correspondence between the composition and image gray by narrow-energy-width or monoenergetic CT. Compared with traditional CT, spectral CT can distinguish the different component. The implement method based on the photon counting detector is limited in temporal resolution and spatial resolution. The implement method based on filter doesn’t have enough discrimination. And the implement method based on blind separation of multi-voltage projection sequences has better practicability. It guarantees the correspondence between the materials and the gray of reconstructed image. However, the attenuation coefficients are unknown and energy spectrum division is uncertain, the analytic energy value of reconstructed images is ambiguous. The error between the analytic energy and referenced energy is higher. Then it will influence the precision of the multi-component quantitative analysis. For this problem, an improved method is proposed. It utilizes the attenuation coefficient decomposition of compton effect and photoelectric effect as an energy constraint to eliminate the uncertainty energy partition. The error between the energy value of reconstructed images with decomposed projection and the reference energy value is reduced. In the decomposition model, the optimum object function is local variance sum of residual error minimum. The attenuation coefficient is decomposed as energy dependency term and material dependency term according to compton effect and photoelectric effect. The energy dependency term can be known in advance. It can be as energy constraint and used to fix the energy value of narrow-energy-width. Then the energy of decomposed projection is known and the energy of corresponding reconstructed images also is known. A cylinder composed of aluminum and silicon is used in the verification experiment since aluminium and silicon have approximate attenuation coefficients. The error between the attenuation coefficients of reconstructed images with the energy constraint is less than the result of reconstructed images without the energy constraint. The contrast tendency of silicon and aluminium with energy is close to the theoretical value. Also the difference with reference energy is reduced. The result shows that the proposed method solves the energy directivity problem of multi-spectrum CT based on the blind separation of multi-voltage projection sequences. The energy spectrum resolution ratio is higher. The composition representation is more accurate.