评价Karl迭代重建技术对胸部CT图像质量的影响

目的 评价Karl迭代重建技术对胸部CT图像质量的影响。 方法 ①模体研究:以管电压120 kVp、管电流140 mAs扫描为常规剂量组；降低管电流50%，以管电压120 kVp、管电流70 mAs扫描为低剂量组，2组分别采用滤波反投影（FBP）和Karl迭代重建技术（重建等级1~9级）进行图像重建，采用噪声功率谱（NPS）和标准差对图像噪声进行测量。②临床研究:基于模体研究结果，选取行胸部CT的受检者120例，其中男性61例、女性59例，年龄35~75岁，BMI为（23.95±0.27） kg/m2。采用随机数字表法分成常规剂量组和低剂量组，每组分别为60例，扫描参数同模体研究，分别采用Karl 5级迭代重建技术和FBP法进行图像重建。比较2组CT容积剂量指数（CTDIvol）、剂量长度乘积（DLP）、有效辐射剂量（ED）及图像噪声、信噪比等客观指标和图像质量主观评分指标。客观指标的比较采用t检验；主观评分的比较采用χ2检验。 结果 ①模体研究:常规剂量组采用Karl迭代重建技术重建图像的平均噪声均低于FBP法重建图像噪声，且随着Karl迭代重建技术等级的升高而降低，差异均有统计学意义（t=5.14~47.50，均P<0.01）。通过NPS曲线对比，Karl 1~9级重建图像在降低图像噪声的同时，保持了与FBP法重建图像的噪声纹理特性，差异无统计学意义（t=2.49， P=0.42）。低剂量组Karl 5级迭代重建图像的噪声[（6.40±0.16） Hu]与常规剂量组FBP重建图像的噪声[（6.30±0.38） Hu]较其他Karl迭代重建技术等级更接近，差异无统计学意义（t=28.34，P=0.423）。②临床研究:低剂量组的CTDIvol[（5.56±0.01） mGy]、DLP[（170.74±18.40） mGy]均明显低于常规剂量组[（11.06±0.01） mGy、（348.93±26.16） mGy·cm]，差异有统计学意义（t=4757.7，P=0.003；t=39.23，P=0.005）；ED[（2.58±0.16） mSv]较常规剂量组[（5.01±0.17） mSv]降低了51.5%，差异有统计学意义（t=37.94，P=0.004）。低剂量Karl 5级迭代重建技术重建图像与常规剂量FBP重建图像比较，噪声（升主动脉:t=0.24，P=0.38; 降主动脉:t=1.51，P=0.70）和信噪比（升主动脉:t=0.45，P=0.45; 降主动脉:t=0.08，P=0.72）的差异均无统计学意义；纵隔窗图像（χ2=2.32，P=0.317; χ2=1.38，P=0.268）和肺窗图像（χ2=0.97，P=0.614; χ2=0.59，P=0.760）的主观图像质量评分比较，差异均无统计学意义。 结论 Karl迭代重建技术可以不同程度地降低图像噪声。降低常规管电流的50%至70 mAs、采用Karl 5级迭代重建技术重建图像可获得与常规剂量FBP相同的图像质量。

Effect of Karl iterative reconstruction technique for chest CT image quality

Objective To investigate the effect of Karl iterative reconstruction on CT image quality enhancement. Methods ① Phantom study: A 120 kVp/140 mAs tube current was set as the standard radiation dose. Low-dose setting is implemented with a 50% current reduction, i.e., 70 mAs, whereas the voltage is set to 120 kVp. Two experimental groups with algorithms of filtered back projection(FBP) and Karl iterative reconstruction(with a noise reduction of 1–9 levels) were compared. The noise power spectrum(NPS) and standard deviation metrics were adopted to assess the noise degree. ② Clinical study: In the phantom study, 120 patients, including 61 males and 59 females with ages ranging from 35 to 75 and a BMI of(23.95±0.27) kg/m2, were recruited and randomly divided into two groups: standard dose and low dose. These groups were scanned via FBP and Karl iterative reconstruction at level 5(Karl 5), respectively. The image qualities of the two groups were assessed with various objective metrics, such as CT dose index of volume(CTDIvol), dose length product(DLP), effective dose(ED), and signal-to-noise ratio(SNR), as well as by human subjective evaluation. Student t-test was adopted to assess the significance of difference for the values of objective metrics from the two groups, whereas subjective evaluation was quantified with an χ2 test. Results ① Phantom study: With the standard dose, the average noise degree for Karl iterative reconstruction was lower than that for FBP. With an increased level of Karl iterative reconstruction, the noise degree will be lowered(t=5.14–47.50, all P<0.01). Referring to the NPS curves, the Karl 1?9 algorithm can attain the goal of noise reduction and remain keep the image texture unchanged(t=2.49, P=0.42). With the low dose, i.e., 50% current reduction, the image noise[(6.40±0.16) Hu] of Karl 5 was close to that[(6.30±0.38) Hu] of FBP with insignificant difference(t=28.34, P=0.423), compared with other levels of Karl iterative reconstruction. Meanwhile, the magnitudes of NPS curves for Karl 5 and FBP insignificantly differed. ② Clinical study: The CTDIvol[(5.56±0.01) mGy] and DLP[(170.74±18.40) mGy·cm]value of the low-dose group were significantly lower than those[(11.06±0.01) mGy, (348.93±26.16) mGy·cm] of the standard-dose group(t=4757.7, P=0.003; t=39.23, P=0.005). The ED values[(2.58±0.16) mSv] of the low-dose group were significantly less by 51.5% than those[(5.01±0.17) mSv] of the standard-dose group(t=37.94, P=0.004). The noise degree and SNR values for images from Karl 5 with low dose and FBP with standard dose were insignificantly different(noise degree, t=0.24, P=0.38; t=1.51, P=0.70; SNR, t=0.45, P=0.45; t=0.08, P=0.72). The results of subjective quality assessments for the images from Karl 5 and FBP were insignificantly different in terms of the usage of mediastinum(χ2=2.32, P=0.317; χ2=1.38, P=0.268) and lung(χ2=0.97, P=0.614; χ2=0.59, P=0.760 ) window settings for image reading. Conclusions Karl iterative reconstruction at different levels can effectively reduce noise with different degrees. For the 50% current reduction to 70 mAs, the image quality of Karl 5 is comparable with that of standard-dose FBP.