深度学习像素闪耀算法改善中低体重指数者腹部70kVp动脉期图像质量

Image Quality Improvement of Abdominal 70kVp Arterial Phase in Patients with Low and Moderate Body Mass Index Using Deep-learning Pixel Shine Algorithm

目的探讨深度学习像素闪耀(PS)算法在提高中低体重指数(BMI)者70 k Vp腹部CT血管成像(CTA)动脉期实质器官图像质量的价值。资料与方法回顾性收集2015年11月至2016年4月因腹部肿瘤或血管性疾病于大连医科大学附属第一医院施行全腹CTA扫描、BMI<22 kg/m2的患者62例,包括低管电压组(70 k Vp,A组)42例、常规管电压组(120 k Vp,B组)20例,A组经PS处理后的图像为C组(42例)。测量各组肝脏CT值、竖脊肌CT值及标准差(SD),并计算肝脏对比噪声比(CNR)、信噪比(SNR)。记录辐射剂量,并对各组图像右肾动脉进行主观质量评分。结果 A、B、C组间肝脏CT值、竖脊肌CT值差异均无统计学意义(P均>0.05);竖脊肌SD值、肝CNR、肝SNR差异均有统计学意义(P均<0.05),C组较A组CNR提高了133.07%(1.69/1.27),SNR提高了137.74%(10.00/7.26)。3组右肾动脉的重建图像评分中位数均为5.00(1.00)分,差异无统计学意义(χ~2=0.381,P>0.05)。A组和B组的CT容积剂量指数分别为(2.05±0.12)m Gy、(7.36±0.89)m Gy,有效剂量分别为(1.53±0.12)m Sv、(5.51±0.72)m Sv,A组的CT容积剂量指数和有效剂量明显低于B组,差异均有统计学意义(P<0.05)。结论针对BMI<22 kg/m^2患者,采用低管电压扫描并PS处理后图像,在不影响实质器官CT值并保证血管成像质量的同时,提高了实质器官的图像质量,有效降低辐射剂量。

Purpose To explore the value of deep learning pixel shine（PS） algorithm in improving the image quality of parenchyma organs of 70 k Vp abdominal CTA arteries in patients with low and moderate body mass index（BMI）.Materials and Methods Sixtytwo patients with abdominal neoplasms or vascular diseases who were treated in the First Affiliated Hospital of Dalian Medical University from November 2015 to April 2016 and underwent full abdominal CTA scanning,BMI〈22 kg/m2 were retrospectively collected： 42 cases in low tube voltage group（70 k Vp,group A）,20 cases in conventional tube voltage group（120 k Vp,group B）,and images after PS treatment in group A were classified as group C（42 cases）.The CT values of the liver,CT values of erector spinae,and standard deviation（SD） in each group were measured,and the liver contrast to noise ratio（CNR） and signal to noise ratio（SNR） were calculated.The radiation dose was recorded,and subjective quality were scored towards the right renal artery in each group of images.All parameters were tested for normality and homogeneity of variance among the three groups.The differences between the parameters were compared.Results The difference of liver CT values and CT values of erector spinae between the three groups showed no statistical significance（P〉0.05）; the difference in terms of SD values of erector spinae,liver CNR and liver SNR all displayed statistical significance（P〈0.05）,with CNR in group C rising to 133.07%（1.69/1.27） as compared with group A,and SNR rising to 137.74%（10.00/7.26）.The median score of the reconstructed images of the right renal artery in the 3 groups was all 5.00（1.00） points,the difference of which was not statistically significant（χ2=0.381,P〉0.05）.The CT dose index of volume in group A and B was（2.05±0.12） m Gy and（7.36±0.89） m Gy,respectively; and effective dose（1.53±0.12） m Sv and（5.51±0.72） m Sv,respectively; the difference of which was statistically significant（P〈0.05）.The CT dose index of volume and effective dose in group A were evidently lower than those in group B.Conclusion For patients with BMI 〈22 kg/m2,the adoption of low tube voltage scan and PS-treated images improved the image quality of parenchyma organs and effectively reduced radiation dose while not affecting the CT value of parenchyma organs and ensuring the quality of angiography.