低剂量条件下深度学习图像重组算法与多模型迭代重组算法对腹部增强CT图像质量的影响

Effects of Deep Learning Image Reconstruction and Adaptive Statistical Iterative Reconstruction Veo on the Image Quality of Abdominal Enhanced CT at Low Dose

目的比较深度学习重组算法(DLIR)和多模型迭代重组算法(ASIR-V)在腹部低剂量CT增强扫描中对图像质量的影响,探讨DLIR在降噪以及提高图像质量等方面的价值。方法采取前瞻性队列研究的方法,对30例患者行腹部四期增强扫描。选择动脉晚期行低剂量扫描,其他三期行常规扫描。所得动脉晚期低剂量原始数据行ASIR-V40%、DLIR-M(中度)、DLIR-H(高度)3种重组方式进行重组。对3组图像进行肝脏、脾脏、主动脉、腰大肌及腹壁脂肪测量CT值、SD值,并计算相应信噪比及对比噪声比。由两位放射科医师对图像质量进行主观评分。结果两位医师对图像的主观评价一致性高。DLIR组所得评分均高于ASIR-V组;DLIR-H与DLIR-M所得肝脏、脾脏、主动脉及腹壁脂肪的SD值均小于ASIR-V40%(P值均<0.001);DLIR-H与DLIR-M肝脏、脾脏、主动脉的SNR、CNR值均大于ASIR-V40%(P值均<0.001),且DLIR-H较DLIR-M更具优势。结论DLIR相较于ASIR-V对于腹部低剂量图像的重组可以更好的降噪和改善图像质量。

Objective The aim of the study was to evaluate the image quality of abdominal low-dose enhancedCTimages using deep learning image reconstruction(DLIR)by comparing with adaptive statistical iterative reconstruction Veo(ASIR-V),and to explore the clinical value of DLIR in noise reduction and image quality improvement.Methods This study prospectively enrolled 30 patients with four stage enhanced abdominal CT scanning.Low-dose scanning was performed at late artery phase and routine scanning was performed at the other three stages.The raw data of late arterial phase was reconstructed using ASIR-V40%、DLIR-M(moderate)and DLIR-H(high).CT values(Hounsfield units)and SD values of liver,spleen,aorta,psoas major and abdominal fat in three groups of images were measured,and SNR and CNR values of liver,spleen,aorta,psoasmajor were calculated.The image quality was subjective estimated by two radiolgists using 5 point score.Results There was a high degree of agreement on the subjective evaluation of images between the two radiologists.The scores of DLIR group were higher than ASIR-V group.The SD values of liver,spleen,aorta and abdominal fat of DLIR-H and DLIR-M groups were lower than that of ASIR-V40%(P<0.001).The SNR and CNR values in liver,spleen and aorta of DLIR-H and DLIR-M were higher than ASIR-V40%(P<0.001).Image quality of DLIR-H group were better than DLIR-M group.Conclusion Compared with ASIR-V,DLIR could reduce noise and improve image quality for low-dose abdominal CT images.