遵义医科大学学报2024，Vol.47Issue(4)：384-391,8.
人工智能迭代重建算法对超低剂量胸部CT图像质量和计算机辅助肺结节检测的影响
Impacts of artificial intelligence iterative reconstruction algorithm on the image quality and computer-aided pulmonary nodule detection at ultra-low dose chest CT
摘要
Abstract
Objective To investigate the effect of artificial intelligence iterative reconstruction(AIIR)algorithm on the image quality and computer-aided pulmonary nodule detection at ultra-low dose chest CT.Methods Forty-one patients who underwent chest CT examination in our hospital from September to October 2023 were prospec-tively enrolled.Conventional dose(120 kVp,using tube current modulation technology,dose level 2,reference tube current 106 mAs)and ultra-low dose(120 kVp,15 mAs)chest CT were simultaneously collected,the ra-diation dose parameters were recorded.Conventional dose CT was reconstructed by Karl iterative reconstruction and ultra-low dose CT was reconstructed by Karl and AIIR(level 1,3,and 5).The CT value and noise index(Standard deviation,SD)of the aorta,fat and muscle areas in the five groups of images were measured and the signal to noise ratio(SNR)and contrast to noise ratio(CNR)were calculated.The overall image quality of the five groups of images was subjectively evaluated.Using ≥4 mm non-calcified solid pulmonary nodules determined by two chest radiologists who jointly reviewed the images as the reference standard,the true positive,false posi-tive(misdiagnosis),and false negative(missed diagnosis)pulmonary nodule numbers of the five groups of ima-ges CAD pulmonary nodule detection were recorded and compared with the reference standard.The manual re-view time and CAD review time of the ultra-low dose Karl reconstruction group were recorded.Results There was no significant difference in CT value among the five groups(P＞0.05).The SD,SNR and CNR values of ultra-low dose AIIR reconstruction were better than those of conventional dose and ultra-low dose Karl reconstruction,with statistically significant differences(P＜0.005).The overall image quality of ultra-low dose AIIR5 recon-struction was similar to that of conventional dose Karl reconstruction and was better than that of ultra-low dose Karl reconstruction(P＜0.005).There was no significant difference in sensitivity of pulmonary nodule detection among the five groups(P＞0.05).The false positive rate of ultra-low dose AIIR reconstruction was lower than that of conventional dose and ultra-low dose Karl reconstruction,with statistically significant differences(P＜0.005).The radiation dose of the ultra-low dose group was reduced by about 89.3％compared with that of the conventional dose group,the CAD reading time was reduced by 63.3％compared with that of manual reading.Conclusion AIIR can improve the image quality of ultra-low dose chest CT while maintaining the sensitivity of pulmonary nodule detection and reducing the false positive rate of CAD detection.关键词
胸部CT/肺结节/人工智能迭代重建/超低剂量CTKey words
chest CT/pulmonary nodule/artificial intelligence iterative reconstruction/ultra-low dose CT分类
医药卫生引用本文复制引用
张宝平,王睿,黄欣,郝辉,王怡名,杨健,金超,李傲,李宇航,朱书萌,田倩,赵文哲,肖瑶,侯伟,刘哲..人工智能迭代重建算法对超低剂量胸部CT图像质量和计算机辅助肺结节检测的影响[J].遵义医科大学学报,2024,47(4):384-391,8.基金项目
国家自然科学基金数学天元重点专项(NO:12226007). （NO:12226007）
