人工智能CT定量参数与磨玻璃结节样肺腺癌PARP9表达水平的关系及预测患者预后的价值分析  

Relationship between artificial intelligence CT quantitative parameters and the expression level of PARP9 in ground-glass nodular lung adenocarcinoma and its prognostic value

作　　者：高洁 欧阳雪晖 吴伟 韩雪 张鹏[3] GAO Jie;OUYANG Xuehui;WU Wei;HAN Xue;ZHANG Peng(Department of Imaging Medicine,People's Hospital of Inner Mongolia Autonomous Region,Hohhot 010020,China;Department of Medical Records,People's Hospital of Inner Mongolia Autonomous Region,Hohhot 010020,China;Department of Imaging Diagnosis,Affiliated Hospital of Inner Mongolia Medical University,Hohhot 010050,China)

机构地区：[1]内蒙古自治区人民医院影像医学科,呼和浩特010020 [2]内蒙古自治区人民医院病案室,呼和浩特010020 [3]内蒙古医科大学附属医院影像诊断科,呼和浩特010050

出　　处：《新疆医科大学学报》2023年第10期1301-1306,共6页Journal of Xinjiang Medical University

基　　金：内蒙古自治区卫生健康科技计划项目(202201031);内蒙古医科大学高等教育教学改革研究项目(NYTXGG2022169)。

摘　　要：目的 研究人工智能CT定量参数与磨玻璃结节样肺腺癌PARP9表达水平的关系及患者预后预测的价值。方法 选择2017年1月-2019年6月内蒙古自治区人民医院影像医学科收治的137例磨玻璃结节样肺腺癌患者为研究对象。对所有患者进行胸部高分辨率CT扫描,图像导入AI系统中得到肺结节长径、平均CT值、实性成分占比等定量参数。对所有患者进行定期随访,随访日期截至2022年6月30日,以无进展生存(Progress free survival, PFS)时间作为随访时间。结果 137例患者中,11例(8.03%)发生淋巴结转移,126例(91.97%)未发生淋巴结转移;83例(60.58%)为浸润性肺腺癌,54例(39.42%)为非浸润性肺腺癌;平均肺结节长径为(11.15±3.32)mm,平均CT值为(-565.52±95.67)Hu,实性成分占比(55.42±14.27)%。与非浸润性肺癌患者比较,浸润性肺癌患者肺结节长径、平均CT值和实性成分占比均升高,差异有统计学意义(P<0.05)。与PARP9低表达患者比较,PARP9高表达患者浸润性、肺结节长径、平均CT值和实性成分占比均较大,差异有统计学意义(P<0.05)。本研究患者无复发生存率(Recurrence free survival, RFS)为77.37%。单因素分析结果显示,患者预后与淋巴结转移、病理分期、平均CT值及实性成分占比有关,差异均有统计学意义(P<0.05)。选取单因素分析中P<0.1的变量纳入多因素Cox回归分析,结果显示病理分型和平均CT值是影响患者RFS的独立预后因素(P<0.05)。结论 人工智能CT定量参数及病理分型有助于预测磨玻璃结节样肺腺癌患者预后,并且与PARP9表达水平有关,具有一定的参考价值。Objective To study the relationship between artificial intelligence CT quantitative parameters and the expression level of PARP9 in ground-glass nodular lung adenocarcinoma and the prognostic value of the patients.Methods A total of 137 patients with ground-glass nodular lung adenocarcinoma treated in the hospital from January 2017 to June 2019 were selected as the research objects.All patients were scanned with high-resolution chest CT,and the images were imported into AI system to obtain quantita-tive parameters such as lung nodule diameter,average CT value,and proportion of solid components.All patients were followed up regularly until June 30,2022,with progress free survival(PFS)as the follow-up time.Results Among 137 patients,11(8.03%)had lymph node metastasis,126(91.97%)had no lymph node metastasis,83(60.58%)had invasive lung adenocarcinoma,and 54(39.42%)had non-inva-sive lung adenocarcinoma,the lung nodule diameter was(11.15±3.32)mm,the mean CT value was(-565.52±95.67)Hu,and the proportion of solid components was(55.42±14.27)%.Compared with non-invasive lung cancer patients,the lung nodule length,average CT value and the proportion of solid components in invasive lung cancer patients were significantly increased,with statistical significance(P<0.05).Compared with PARP9 group with low expression,the proportion of infiltrative,larger lung nod-ule length,larger average CT value and larger solid component in PARP9 group with high expression was significantly increased,with statistical significance(P<0.05).The recurrence free survival(RFS)rate of all patients was 77.37%.The results of univariate analysis showed that the prognosis of the patients was related to lymph node metastasis,pathological stage,average CT value and proportion of solid compo-nents,and the differences were statistically significant(P<0.05).Multivariate Cox regression analysis showed that pathological classification and average CT value were independent prognostic factors affecting RFS in patients(P<0.05).Conclusion Artificial intelligenc

关 键 词：人工智能 CT定量参数 磨玻璃结节 肺腺癌 预后 COX回归 

分 类 号：R734[医药卫生—肿瘤]