基于AI定量检测pGGN体积与质量并预测其生长规律

Quantitative Detection of the Volume and Mass of pGGN and Predictedits Growth Law Based on AI

目的 利用人工智能(AI)肺结节体积与质量的三维定量检测功能评估持续存在肺纯磨玻璃结节(pGGN)生长速率和探索其自然生长规律。方法 回顾性搜集患者78例,共计141个pGGN;共有510次CT扫描(基线78次、随访432次)纳入本研究。使用基于卷积神经网络“Infervision软件”对基线和所有随访CT扫描的pGGN进行定量检测,包括体积、质量、体积倍增时间(VDT)和质量倍增时间(MDT)。分别利用Kaplan-Meier生存曲线、Wilcoxon秩和检验及Mann-Whitney U检验对比分析不同组别间pGGN生长曲线、VDT和MDT的差异等指标。结果 所有pGGN的平均随访时间为(55.7±30.8)个月。其中生长组63个pGGN、稳定组78个pGGN。随访2、3、5年pGGN的累积生长率分别为23.8%、46%、79.4%。Kaplan-Meier生存分析曲线显示,约50个月时,初始体积≥523.6 mm~3组及初始质量≥100 mg组pGGN增长率明显高于初始体积<523.6 mm~3组和初始质量<100 mg组。生长组pGGN的中位VDT和MDT(分别为809.67天、784.76天)均明显低于稳定组pGGN(分别为1890.10天、 1601.11天)。生长组pGGN的中位MDT低于VDT;稳定组pGGN的中位MDT也低于VDT。结论 持续存在的pGGN呈惰性生长过程,AI定量检测有助于评估预测其生长规律;通常体积和质量较大的pGGN更可能出现增长,且质量指标(MDT)变化预测价值优于体积指标(VDT)变化。

Objective To evaluate the growth rate of persistent pure ground glass nodule(pGGN)and explore its natural growth rules,with the three-dimensional quantitative detection of the volume and mass of lung nodules using.Methods 141 pGGNs from 78 patients with 510 CT scans(baseline 78 times,follow-up 432 Times)were included in this retrospective study.pGGN quantitative detection was performed on initial and all follow-up CT scans using the Infervision software based on convolution neural networks.Subsequently,pGGN volume,mass,volume doubling time(VDT)and mass doubling time(MDT)were calculated automatically.Kaplan-Meier analyses with the log-rank test,Wilcoxon rank sum test,and Mann-Whitney U test were conducted to compare and analyze the differences in pGGN growth curve,VDT and MDT among different groups.Results The mean follow-up period of enrolled pGGNs was(55.7±30.8)months.All pGGNs were categorized into 63 growth group,78 stable group.The 2-year,3-year and 5-year cumulative percentages of pGGN growth were 23.8%,46%,and 79.4%,respectively.Kaplan-Meier survival analysis curve shows that at about 50 months,the pGGN growth rate of the initial volume≥523.6 mm~3 group and the initial mass≥100 mg group were significantly higher than that of the initial volume<523.6 mm~3 group and the initial mass<100 mg group.The median VDT and MDT of the pGGNs having grown were 809.67 days and 784.76 days,respectively,which were significantly lower than the stable group with 1890.10 days and 1601.11 days.The median MDT of pGGN in the growth group was lower than VDT;the median MDT of pGGN in the stable group was also lower than VDT.Conclusion Persistent pGGNs showes an indolent course.AI quantitative detection can assist in evaluating and predicting its growth law;pGGNs with larger initial volume and mass are more likely to grow,and the predictive value of the change in quality index(MDT)is better than the change in volume(VDT).