人工智能辅助下儿童胸部CT低剂量扫描参数自适应优化及多中心临床应用研究OA
Adaptive Optimization of Low-dose Chest CT Scanning Parameters for Chil-dren with Artificial Intelligence Assistance and Its Multi-center Clinical Application
目的:本研究旨在建立基于人工智能的儿童胸部计算机断层扫描(computed tomography,CT)扫描参数自适应优化策略,通过基于MLP-CNN混合架构的CT参数自适应优化系统(CT adaptive parameter optimization,CT-APO)实时优化管电流、管电压、旋转时间及准直器宽度等参数,在保证图像质量的前提下最大程度降低辐射剂量,并通过多中心临床应用验证其有效性.方法:本研究采用前瞻性多中心设计,纳入2023年1月至2024年9月期间需行胸部CT检查的0～16岁儿童患者共2 576例.患者按年龄、体重分层随机分组,传统固定参数扫描方案组(对照组,n=1283)采用基于体重的固定参数,CT-APO系统组(研究组,n=1 293)使用基于卷积神经网络的自适应管电压选择技术,对两组容积CT剂量指数(volume CT dose index,CTDIvol)、剂量长度乘积(dose-length product,DLP)、有效辐射剂量(effective radiation dose,ED)、信噪比(signal-to-noise ra-tio,SNR)、对比噪声比(contrast-to-noise ratio,CNR)及5分制主观图像质量评分进行评估.结果:研究组的平均管电压较对照组的固定参数明显降低(P＜0.001).总体上,研究组的CTDIvol、DLP及估算有效剂量较对照组降低(P＜0.001).研究组与对照组在客观图像质量参数(图像噪声、SNR、CNR)和主观图像质量评分方面均无统计学差异(P＞0.05).对于肺实质小结节(≤5 mm)、支气管壁增厚、肺间质改变等病变,两组检出率均无统计学差异(P＞0.05).结论:基于人工智能的儿童胸部CT扫描参数自适应优化系统能够在保持诊断图像质量和临床诊断效能的前提下,显著降低儿童胸部CT检查的辐射剂量,具有良好的跨平台通用性与自适应学习能力.该系统特别适用于婴幼儿和低体重患者群体,有望优化儿科CT检查实践,降低辐射风险,提高医疗安全性.
Objective:This study aims to establish an artificial intelligence-based adaptive optimization strategy for pediatric chest CT scanning parameters using an MLP-CNN hybrid architecture CT adaptive parameter optimization(CT-APO)system.The system dynamically optimizes parameters including kVp,mA,rotation time,and collimator width to maximize radiation dose reduction while maintaining image quality,with validation through multi-center clinical applications.Methods:A prospective,multi-center,randomized controlled study was conducted,enrolling 2 576 pediatric patients aged 0～16 years who underwent chest CT examinations from January 2023 to September 2024.Patients were stratified by age and weight,then randomly assigned concealment to either the control group(n=1 283)or the study group(CT-APO system,n=1 293)using convolutional neural network-based adaptive tube voltage selection technology.Primary evaluation metrics included volume CT dose index(CTDIvol),dose-length product(DLP),effective radiation dose(ED),signal-to-noise ratio(SNR),contrast-to-noise ratio(CNR),and subjective image quality scores on a 5-point scale.Results:The average tube voltage in the study group was significantly lower than that of the control group(P＜0.001).Overall,the CTDIvol,DLP,and estimated effective dose in the study group were reduced compared to the control group(P＜0.001).There were no statistically significant differences between the study group and the control group in objective image quality parameters(image noise,SNR,CNR)or subjective image quality scores(P＞0.05).For lesions such as pulmonary parenchymal nodules(≤5 mm),bronchial wall thickening,and interstitial lung changes,the detection rates in both groups showed no statistically significant differences(P＞0.05).Conclusion:The artificial intelligence-based adaptive optimization system for pediatric chest CT scanning parameters can significantly reduce radiation dose while maintaining diagnostic image quality and clinical diagnostic efficacy.The system demonstrates excellent cross-platform applicability and adaptive learning capabilities,making it particularly suitable for infants and low-weight patient populations.This technology holds promise for optimizing pediatric CT examination practices,reducing radiation risks,and improving medical safety.
冉崇荣;邓丽佳;卢晓玉;雷敏
绵阳市妇幼保健院绵阳市儿童医院,四川 621000绵阳市妇幼保健院绵阳市儿童医院,四川 621000绵阳市妇幼保健院绵阳市儿童医院,四川 621000绵阳市妇幼保健院绵阳市儿童医院,四川 621000
医药卫生
儿童胸部CT人工智能辐射剂量图像质量自适应参数优化深度学习
pediatric chest CTartificial intelligenceradiation doseimage qualityadaptive parameter optimizationdeep learning
《影像科学与光化学》 2026 (4)
144-153,10
四川省自然科学基金项目(2025ZNSFSC1772)绵阳市妇幼保健院.绵阳市儿童医院2024年院级科研项目(2024-KY-012).
评论