基于深度算法的肝外胆管癌术前MRI动态增强扫描自动分期系统的构建及验证

扫码查看

原文链接

万方数据
维普

中文摘要：目的基于深度学习算法构建肝外胆管癌(ECC)术前磁共振成像(MRI)动态增强扫描自动分期系统,并验证其效能.方法回顾性分析 2020 年 1 月—2022 年 12 月本院收治的 270 例ECC患者的临床资料,作为训练集,收集其经MRI动态增强扫描获得的肿瘤位置、侵犯范围、淋巴结转移等特征,基于深度学习算法对不同临床分期的ECC进行分类建模,构建ECC术前MRI动态增强扫描自动分期系统.另回顾性分析 2023 年 1-12 月本院收治的 94 例ECC患者临床资料作为验证集,以病理学诊断的临床分期结果作为金标准,分析基于深度算法的ECC术前MRI动态增强扫描自动分期系统对患者临床分期的诊断效能.结果基于深度学习算法的ECC术前MRI动态增强扫描自动分期系统对临床分期Ⅰ期、Ⅱ期、Ⅲ期、Ⅳ期的诊断灵敏度(100.00%、93.33%、97.78%、100.00%)、特异度(100.00%、98.44%、95.92%、100.00%)和准确度(100%、96.81%、96.81%、100.00%)高于MRI动态增强扫描检查方式(灵敏度:91.67%、86.67%、88.89%、85.71%,特异度:97.56%、90.63%、93.88%、100.00%,准确度:96.81%、89.36%、91.49%、98.94%),且与病理学结果的一致性高(Kappa=0.885,P<0.001),MRI动态增强扫描检查方式与病理学结果也存在一致性(Kappa=0.691,P<0.001).结论基于深度学习算法构建的ECC术前MRI动态增强扫描自动分期系统可提高对ECC术前分期的诊断敏感度、特异度和准确度.

外文标题：Construction and verification of automatic staging system of preoperative dynamic enhanced MRI based on depth algorithm for extrahepatic cholangiocarcinoma

外文摘要：Objective Based on the deep learning algorithm,an automatic staging system for preoperative dynamic enhanced magnetic resonance imaging(MRI)for extrahepatic cholangiocarcinoma(ECC)was constructed,and its effectiveness was verified.Methods The clinical data of 270 patients with ECC admitted to our hospital from January 2020 to December 2022 were retrospectively analyzed as a training set.Characteristics including tumor location,invasion range,and lymph node metastasis obtained from dynamic enhanced MRI were collected.The deep learning algorithm was used to classify and model different clinical stages of ECC,constructing an automatic staging system for preoperative dynamic enhanced MRI for ECC.Additionally,the clinical data of 94 patients with ECC admitted from January 2023 to December 2023 were retrospectively analyzed as a verification set.The clinical staging results from pathological diagnosis were used as the gold standard to analyze the diagnostic efficiency of the automatic staging system based on the deep learning algorithm for preoperative dynamic enhanced MRI in patients with ECC.Results The diagnostic sensitivity(100.00%,93.33%,97.78%,100.00%),specificity(100.00%,98.44%,95.92%,100.00%),and accuracy(100.00%,96.81%,96.81%,100.00%)of the automatic staging system based on the deep learning algorithm for preoperative dynamic enhanced MRI for clinical staging Ⅰ,Ⅱ,Ⅲ,and Ⅳ were higher than those of the dynamic enhanced MRI method(sensitivity:91.67%,86.67%,88.89%,85.71%,specificity:97.56%,90.63%,93.88%,100.00%,accuracy:96.81%,89.36%,91.49%,98.94%),and it showed high consistency with pathological results(Kappa=0.885,P<0.001).The dynamic enhanced MRI method also demonstrated consistency with pathological results(Kappa=0.691,P<0.001).Conclusions The automatic staging system for preoperative dynamic enhanced MRI based on the deep learning algorithm can improve the diagnostic sensitivity,specificity,and accuracy of preoperative staging for ECC.

外文关键词：

Deep learning algorithmExtrahepatic cholangiocarcinomaMagnetic resonance imagingDynamic enhanced scanning

作者：

时金凤、王志芳

展开 >

作者单位：

457001 河南濮阳,濮阳油田总医院放射科

关键词：

深度学习算法肝外胆管癌磁共振成像动态增强扫描

出版年：

2025

DOI：