摘要
目的研究多模态影像模型在子宫内膜癌诊断中的应用价值。方法将2019年7月—2020年6月接受诊断的疑似子宫内膜病变患者250例纳入研究。包括子宫内膜癌组150例,良性增生组100例。分别对所有受试者进行经阴道彩色多普勒超声和磁共振成像(MRI)检查,采集影像资料。通过医学影像智能软件,将不同模态的医学图像,借助特定的算法,通过一定的处理,从而获取不同模态图像优点或互补性生成新的图像,并对可融合图像实施数据度量以及统计分析,建立多模态影像模型。比较两组微血管密度(MVD)、阻力指数(RI)以及搏动指数(PI)水平。此外,以病理诊断为金标准,分析不同诊断方式的诊断效能。结果子宫内膜癌组MVD高于良性增生组,而RI以及PI均低于良性增生组(均P<0.05)。拉普拉斯金字塔融合算法和GAN算法多模态影像模型诊断子宫内膜癌的灵敏度、特异度以及准确度均高于阴道超声以及MRI诊断(均P<0.05);其中GAN算法的灵敏度、特异度以及准确度高于拉普拉斯金字塔融合算法,但两组比较差异不明显(均P>0.05)。GAN算法多模态影像模型诊断Ⅰ期、Ⅱ期子宫内膜癌的符合率均高于阴道超声以及MRI检查,且拉普拉斯金字塔融合算法多模态影像模型诊断Ⅰ期、Ⅱ期子宫内膜癌的符合率均高于阴道超声检查(均P<0.05);拉普拉斯金字塔融合算法多模态影像模型与GAN算法多模态影像模型诊断Ⅰ期、Ⅱ期子宫内膜癌的符合率对比,差异不明显(均P>0.05)。结论多模态影像模型应用于子宫内膜癌诊断中的价值较高,可明显提高诊断灵敏度、特异度以及准确度,且GAN算法的应用效果更为理想,值得临床推广应用。
Objective To study the application value of multi-mode image model in the diagnosis of endometrial cancer.Methods 250 patients with suspected endometrial lesions diagnosed in our hospital from July 2019 to June 2020 were included in the study.Including endometrial carcinoma group 150 cases,benign hyperplasia group 100 cases.Transvaginal color Doppler ultrasound,computed tomography(CT),and magnetic resonance imaging(MRI)were performed on all subjects.Through intelligent medical image software,the advantages or complementarities of medical images with different modes can be obtained through certain processing with the help of specific algorithms,and new images can be generated.Moreover,data measurement and statistical analysis can be carried out on the fused images to establish a multi-mode image model.Microvascular density(MVD),resistance index(RI)and pulse index(PI)were compared between the two groups.In addition,pathological diagnosis was used as the gold standard to analyze the diagnostic efficacy of different diagnostic methods.Results MVD in the endometrial cancer group was higher than that in the benign hyperplasia group,while RI and PI were lower than that in the benign hyperplasia group(all P<0.05).The sensitivity,specificity and accuracy of Laplacian pyramid fusion algorithm and GAN algorithm multi-mode image model in the diagnosis of endometrial cancer were all higher than that of ultrasonic vaginal ultrasonography and MRI(all P<0.05).The sensitivity,specificity and accuracy of GAN algorithm were higher than that of Laplace pyramid fusion algorithm,but there was no significant difference between the two groups(all P>0.05).GAN multimodal imaging model diagnosis stageⅠ,Ⅱthe coincidence rate of endometrial carcinoma were higher than ultrasound vaginal ultrasound and MRI,and Laplacian pyramid fusion algorithm for multimodal imaging model diagnosis stageⅠ,Ⅱthe coincidence rate of endometrial carcinoma were higher than ultrasound vaginal ultrasound(all P<0.05);Laplacian pyramid fusion algorithm for multimodal imaging model and the algorithm of GAN multimodal imaging model diagnosis stageⅠ,Ⅱthe coincidence rate of endometrial carcinoma in contrast,no significant difference(P>0.05).Conclusion The application of multi-mode image model in the diagnosis of endometrial cancer has a high value,which can significantly improve the diagnostic sensitivity,specificity and accuracy.Moreover,GAN algorithm has a more ideal application effect,which is worthy of clinical application.
作者
王玲
张玉娟
肖云敏
陈菁特
刘新琼
WANG Ling;ZHANG Yujuan;XIAO Yunmin;CHEN Jingte;LIU Xinqiong(Shenzhen People’s Hospital,Shenzhen 518000,China;不详)
出处
《现代医院》
2021年第3期474-477,共4页
Modern Hospitals
基金
广东省医学科学技术研究基金项目(B2019014)。
关键词
子宫内膜癌
多模态影像模型
应用价值
微血管密度
阻力指数
Endometrial cancer
Multimodal image model
Application value
Microvascular density
Resistance index
