基于知识图谱的腹部前后立位片的智能质量控制

打开文本图片集

[Abstract]Objective：Todesignanintellgentimagequalitycontrolalgorithmmodelforadominalanteroposterioradiographbased onknowledgegraphforclinicalpractice.Methods：UtilizingtheAnhuiProvincialImagingCloudPlatform，4499de-identified abdominalradiographimageswereincluded.Allimagesweredividedintoatrainingset（3136images），avalidationset （457iages）， andatestset（906images）inaratioof7：1：2.Aknowledgegraph forsingle-personandmulti-personqualitycontrollabeling wasconstructed，anartifcialintellgence（AI）qualitycontrolmodelwastrained.Takingtheresultsofthreequalitycontrol expertsasthereferencestandardforqualitycontrol，theimagetechnicalqualityaccuracyandimageclarityacuracyamong thesingle-person，multi-personandAIqualitycontrolmethodswerecompared.Thecorelationsbetweensingle-personquality controlmulti-personqualitycontrolandreferencestandardswereevaluatedResults：Inthetestset，comparedwiththe referencestandard，theaverageacuracyofAIqualitycontrolwashigherthanthatofsingle-personqualitycontrol（A，B）（ P< 0.01）.IntermsofimageclaritythecorelationofIqualitycontrolwasthehighest（O.754），followedbymulti-personquality control（0.671），single-personqualitycontrolA（0.625）andB（O.617）insequence.Conclusions：Inthequalityassessment ofimageprojectiontechniqueandimageclarity，theefectofAIqualitycontrolissignificantlybetterthanthatof single-personqualitycontrol，andevenforsomelabels，itisslightlybeterthanthatofmuli-personqualityontrol，owing potential for clinical application.

[Keywords] Abdominal Radiograph;Image quality control;Knowledge graph;Artificial intellgence

腹部前后立位片（腹部平片）作为一种低成本、易获取且非侵入性的常规检查手段1，对腹部疾病的诊断至关重要。（剩余6254字）