融合ESCA注意力机制的迁移学习在X线肺炎检测中的研究

打开文本图片集

中图分类号：TP181 文献标识码：A文章编号：1006-8228（2025）08-46-06

Abstract：Thispaperproposesatransferlearningframeworkbasedonchanelspatialatentionmechanismsfordetecting pneumoniainchestX-rayimages.Theframeworkintegratesthefeatureextractioncapabilitiesofthreepre-trainedmodelsResNet101，ResNet50ndReNet34andicorporatesanicientcaelspatialaetionmechansmECA）toddressise thattraditionalmethodsfocusonlyonchannelinformationwhileneglectingspatialinformation.Byleveragingthechanelspatial attentionmechanism，themodelcansimultaneouslyfocusonbothchannelandspatialinformation，therebysignificantlyiproving clasificationaccuracyExperimentalresultsdemonstratethatthisapproachachievesexcelentperformanceintermsofaccuracy （2 （98%） ，precision （97.2%） ，recall （99%） ，and specificity （95.3%） .This study provides strong support for pneumonia detection and offersanew efficient solution for medical image classification.

eywords：Channel Spatial Attention Mechanism；Transfer Learning；Pneumonia Detection;Accuracy

0引言

2019冠状病毒病（Coronavirus Disease 2019，COVID-19）是一种严重的疾病，全球公共卫生部门目前正试图通过早期发现和有效的防控手段，来遏制这种病毒的传播l。（剩余9668字）