Menna Allah Mahmoud, Yanhua Wen, Yuling Liufu, Xiaohuan Pan, Ruihua Su, Yubao Guan
{"title":"A Lightweight Dual-Output Vision Transformer for Enhanced Lung Nodule Classification Using CT Images.","authors":"Menna Allah Mahmoud, Yanhua Wen, Yuling Liufu, Xiaohuan Pan, Ruihua Su, Yubao Guan","doi":"10.1177/15330338251370439","DOIUrl":null,"url":null,"abstract":"<p><p>IntroductionThis study evaluates the effectiveness of a lightweight vision transformer (EfficientFormerV2-S2) with a dual-output architecture for lung nodule classification, assessing its performance and generalizability across multiple datasets.MethodsThe study utilized datasets from three sources: Institution 1 (936 images), Institution 2 (280 images), and a public Zenodo dataset (308 images), comprising adenocarcinoma, squamous cell carcinoma, and benign lesions. Model evaluation included holdout validation, five-fold cross-validation, and benchmarking against the PneumoniaMedMNIST dataset. Comprehensive image preprocessing and augmentation techniques were implemented.ResultsThe model demonstrated robust performance across all datasets, achieving test accuracies of 92.62 ± 1.65%, 97.14 ± 1.78%, and 95.74 ± 1.35% for Institutions 1, 2, and Zenodo respectively. Cross-validation results showed consistent performance with minimal variability (standard deviations <2%). On the PneumoniaMedMNIST benchmark, our optimized model achieved superior performance (accuracy: 0.936, AUC: 0.981) compared to ResNet18 and ResNet50 benchmarks.ConclusionThe lightweight transformer-based model demonstrates excellent performance and generalizability across multiple institutional datasets, suggesting its potential for efficient clinical implementation in lung nodule classification tasks.</p>","PeriodicalId":22203,"journal":{"name":"Technology in Cancer Research & Treatment","volume":"24 ","pages":"15330338251370439"},"PeriodicalIF":2.8000,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12374121/pdf/","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Technology in Cancer Research & Treatment","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1177/15330338251370439","RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"2025/8/21 0:00:00","PubModel":"Epub","JCR":"Q3","JCRName":"ONCOLOGY","Score":null,"Total":0}
引用次数: 0
Abstract
IntroductionThis study evaluates the effectiveness of a lightweight vision transformer (EfficientFormerV2-S2) with a dual-output architecture for lung nodule classification, assessing its performance and generalizability across multiple datasets.MethodsThe study utilized datasets from three sources: Institution 1 (936 images), Institution 2 (280 images), and a public Zenodo dataset (308 images), comprising adenocarcinoma, squamous cell carcinoma, and benign lesions. Model evaluation included holdout validation, five-fold cross-validation, and benchmarking against the PneumoniaMedMNIST dataset. Comprehensive image preprocessing and augmentation techniques were implemented.ResultsThe model demonstrated robust performance across all datasets, achieving test accuracies of 92.62 ± 1.65%, 97.14 ± 1.78%, and 95.74 ± 1.35% for Institutions 1, 2, and Zenodo respectively. Cross-validation results showed consistent performance with minimal variability (standard deviations <2%). On the PneumoniaMedMNIST benchmark, our optimized model achieved superior performance (accuracy: 0.936, AUC: 0.981) compared to ResNet18 and ResNet50 benchmarks.ConclusionThe lightweight transformer-based model demonstrates excellent performance and generalizability across multiple institutional datasets, suggesting its potential for efficient clinical implementation in lung nodule classification tasks.
期刊介绍:
Technology in Cancer Research & Treatment (TCRT) is a JCR-ranked, broad-spectrum, open access, peer-reviewed publication whose aim is to provide researchers and clinicians with a platform to share and discuss developments in the prevention, diagnosis, treatment, and monitoring of cancer.