2026-10-07T15:04:19Z https://api.figshare.com/v2/oai

oai:figshare.com:article/34073485 2026-10-05T17:48:33Z category_146 category_7 category_12 category_45 category_734 category_106 portal_5 item_type_3 month_year_10_2026

<p>Architectural ablation study.</p> Muhammad Ahsan Jamil (25317220) Malik Daler Ali Awan (25317223) Fatima Bukhari (25317226) Malik Muhammad Saad Missen (25317229) Nadeem Iqbal Kajla (25317232) Space Science Medicine Cell Biology Sociology Biological Sciences not elsewhere classified Science Policy paired errors differed guided efficientnet framework balanced working corpus underlying histopathology images released lc25000 images augmentation source identifiers mean accuracy ), level test partition level benchmark performance image level could div >< p study presents effatt lungnet reached 99 earliest source 650 images 500 images test partitions smaller collection refine features p </ model combines mcnemar ’ lung adenocarcinoma generated 10 f1 score experiments used efficientnetb3 backbone clinical claim centre datasets 9997 auc 001 ). <div><p>This study presents EffAtt-LungNet, an attention-guided deep learning framework for three-class lung histopathology classification. The model combines an EfficientNetB3 backbone with the Convolutional Block Attention Module (CBAM) to refine features and improve discrimination among benign lung tissue, lung adenocarcinoma, and lung squamous cell carcinoma. Experiments used the publicly released LC25000 lung subset, which contains 15,000 images across three classes and was itself produced by augmenting a smaller collection of underlying histopathology images. We generated 10,500 additional fixed variants, creating a balanced working corpus of 25,500 images with training, validation, and test partitions of 15,300, 2,550, and 7,650 images. EffAtt-LungNet reached 99.86% accuracy, 99.86% precision, 99.86% recall, 99.86% F1 score, and 0.9997 AUC on the image-level test partition. Performance was stable across five random seeds (mean accuracy ), and paired errors differed from those of the EfficientNetB3 backbone (McNemar’s test, , <i>p</i> < 0.001). Because pre-augmentation source identifiers for the released LC25000 images were unavailable, independence at the earliest source-image level could not be confirmed. The results therefore represent internal tile-level benchmark performance and require validation on independent, non-augmented, multi-centre datasets before any translational or clinical claim can be made.</p></div> 2026-10-05T17:48:13Z Dataset Dataset 10.1371/journal.pone.0359302.t011 https://figshare.com/articles/dataset/_p_Architectural_ablation_study_p_/34073485 CC BY 4.0