A CT-based deep learning approach to differentiate multiple primary lung cancers, metastases, and benign nodules - Research Open Access

Sep 22, 2026

Upload by Patrick

Lung cancer diagnosis faces persistent difficulty when multiple lesions coexist, because imaging findings on chest CT must be separated into multiple primary lung cancers (MPLCs), intrapulmonary metastases (IPMs), and multiple benign pulmonary lesions (MBPLs). This study develops and validates an automated deep learning classification system using CT data, comparing six pretrained architectures via ablation experiments and aggregated DeLong tests across seeds. The final MambaOut-Kobe model shows high macro-AUC and accuracy, supported by decision curve analysis and Grad-CAM interpretability.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages14

Reading Time35 min

AccessFree to download

Abstract

Background and Objectives

Materials and Methods

Model comparison and validation

Clinical utility and interpretability

Results

Conclusions

Keywords

Introduction

Liufu etal. BMC Cancer (2026) 26:177 BMC Cancer [https://doi.org/10.1186/s12885-025-15501-1](https://doi.org/10.1186/s12885-025-15501-1) RESEARCH Open Access A CT-based deep learning approach to differentiate multiple primary lung cancers, metastases, and benign nodules Yuling Liufu 1†, Ruihua Su1†, Yanhua Wen1, Yubao Guan1* and Menna Allah Mahmoud1*† Abstract Background Lung cancer, particularly adenocarcinoma and squamous cell carcinoma, remains a leading cause of cancer-related deaths globally. The diagnosis of multiple primary lung cancers (MPLCs) has become more frequent due to advanced chest CT technology and improved health surveillance. However, differentiating MPLCs from intrapulmonary metastases (IPMs) and multiple benign pulmonary lesions (MBPLs) remains challenging. Objectives Distinguishing multiple primary lung cancers from metastases and benign lesions on CT remains challenging yet critical for treatment planning. Current approaches rely on subjective interpretation and invasive procedures. This study aims to develop and validate an automated deep learning classification system to provide rapid, objective diagnoses for optimizing patient management. Materials and methods We studied 260 patients (MPLC = 83, IPM = 81, MBPL = 96; 881 axial CT slices) . Six pretrained architectures (DenseNet−121, EfficientNet-B1, MambaOut-Kobe, ResNet−50, SwinV2-CR-Tiny−224, ViT-TinyPatch16−224) were compared in a five-seed ablation (seeds 42, 789, 1011, 2025, 2048) . Pairwise one-vs-rest DeLong tests were aggregated across seeds to compare AUCs. Clinical utility was assessed using decision curve analysis (DCA) . The final model (MambaOut-Kobe) underwent stratified five-fold cross-validation. Results Considering efficiency, MambaOut-Kobe combined competitive accuracy with the lowest memory (~ 100 ± 14 MB) and low latency (~ 0.0093 ± 0.0017 s/image) . Aggregated DeLong testing found no significant AUC differences among these models after multiplicity control. On five-fold cross-validation, MambaOut-Kobe achieved a macro-AUC of 0.946 ± 0.004 (95% CI 0.942–0. 950), and an accuracy 0.829 ± 0.029 (95% CI 0.800–0. 858) . DCA demonstrated a positive net benefit across clinically relevant threshold probabilities compared with treat-all and treatnone strategies. Grad-CAM visualizations highlighted diagnostically relevant regions in CT images, providing interpretable decisionmaking support. †Yuling Liufu, Ruihua Su and Menna Allah Mahmoud contributed equally as first co-authors. *Correspondence: Yubao Guan [yubaoguan@163.com](yubaoguan@163.com)[ ](yubaoguan@163.com)Menna Allah Mahmoud [menna22a@yahoo.com](menna22a@yahoo.com) Full list of author information is available at the end of the article © The Author(s) 2026. Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit [http://creati](http://creati)[vecommons.org/licenses/by-nc-nd/4.0/](vecommons.org/licenses/by-nc-nd/4.0/.)[.](vecommons.org/licenses/by-nc-nd/4.0/.) Liufu et al. BMC Cancer (2026) 26:177 Page 2 of 14 Conclusions The MambaOut Kobe model demonst

FAQs

Why is differentiating MPLCs from IPMs and MBPLs on CT challenging?

Diagnosing multiple primary lung cancers versus metastases and benign lesions remains difficult because current practice often relies on subjective interpretation and invasive procedures, while CT presentations can overlap.

Which deep learning architectures were compared in the study?

The study compared six pretrained architectures: DenseNet-121, EfficientNet-B1, MambaOut-Kobe, ResNet-50, SwinV2-CR-Tiny-224, and ViT-TinyPatch16-224.

What performance did the final MambaOut-Kobe model achieve?

On five-fold cross-validation, MambaOut-Kobe achieved a macro-AUC of 0.946 ± 0.004 and an accuracy of 0.829 ± 0.029, with decision curve analysis showing positive net benefit across relevant thresholds.

Sep 22, 2026

Upload by Patrick