Differentiating sub-centimeter lung metastases in colorectal cancer by deep learning - a multicenter retrospective study

Sep 23, 2026

Uploaded by dhado

Early, noninvasive identification of sub-centimeter lung metastases is crucial for timely decisions and improved outcomes in colorectal cancer. Indeterminate small lung nodules remain a major diagnostic challenge. This study develops and validates a deep learning model to differentiate benign from metastatic sub-centimeter lung nodules using pretreatment non-contrast CT, supporting accurate classification to guide individualized management.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Abstract

Background

Methods

Results

Conclusion

Background

Gao et al. BMC Medical Imaging (2026) 26:424 BMC Medical Imaging [https://doi.org/10.1186/s12880-026-02508-8](https://doi.org/10.1186/s12880-026-02508-8) RESEARCH Open Access Differentiating sub-centimeter lung metastases in colorectal cancer by deep learning: a multicenter retrospective study Xinyi Gao1†, Xiao Ma2†, Zhiyuan Zhang3,4,5, Jie Yuan6, Qianling Li1, Pengwen Zheng7, Min Lv8, Dening Ma9,10*† and Jingjing Sun1*† Abstract Background Early diagnosis of sub-centimeter lung metastases is critical for timely decision-making and improved prognosis in patients with colorectal cancer. The diagnostic evaluation of indeterminate sub-centimeter lung nodules in colorectal cancer patients remains a crucial challenge. We aim to develop and validate a deep learning model for differentiating sub-centimeter lung metastases noninvasively. Methods Our retrospective study included 1335 colorectal cancer patients with pretreated sub-centimeter lung metastases from 3 centers and 1335 benign lung nodules from one center. The primary cohort comprised 1194 patients, who were randomly assigned (8:2) to training and internal validation cohorts. Two external validation cohorts (EVC) consisted of 101 (EVC1) and 40 (EVC2) patients. The deep learning framework employed a fully automated VNetbased segmentation model for non-contrast computed tomography (CT) scans, integrated with a ResNet18 classifier for discriminating whether sub-centimeter lung nodules are benign or metastatic. Moreover, stepwise validation of subgroups was performed according to the maximum diameter of the nodules (10, 9, 8, 7, 6, 5, ≤ 4 mm) . Results The automatic segmentation model achieved a dice coefficient of 0.825. In primary cohort, the accuracy by radiologists was 0.705. The AUC in deep learning model showed 0.953 (95%CI: 0 .937–0. 967), 0.906 (95%CI: 0.874– 0. 926), and 0.951 (95%CI: 0 .938–0. 965) in internal and two external validation cohorts. Furthermore, the stepwise validation demonstrated that the diagnostic performance remain stable as the nodule maximum diameter shrinks when size≥5 mm. Conclusion The deep learning model can accurately and non-invasively distinguish between benign and metastatic sub-centimeter lung nodules in patients with colorectal cancer. | †Xinyi Gao and Xiao Ma contributed equally to this work as co-first authors. | | --- | | †Dening Ma and Jingjing Sun contributed equally to this work asco-corresponding authors. | *Correspondence: Dening Ma [madeningsdu@163.com](madeningsdu@163.com)[ ](madeningsdu@163.com)Jingjing Sun [sophia-j@163.com](sophia-j@163.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/.) Gao et al. BMC Medical Imaging (2026) 26:424 Page 2 of 12 Keywords Colorectal cancer, Computed tomography, Sub-centimeter, Lung metastases, Deep learning Background

FAQs

What problem does this study address for colorectal cancer patients?

It targets the difficulty of distinguishing indeterminate sub-centimeter lung nodules as benign versus metastatic to support early treatment decisions.

How was the deep learning model built and validated?

A retrospective multicenter dataset was used, including a primary training/validation cohort and two external validation cohorts, combining automated VNet-based segmentation with a ResNet18 classifier for benign versus metastatic discrimination.

What performance metrics were reported for distinguishing metastases?

The automatic segmentation achieved a dice coefficient of 0.825, and the deep learning classifier showed AUC values around 0.906–0.953 across internal and external validation cohorts.

Sep 23, 2026

Uploaded by dhado