Development and validation of an interpretable clinical-radiomics model for predicting 3-year postoperative recurrence in clinical stage I lung adenocarcinoma: a multicentre study.
- 1Department of Radiology, The Affiliated Hospital of Qingdao University, Qingdao, China.
- 2Department of Ultrasound, The Fourth Affiliated Hospital of Zhejiang University School of Medicine, Yiwu, China.
Abstract
Although surgical resection is the standard curative treatment for clinical stage I lung adenocarcinoma (LUAD), a clinically meaningful proportion of patients still experience postoperative recurrence. Conventional clinical and computed tomography (CT) imaging features have limited ability to provide individualized preoperative risk stratification. This study aimed to develop and externally validate an interpretable CT-based clinical-radiomics model for predicting recurrence within 3 years after surgery in patients with clinical stage I LUAD. This multicentre retrospective prediction model development and validation study included 333 patients with pathologically confirmed clinical stage I LUAD who underwent surgical resection between January 2020 and December 2021. Patients from Center 1 (The Affiliated Hospital of Qingdao University) (n=199) constituted the training cohort, and those from Center 2 (The Fourth Affiliated Hospital of Zhejiang University School of Medicine) (n=134) served as the validation cohort. The outcome was recurrence within 3 years after surgery (determined by imaging findings, clinical follow-up records, and pathological confirmation when available). Patients without recurrence were required to have at least 3 years of follow-up. Radiomics features were extracted from preoperative CT images and selected using analysis of variance (ANOVA) and least absolute shrinkage and selection operator (LASSO) regression, while clinical predictors were identified using univariable and multivariable logistic regression. Radiomics, clinical, and combined models were evaluated using the area under the receiver operating characteristic (ROC) curve, calibration, and decision curve analysis (DCA). SHapley Additive exPlanations (SHAP) analysis was used to interpret the optimal model. Overall, 63 of 333 patients (18.9%) experienced recurrence within 3 years after surgery. In the validation cohort, the clinical, radiomics, and combined models achieved the areas under the curve (AUCs) of 0.674 [95% confidence interval (CI): 0.550-0.799], 0.781 (95% CI: 0.668-0.893), and 0.850 (95% CI: 0.752-0.948), respectively, with corresponding sensitivities/specificities of 0.880/0.385, 0.920/0.532, and 0.840/0.798. The combined model showed significantly better discrimination than the clinical model (DeLong P=0.02). The combined model showed acceptable calibration in the validation cohort, with the calibration intercept, calibration slope, and Brier score were -0.157, 0.763, and 0.114, respectively. DCA suggested a potential net benefit of the combined model across a range of threshold probabilities in both cohorts, while SHAP analysis provided individualized interpretation of model predictions. The combined clinical-radiomics model based on preoperative CT and clinical characteristics showed potential for stratifying the risk of recurrence within 3 years after surgery in patients with clinical stage I LUAD. Further validation in larger, prospective cohorts is warranted before clinical implementation.
Related
- Development and validation of an interpretable machine learning model for early prediction of mortality in ICU patients with chronic obstructive pulmonary disease and heart failure
- Development and validation of an interpretable prediction model for high-grade vesicoureteral reflux in children with urinary tract infections
- Development and validation of a postoperative prognostic nomogram for early-stage pancreatic cancer
- Development and internal validation of a machine learning prediction model for postoperative hemorrhage following modified radical mastectomy for breast cancer
- Development and Validation of a Perioperative Risk Update Model for Postoperative Delirium in Patients Aged ≥ 80 Years Undergoing Hip Fracture Surgery: A Retrospective Cohort Study
BibTeX
@article{su2026development,
title = {{Development and validation of an interpretable clinical-radiomics model for predicting 3-year postoperative recurrence in clinical stage I lung adenocarcinoma: a multicentre study.}},
author = {Qiushi Su and Bingyan Wang and Pei Nie and Wenjian Xu},
journal = {Journal of thoracic disease},
year = {2026},
month = sep,
doi = {10.21037/jtd-2026-1793},
url = {https://europepmc.org/article/MED/42996697},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - Development and validation of an interpretable clinical-radiomics model for predicting 3-year postoperative recurrence in clinical stage I lung adenocarcinoma: a multicentre study.
AU - Qiushi Su
AU - Bingyan Wang
AU - Pei Nie
AU - Wenjian Xu
PY - 2026
DA - 2026/09/28
JO - Journal of thoracic disease
DO - 10.21037/jtd-2026-1793
UR - https://europepmc.org/article/MED/42996697
AB - Although surgical resection is the standard curative treatment for clinical stage I lung adenocarcinoma (LUAD), a clinically meaningful proportion of patients still experience postoperative recurrence. Conventional clinical and computed tomography (CT) imaging features have limited ability to provide individualized preoperative risk stratification. This study aimed to develop and externally validate an interpretable CT-based clinical-radiomics model for predicting recurrence within 3 years after surgery in patients with clinical stage I LUAD. This multicentre retrospective prediction model development and validation study included 333 patients with pathologically confirmed clinical stage I LUAD who underwent surgical resection between January 2020 and December 2021. Patients from Center 1 (The Affiliated Hospital of Qingdao University) (n=199) constituted the training cohort, and those from Center 2 (The Fourth Affiliated Hospital of Zhejiang University School of Medicine) (n=134) served as the validation cohort. The outcome was recurrence within 3 years after surgery (determined by imaging findings, clinical follow-up records, and pathological confirmation when available). Patients without recurrence were required to have at least 3 years of follow-up. Radiomics features were extracted from preoperative CT images and selected using analysis of variance (ANOVA) and least absolute shrinkage and selection operator (LASSO) regression, while clinical predictors were identified using univariable and multivariable logistic regression. Radiomics, clinical, and combined models were evaluated using the area under the receiver operating characteristic (ROC) curve, calibration, and decision curve analysis (DCA). SHapley Additive exPlanations (SHAP) analysis was used to interpret the optimal model. Overall, 63 of 333 patients (18.9%) experienced recurrence within 3 years after surgery. In the validation cohort, the clinical, radiomics, and combined models achieved the areas under the curve (AUCs) of 0.674 [95% confidence interval (CI): 0.550-0.799], 0.781 (95% CI: 0.668-0.893), and 0.850 (95% CI: 0.752-0.948), respectively, with corresponding sensitivities/specificities of 0.880/0.385, 0.920/0.532, and 0.840/0.798. The combined model showed significantly better discrimination than the clinical model (DeLong P=0.02). The combined model showed acceptable calibration in the validation cohort, with the calibration intercept, calibration slope, and Brier score were -0.157, 0.763, and 0.114, respectively. DCA suggested a potential net benefit of the combined model across a range of threshold probabilities in both cohorts, while SHAP analysis provided individualized interpretation of model predictions. The combined clinical-radiomics model based on preoperative CT and clinical characteristics showed potential for stratifying the risk of recurrence within 3 years after surgery in patients with clinical stage I LUAD. Further validation in larger, prospective cohorts is warranted before clinical implementation.
SN - 2072-1439
SN - 2077-6624
ER -