Computed tomography-based radiomics model for predicting station 4 lymph node metastasis in non-small cell lung
Yanru Kang1, Mei Li1, Xizi Xing1
1School of Clinical and Basic Medicine, Shandong First Medical University, Jinan, 250117, China.
BMC Medical Imaging
|June 4, 2025
Summary
Machine learning models accurately identify non-small cell lung cancer (NSCLC) mediastinal lymph node metastasis. This approach aids clinical decisions for pN0-pN2 stage patients, optimizing treatment and prognosis.
Area of Science:
- Radiology and Medical Imaging
- Artificial Intelligence in Oncology
- Thoracic Surgery
Background:
- Accurate staging of mediastinal lymph node metastasis is crucial for non-small cell lung cancer (NSCLC) treatment planning.
- Preoperative identification of metastasis to station 4 mediastinal lymph nodes (MLNM) in pN0-pN2 NSCLC remains challenging.
- Enhanced precision in identifying MLNM can significantly impact clinical decision-making and patient outcomes.
Purpose of the Study:
- To develop and validate machine learning (ML) models for the preoperative identification of MLNM in NSCLC patients at pN0-pN2 stage.
- To integrate radiomics features from computed tomography (CT) images with clinical predictors for improved diagnostic accuracy.
- To establish a robust auxiliary tool for optimizing treatment strategies and prognostic assessment.
Main Methods:
- A cohort of 356 NSCLC patients (pN0-pN2) was divided into training, internal test, and independent test sets.
- Radiomics features were extracted from station 4 mediastinal lymph node regions of interest on CT images.
- Least Absolute Shrinkage and Selection Operator (LASSO) regression selected key features, and Random Forest (RF) was used to build combined radiomics-clinical models, evaluated using ROC analysis, calibration curves, and DCA.
Main Results:
- Eight radiomics features were selected from 1721 initial features using LASSO regression.
- The RF-based combined model demonstrated strong discriminative power with an Area Under the Curve (AUC) of 0.934 (training) and 0.889 (internal test).
- The combined model showed superior performance, validated for robustness on an independent test set.
Conclusions:
- A combined model integrating radiomics and clinical features using RF effectively identifies station 4 MLNM in NSCLC (pN0-pN2 stage).
- This non-invasive model serves as a valuable auxiliary tool for clinical decision-making.
- The model has the potential to optimize treatment strategies and improve prognostic assessment for NSCLC patients.
