A deep learning-based radiomics model for predicting lymph node status from lung adenocarcinoma.

Area of Science:

• Radiology and Medical Imaging
• Oncology
• Computational Pathology

Background:

• Accurate evaluation of lymph node metastasis is crucial for lung adenocarcinoma staging and treatment.
• Current methods for assessing lymph node status have limitations, necessitating safer and more precise predictive tools.

Purpose of the Study:

• To develop and validate a radiomics-based noninvasive classification model for predicting lymph node metastasis in lung adenocarcinoma.
• To assess the efficacy of the extreme gradient boosting (XGBoost) method in distinguishing lymph node metastasis status using contrast-enhanced CT features.

Main Methods:

• A retrospective analysis of 503 lung adenocarcinoma patients from two hospitals was performed.
• Radiomics features were extracted from contrast-enhanced CT images using traditional and deep learning methods, followed by feature selection using Spearman test and LASSO.
• A classification model was built using machine learning algorithms, with performance evaluated by Accuracy, AUC, Specificity, Precision, Recall, and F1 scores.

Main Results:

• The XGBoost model demonstrated strong performance on the external test set with Accuracy of 0.765, AUC of 0.845, Specificity of 0.705, Precision of 0.784, Recall of 0.811, and F1 score of 0.797.
• Decision curve analysis, calibration curves, and confusion matrix confirmed the model's stability and accuracy in predicting lymph node metastasis.

Conclusions:

• A noninvasive radiomics classification model based on XGBoost, utilizing enhanced CT images, accurately predicts lymph node metastasis in lung adenocarcinoma.
• This method offers a safe and precise alternative to invasive diagnostic procedures, potentially improving patient management and clinical outcomes.