July 1, 2023HeliyonOpen Access

Interpretable prediction of cardiopulmonary complications after non-small cell lung cancer surgery based on machine learning and SHapley additive exPlanations

Q: What is the clinical evidence from this study?

Study design: Observational. Population: Non-small cell lung cancer (NSCLC) (n=718). Intervention: Machine learning prediction models (Random Forest). Primary outcome: Postoperative cardiopulmonary complications (AUC 0.856, 95% CI 0.815-0.898).

View Full Paper

Ask AI

Bookmark

Key result

A Random Forest machine learning model predicted postoperative cardiopulmonary complications after lobectomy for non-small cell lung cancer with an AUC of 0.856 (95% CI 0.815-0.898).

Why the study?

Approximately 20% of patients develop cardiopulmonary complications after lobectomy for lung cancer, necessitating an accurate and personalized machine learning-based method to help prevent them.

Can machine learning models accurately predict postoperative cardiopulmonary complications in patients undergoing lobectomy for NSCLC?

Population

718 patients who underwent lobectomy for NSCLC

Comparison

Five machine learning models for predicting postoperative cardiopulmonary complications

Design

Retrospective cohort study

Authors

YZYihai ZhaiFirst Affiliated Hospital of GuangXi Medical University XLXue LinBoston University QWQiaolin WeiGuangxi Medical University

Discussion

Member takes

Implication

May aid preoperative risk stratification in NSCLC lobectomy; leaves open clinical utility pending prospective validation.

Key Points

Develop and evaluate interpretable machine learning algorithms to predict postoperative cardiopulmonary complications in patients undergoing lobectomy for non-small cell lung cancer.
Retrospectively analyzed clinical records of 718 patients who underwent lobectomy for non-small cell lung cancer between 2017 and 2021.
Trained and tuned five machine learning algorithms—logistic regression, decision tree, random forest, gradient boosted decision tree, and XGBoost—using grid search and 10-fold cross-validation.
Applied SHapley Additive exPlanations (SHAP) to interpret model predictions and identify the clinical features most strongly associated with complication risk.
Postoperative cardiopulmonary complications occurred in 20.89% of patients (150 of 718).
The random forest model yielded the highest predictive performance, achieving an AUC of 0.856 (95% CI, 0.815–0.898), an accuracy of 0.786 (95% CI, 0.738–0.834), and a Kappa value of 0.696 (95% CI, 0.617–0.768).
SHAP interpretation identified patient age, gender, and intraoperative blood loss as the most influential variables driving the risk of cardiopulmonary complications.

Study Design

Type

Observational (n=718)

Structured PICO

Can machine learning models accurately predict postoperative cardiopulmonary complications in patients undergoing lobectomy for NSCLC?

Population

718 patients who underwent lobectomy for non-small cell lung cancer between 2017 and 2021, with a 20.89% incidence of postoperative cardiopulmonary complications.

Exposure

Machine learning models (logical regression, decision tree, random forest, gradient boost DT, and eXtreme gradient boosting) combined with SHAP interpretability

Outcome

Postoperative cardiopulmonary complications

Main Result

Effect estimate: AUC 0.856 (95% CI 0.815-0.898)

A random forest machine learning model combined with SHAP interpretability can accurately predict postoperative cardiopulmonary complications in patients undergoing lobectomy for non-small cell lung cancer.

Cite This Study

Zhai et al. (2023) conducted an observational in Non-small cell lung cancer (NSCLC) (n=718). Machine learning prediction models (Random Forest) was evaluated on Postoperative cardiopulmonary complications (AUC 0.856, 95% CI 0.815-0.898). A Random Forest machine learning model predicted postoperative cardiopulmonary complications after lobectomy for non-small cell lung cancer with an AUC of 0.856 (95% CI 0.815-0.898).

synapsesocial.com/papers/6a8c239c1692ae2bc0f47f5b https://doi.org/10.1016/j.heliyon.2023.e17772

View Full Paper

Ask AI

Bookmark

Interpretable prediction of cardiopulmonary complications after non-small cell lung cancer surgery based on machine learning and SHapley additive exPlanations

Discussion

Member takes

Implication

Key Points

Study Design

Structured PICO

Main Result

Cite This Study

Interpretable prediction of cardiopulmonary complications after non-small cell lung cancer surgery based on machine learning and SHapley additive exPlanations

Discussion

Member takes

Implication

Key Points

Study Design

Structured PICO

Main Result

Cite This Study

Also Consider

Also Consider