Discrimination of lung tumor and boundary tissues based on laser-induced breakdown spectroscopy and machine learning

Abstract

The boundary between lung tumor and boundary tissue is blurry. If tumor resection is not complete, recurrence may affect patient survival. Laser-induced breakdown spectroscopy (LIBS) is used to distinguish lung tumor and boundary tissues. Support Vector Machine(SVM)and Boosting Tree classification models combine Principle component analysis (PCA) or Random forest (RF) to optimize input variables to enhance accuracy, sensitivity, specificity, ROC curves, and AUC through 10-fold cross-validation. The RF-Boosting Tree model shows the highest accuracy (98.9%). RF feature selection better retains data information, removes redundant information and interference, and reduces training time. LIBS and an RF-Boosting Tree model is a rapid and robust method for identification of lung tumor boundaries.