Laser-Induced Breakdown Spectroscopy Combined with Machine Learning for the Identification of Lung Cancer Tumors

Abstract

The diagnosis of lung cancer has always been a challenging clinical issue. In this work, we use laser-induced breakdown spectroscopy (LIBS) combined with machine learning to differentiate samples of lung cancer tumors from those of normal tissues. Sample plasma was collected by laser ablation at 1064 nm to obtain the characteristic spectra of lung tumor and normal tissue samples. Twelve lines of C, Mg, Ca, C–N, Na, and K were selected for the diagnosis of malignancy. Principal component analysis (PCA), support vector machine (SVM), k-nearest neighbors (KNN), Decision Tree, and Bagged Tree were used to establish the discrimination model for tumors and normal tissue. A 10-fold cross-validation method was used to evaluate the discrimination model. The results showed that the integrated learning Bagged Tree model performed best, with an overall accuracy of 98.9%, sensitivity and specificity of 98.6 and 99.3%, respectively, and an area under the curve (AUC) of 0.982. This study suggests that LIBS can be used as a fast and accurate means of identifying human lung tumors.