Precise and rapid diagnosis of lung cancer: leveraging laser-induced breakdown spectroscopy with optimized kernel methods in machine learning

Abstract

Improving the efficiency of laser-induced breakdown spectroscopy (LIBS) is crucial for its clinical applicability in tumor diagnosis. This study presents an accelerated diagnostic approach based on a kernel principal component analysis-support vector machine (KPCA-SVM) model. Initially, elemental features—calcium (Ca), sodium (Na), magnesium (Mg), and copper (Cu)—were selected due to noticeable differences in spectral intensity between tumor and normal tissues. Subsequently, employing KPCA facilitated the projection of LIBS features into a high-dimensional space, capturing nonlinear data relationships. Dimensionality reduction within this space was then performed to retain essential nonlinear features while eliminating redundancy. The resulting reduced matrix was input into the SVM classifier. Both the Gaussian kernel of KPCA and the Radial Basis Function (RBF) kernel of the SVM exhibited exceptional diagnostic efficacy. Optimal results were attained using 15 principal components, achieving a classification accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of 99.03%, 99.72%, 98.89%, 98.90%, and 99.72%, respectively. Importantly, the model's runtime was only 6.77 seconds, highlighting the potential of KPCA and SVM kernel methodologies for rapid lung cancer diagnosis.