Diagnosis and staging of multiple myeloma using serum-based laser-induced breakdown spectroscopy combined with machine learning methods.

Abstract

Diagnosis and staging of multiple myeloma (MM) have been achieved using serum-based laser-induced breakdown spectroscopy (LIBS) in combination with machine learning methods. 130 cases of serum samples collected from registered MM patients in different progressive stages and healthy controls were deposited onto standard quantitative filter papers and ablated with a Q-switched Nd:YAG laser. Emissions of Ca, Na, K, Mg, C, H, O, N and CN were selected for malignancy diagnosis and staging. Multivariate statistics and machine learning methods, including principal component analysis (PCA), k-nearest neighbor (kNN), support vector machine (SVM) and artificial neural network (ANN) classifiers, were used to build the discrimination models. The performances of the classifiers were optimized via 10-fold cross-validation and evaluated in terms of accuracy, sensitivity, specificity, and area under the receiver operating characteristic curves (AUC). The kNN, SVM and ANN classifiers achieved comparable discrimination performances with accuracies of over 90% for both diagnosis and staging of MM. For diagnosis of MM, the classifiers achieved performances with AUC of ∼0.970, sensitivity of ∼0.930 and specificity of ∼0.910; for staging of MM, the corresponding values were AUC of ∼0.970, sensitivity of ∼0.910 and specificity of ∼0.930. These results show that the serum-based LIBS in combination with machine learning methods can serve as a fast, less invasive, cost-effective, and robust technique for diagnosis and staging of human malignancies.