Analysis of Specificity and Limitations Applying the Receiver Operating Characteristic Curve and Laser-Induced...

Laser-induced breakdown spectroscopy (LIBS) offers a promising alternative due to its minimal sample preparation, real-time analysis capabilities, and versatility in analyzing a broad range of materials. However, the challenge lies in determining its ability to effectively distinguish high-iron ore content from mineralogically similar ores with lower iron content. This study focuses on differentiating iron ore from a variety of ores with lower iron content, including calcite, biotite, dolomite, chalcopyrite, rutile, chromite, olivine, limonite, and astrophyllite, using LIBS. By comparing the obtained spectra and applying receiver operating characteristic (ROC) curve analysis, the study assesses the specificity of the technique. The results demonstrate a high specificity (>70%) in distinguishing iron ore from biotite, dolomite, chalcopyrite, rutile, olivine, and astrophyllite, revealing the potential of LIBS for effectively identifying iron ore from some ore types. However, when comparing iron ore to other ore types, such as limonite, chromite, and calcite, the results are not statistically significant. This means that the spectral or compositional similarities between these ores may limit the method's capacity to give clear separation in certain situations. To further validate the results, two common classification models, principal component analysis followed by linear discriminant analysis (PCA + LDA) and k-nearest neighbors (KNN) were applied to the spectral data. The comparison results demonstrate the resilience of LIBS classification and the impact of mineral matrix influences on diagnostic performance.