Key points are not available for this paper at this time.
Bacteria are the primary cause of infectious diseases, making the rapid and accurate identification of bacterial species crucial for timely diagnosis and control of disease transmission. However, traditional techniques like Polymerase Chain Reaction (PCR) and Loop-Mediated Isothermal Amplification (LAMP) are complex, time-consuming, and pose potential infection risks due to close contact. This study utilizes a Cassegrain reflective telescope combined with Laser-Induced Breakdown Spectroscopy (LIBS) for remote (~3 meters) bacterial identification. After dimensionality reduction of the collected LIBS spectral data using Principal Component Analysis (PCA), the identification accuracy of two classification algorithms, Support Vector Machine (SVM) and Random Forest (RF), was compared. Five-fold cross-validation results show that the SVM model with an RBF kernel achieved a classification accuracy, recall, precision, and F1 score of 98.11%, 98.90%, 99.00%, and 0.9895, respectively. In contrast, the RF model achieved 99.81%, 99.80%, 99.79%, and 0.9979, respectively, demonstrating superior performance. This study indicates that the method of bacterial identification based on LIBS combined with machine learning has broad application prospects, providing crucial support for timely, non-contact diagnosis of infectious diseases, enhancing public health standards, and promoting advancements in medical research and technology development.
Cheng et al. (Thu,) studied this question.