Non-invasive Diabetes Mellitus Diagnostics Using Laser-Induced Breakdown Spectroscopy and Support Vector Machine Algorithm

This work employs a support vector machine (SVM) algorithm and laser-induced breakdown spectroscopy (LIBS) to identify diabetes mellitus in human urine. A total of 35 pathological urine samples from people with diabetes mellitus were obtained, whereas 35 non-diabetic control urine samples were gathered and used for this investigation. The LIBS spectra of both diabetic and normal urine samples were captured using a LIBS spectrometer that was equipped with a laser that operated at the fundamental wavelength of 1064 nm. At particular elemental peaks in the LIBS spectra, significant intensity differences between the two (diseased and normal) groups were observed using principal component analysis (PCA) combined with a support vector classifier. In order to achieve linear separability, support vector machine has the ability to transform data that is linearly inseparable into a high-dimensional space using a mathematical function called a kernel. The SVM model was developed in this work and applied utilizing various kernel functions, notably the linear kernel, polynomial kernel, and Gaussian radial basis function, to distinguish between diseased samples and normal samples. Performance evaluation of the model was carried out using a k-fold cross-validation procedure with k = 10. Of all the kernel functions used in this investigation, the radial basis function (RBF) kernel produced the highest performance efficiency. The results of our study highlight the LIBS approach's potency as a quick and exceptional technique for real-time assessment of diabetes mellitus in clinical trials.