Screening of neoplastic diseases by statistical analysis of urine fluorescence spectroscopic data. Application of multivariate techniques for enhancing classification.

Autor: Corti A; Departamento de Física, Facultad de Ciencias Exactas, UNLP, La Plata, Argentina; Instituto de Investigaciones Fisicoquímicas Teóricas y Aplicadas (INIFTA), Departamento de Química, Facultad de Ciencias Exactas, UNLP, CONICET), Sucursal 4, Casilla de Correo 16, 1900 La Plata, Argentina. Electronic address: agustina.corti@fisica.unlp.edu.ar., Pasquale MA; Instituto de Investigaciones Fisicoquímicas Teóricas y Aplicadas (INIFTA), Departamento de Química, Facultad de Ciencias Exactas, UNLP, CONICET), Sucursal 4, Casilla de Correo 16, 1900 La Plata, Argentina., García Einschlag FS; Instituto de Investigaciones Fisicoquímicas Teóricas y Aplicadas (INIFTA), Departamento de Química, Facultad de Ciencias Exactas, UNLP, CONICET), Sucursal 4, Casilla de Correo 16, 1900 La Plata, Argentina.
Jazyk: angličtina
Zdroj: Journal of photochemistry and photobiology. B, Biology [J Photochem Photobiol B] 2023 Jan; Vol. 238, pp. 112598. Date of Electronic Publication: 2022 Nov 14.
DOI: 10.1016/j.jphotobiol.2022.112598
Abstrakt: The composition of human fluids is modified during the course of neoplastic diseases. Urine analysis offers the advantage of being a noninvasive method for which samples are easily and routinely collected from patients. In this work, urine fluorescence spectra recorded upon excitation at 405 nm were obtained from healthy volunteers and individuals with different oncologic pathologies. A large number of indexes, i.e., parameters obtained from spectral data which assist spectral features characterization, were developed to classify healthy and pathological populations. The discrimination ability of simple predictive indexes, obtained from spectra pretreated with different normalization procedures and by taking their derivatives, was statistically assessed. In addition, multivariate methods, such as principal component analysis and multivariate curve resolution by alternating least squares, were used to develop more elaborate indexes for distinguishing between healthy and pathological populations. All indexes were systematically evaluated on a statistical basis by in lab-developed routines capable of detecting outliers, judging the normal distribution of the indexes, evaluating variance homogeneity, testing the difference between the means of healthy and pathological populations, as well as performing a receiver operator curve analysis to assess the classification power of each index. Those indexes with the best performances were further combined to perform a linear discriminant analysis, which yielded a powerful classification algorithm with an area under the receiver operator curve of 0.986, a sensitivity of 97.7%, a specificity of 100%, and an overall accuracy of 98.8%. The present study shows that the statistical analysis of urine fluorescence data with a proper combination of multivariate techniques bears a high potential to develop massive screening tests for the early detection of oncologic pathologies.
Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
(Copyright © 2022 Elsevier B.V. All rights reserved.)
Databáze: MEDLINE
Externí odkaz: