Banca de DEFESA: LAURINDA FERNANDA SALDANHA SIQUEIRA

Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.
DISCENTE : LAURINDA FERNANDA SALDANHA SIQUEIRA
DATA : 30/01/2017
HORA: 14:00
LOCAL: Anfiteatro A do CCET
TÍTULO:

MULTIVARIATE CLASSIFICATION AND MID-INFRARED SPECTROSCOPY IN PROSTATE CANCER TISSUES

PALAVRAS-CHAVES:

Cancer. Multivariate Classification. FT-MIR. 

PÁGINAS: 200
GRANDE ÁREA: Ciências Exatas e da Terra
ÁREA: Química
SUBÁREA: Química Analítica
ESPECIALIDADE: Instrumentação Analítica
RESUMO:

This research is a theoretical and practical support to the differentiation of prostate cancer types by multivariate classification applied in MIR spectra derived from human tissues. For this, it sought to identify spectral differences between the prostate cancer stages, to determine biochemical markers responsible for differentiation and to compare and evaluate the performance of multivariate classification methods. Prostatic tissue samples classified into Gleason II, III and IV were used. 40 to 100 transmission spectra were collected per sample (resolution 8 cm^-1, 32 scans). FT-MIR spectra were cut from 900–1800 cm^-1 (fingerprint region) and preprocessed by ESMC, 1st order Savitzky-Golay smoothing (15 points) and normalized to amide I peak (»1,650 cm^-1). In a first study, the multivariate classification models PCA-LDA, SPA-LDA and GA-LDA were constructed aiming a methodology for prostate cancer discrimination, considering Gleason's graduation and a categorization of 'Low and High grades'; and also biochemical markers identification. The models performances were compared. GA-LDA obtained the most satisfactory results, being better in the perspective of 'Low and High grades'. Correct classification rates »81% and »83% for the two categorizations, respectively; and sensitivity values around »72-100% and specificity between »77-80%, respectively, were obtained. In a second work, PCA-LDA/QDA and GA-LDA/QDA were applied to compare linear and quadratic classification approaches and to add up methodological tools for prostate cancer screening, working with a perspective of ‘Low and High grades' differentiation. The QDA models obtained better results than the LDA, as well as variable selection methods (GA) were better than the variables reduction method (PCA). GA-QDA obtained better performance with correct classification rates for calibration and prediction sets of »97% and »100%, respectively; and sensitivity and specificity of »75% and »100%, respectively. In a third study, SVM models (linear, polynomial, RBF and quadratic) and the PCA-SVM, SPA-SVM and GA-SVM algorithms were applied to evaluate the use of variables reduction and selection methods in a non- linear classification approach for screening 'Low and High grades' of prostate cancer. The SVM models obtained lower performances than the others. The best model was GA-SVM with »100% and »90% correctly classified calibration and prediction samples, respectively; and sensitivity and specificity of »90%. Potential biomarker identified by the studies were: the regions of secondary proteins (≈ 1591–1483 cm^-1), DNA and RNA (≈1000–1490 cm⁻¹) and protein phosphorylation ( ≈ 970 cm^-1). Changes in these regions can indicate metabolic alterations caused by cancer stages. The results indicated that multivariate classification coupled to FT-MIR enables differentiate pathological states of tissues even in the early stages of the prostate cancer with objectivity, speed, accuracy, easy procedure, independence of intra and inter-observer variability, results reliability and high sensibility and specificity; in comparison to traditional techniques which suffer with operator-dependence, high intra and inter-observer variability, time consuming, difficult proceedings, and  lower sensitivity and specificity. The techniques proposed may cause economic and social beneficial impacts provided by the precocious diagnosis and feasibility of the treatment in the early stages, which it will can allow gains in quality of life and survival of patients.

MEMBROS DA BANCA:
Externo ao Programa - 1913849 - EDGAR PERIN MORAES
Externo à Instituição - EDVAN CIRINO DA SILVA - UFPB
Presidente - 1714946 - KASSIO MICHELL GOMES DE LIMA
Externo à Instituição - LUCIANO FARIAS DE ALMEIDA - UFPB
Interno - 1523912 - TATIANA DE CAMPOS BICUDO