The Applications of Raman Spectroscopy Assisted Urinalysis: Hematuria and Bladder Cancer Detection

Early detection and screening for urinary tract illnesses is a complex and widespread process which has implications for both preventative care, diagnosis, and treatment monitoring. In this paper, we investigate the use of Raman spectroscopy (RS) for the analysis of urine, a complex biological solution, for the detection of bladder cancer (BCa) and hematuria. Raman spectroscopy is a rapid, low cost, non-destructive analysis method with wide-ranging applicability due to the holistic data capturing nature of the scanning technique. Each Raman scan can be considered a 'snapshot' of the molecular makeup of the sample, and, through proper applications of algorithmic transformation and statistical analysis, many types of assessments can be performed on each sample. In this paper we address creating and utilizing a data pipeline for the purposes of analyzing and characterizing potential samples with hematuria and BCa. The algorithmic transformations utilized include baselining using either the Goldindec or ISREA methods, and intensity normalization. The statistical analysis methods utilized include principal component analysis (PCA), discriminant analysis of principal components (DAPC), analysis of variance (ANOVA), pairwise ANOVA, leave-one-out cross-validation (LOOCV), and partial least squares regression (PLSR). These components of the data pipeline serve to output qualitative or quantitative data, depending on the application. The Rametrix toolbox encompasses the tools required to transform and assess Raman spectra with PCA and DAPC. Using the Rametrix toolbox as well as ANOVA, pairwise ANOVA, and LOOCV, we were able to significantly detect the presence of bladder cancer in a specimen with 80% accuracy. Using the Rametrix toolbox, ANOVA, pairwise ANOVA, LOOCV, and PLSR, we were able to classify samples as pure urine, micro-, or macrohematuria with a greater than 91% accuracy, and quantify the amount of blood in the sample with a high correlation (R-squared value of 0.92). In combination, this style of data pipeline is shown to rapidly and accurately test for multiple symptoms or diseases using similar methodologies.