Feasibility for early cancer diagnostics by machine learning enabled synchrotron radiation based micro X-ray fluorescence imaging of trace biometals as cancer biomarkers

Abstract

Trace quantitative spectroscopic imaging has the potential to provide the location and distribution of trace biometals in a biological sample for disease (cancer) diagnostics. However, spatial qualitative analysis of the trace biometals for cancer diagnostics remains a challenge due to the complex biological matrices that result in weak analyte signals and intricate multivariate relationships between the analyte spatial distribution and disease (cancer) state. In this study, principal component analysis (PCA)-enabled artificial neural networks (ANN) for micro X-ray fluorescence for simultaneous determination of biometal (Mn, Fe, Cu and Se) spatial profiles as biomarkers for cancer diagnosis in model human cell cultures (DU145 and Vero). The cell lines were cultured on silicon nitride membranes and micro XRF analysis carried out at TwinMic beamline, Elletra synchrotron source at beam excitation energy of 1.7 keV. These enabled 2D mapping of Mn, Fe, Cu and Se using cylindrical beam dimensions of Æ690 nm in stepper motor-controlled step sizes of 0.6μm with dwell times of 10s per pixel. Python multichannel analyzer (PyMca) software was used for spectral deconvolution and determination of 2D maps of the trace biometals. Principal component analysis (PCA) reduced the data dimensions for optimal artificial neural networks (ANN) exploratory modelling of cancer pathogenesis stages utilizing the pixel spectral profile of the trace biometals (Mn, Fe, Cu, Se). The 2D spatial distribution maps of the trace biometals revealed high spatial correlations between Cu and Fe ((0.941) and (0.923)) in DU_D3 and DU_D4 in cancerous compared to normal cell culture stages (V_D3 and V_D4) at 0.661 and 0.203 respectively. Utilizing PC1 and PC2 scores from selected fluorescence L (Lα and Lβ) lines of Fe and Cu, ANN distinctively classified the cell cultures into cancerous and healthy groups. Further, the selected fluorescence L-lines of Fe, Cu and Compton scatter spectral profiles enabled ANN classification of cancerous cultured cells into early, intermediate and advanced stages. The study has provided proof of concept for early diagnosis of cancer based on the multivariate alterations and spatial distribution of the trace biometals as cancer biomarkers.