Validating urea-nitrogen CARS spectra with collaborative filtering

As someone passionate about improving end stage renal disease (ESRD) treatment, I covered a program that I developed which enhances the success of dialysis. ESRD is a condition where the kidneys lose function, requiring dialysis to remove urea nitrogen from the blood via dialysate fluid. Accurate analysis of dialysate samples is important, as it reveals excretory success, and Coherent Anti-Stokes Raman Spectroscopy (CARS) is ideal for this purpose. CARS measures the molecular composition of a sample by using light interactions to produce a molecule vibrational signal, revealing urea nitrogen concentrations. To ensure the validity of CARS results, as measurements can be inaccurate, I’ve integrated collaborative filtering techniques into a program. These techniques find patterns from user input to identify valid samples. My program has two algorithms: Byzantine Generals and Iyengar, the latter providing more in-depth analysis. Both algorithms compare user input parameters—such as wavelengths, intensities, and concentrations—to a database of lab values from Atonarp's pilot study. The tool interface allows users to input their spectral data, select an algorithm for validation, and get results with comparison graphs. If no data is available, sample data can be copied and pasted. Tested with real-life data from Atonarp, this tool can be used to make the dialysis process safer and more accurate.