Partial least-squares regression for routine analysis of urinary calculus composition with Fourier transform infrared analysis.

Quantitative assessment of urinary calculus (renal stone) constituents by infrared analysis (IR) is hampered by the need of expert knowledge for spectrum interpretation. Our laboratory performed a computerized search of several libraries, containing 235 reference spectra from various mixtures with different proportions. Library search was followed by visual interpretation of band intensities for more precise semiquantitative determination of the composition. We tested partial least-squares (PLS) regression for the most frequently occurring compositions of urinary calculi. Using a constrained mixture design, we prepared various samples containing whewellite, weddellite, and carbonate apatite and used these as a calibration set for PLS regression. The value of PLS analysis was investigated by the assay of known artificial mixtures and selected patients' samples for which the semiquantitative compositions were determined by computerized library search followed by visual interpretation. Compared with that method, PLS analysis was superior with respect to accuracy and necessity of expert knowledge. Apart from some practical limitations in data-handling facilities, we believe that PLS regression offers a promising tool for routine quantification, not only for whewellite, weddellite, and carbonate apatite, but also for other compositions of the urinary calculus.