NON-DESTRUCTIVE ANALYSIS OF BARIUM SULPHATE TABLETS BY NEAR-INFRARED SPECTROSCOPY

A novel formulation of barium sulphate tablets was developed to replace the commercial radiopaque markers, since the commercial products are quite expensive and there is a restriction access for Thai patients.The quantitative determination of barium sulphate according to the United State Pharmacopoeia is performed using a gravimetric method which requires tedious sample preparation steps.In this study, fast and non-destructive analysis of barium sulphate tablets by benchtop and portable near infrared spectrometry (NIR) combined with chemometrics approach was developed and compared with reliable inductively coupled plasma-optical emission spectrometry (ICP-OES).Partial least square regression (PLSR) was the chemometrics tool selected in this study.The optimum PLSR model for benchtop NIR was obtained from standard normal variate (SNV) and orthogonal signal correction (OSC) pretreated NIR spectral data with 1 latent factor, R 2 Pearson for prediction was 0.9727, root mean square error of prediction (RMSEP) and bias were 3.7265 and 0.6614.While the best model for portable NIR was also obtained from SNV and OSC pretreated NIR spectral data with 2 latent factors, R 2 Pearson for prediction was 0.9371, RMSEP and bias were 5.6147 and -0.2604.The assay results of benchtop and portable NIR were not significantly different from ICP-OES. RezumatUlterior obinerii unei formulri cu sulfat de bariu sub form de comprimate, menit s nlocuiasc markerii comerciali actuali de radioopacitate, determinarea cantitativ a API a fost realizat conform printr-o metod rapid i nedistructiv de spectrometrie benchtop n infrarou apropiat (NIR), combinat cu abordarea chemometric i a fost comparat cu spectrometria de emisie optic cu plasm cuplat inductiv (ICP-OES).Regresia parial a celor mai mici ptrate (PLSR) a fost instrumentul chemometric selectat n acest studiu.Modelul PLSR optim pentru NIR de benchtop a fost obinut din date spectrale NIR cu variaii normale standard (SNV) i corecie a semnalului ortogonal (OSC) cu 1 factor latent, factorul R2 Pearson pentru predicie fiind regsit 0,9727, eroarea ptrat medie a prediciei (RMSEP) i bias au fost calculate 3,7265 i respective 0,6614.Cel mai bun model pentru NIR portabil a fost, de asemenea, obinut din datele spectrale NIR pretratate SNV i OSC, dar cu 2 factori lateni, factorul R2 Pearson pentru predicie a fost 0,9371, RMSEP i bias-ul au fost 5,6147 i respectiv -0,2604.Rezultatele testului NIR de benchtop portabil nu au fost semnificativ diferite de cele obinute prin intermediul ICP-OES.