Speaker
Description
This presentation is an overview on radionuclide analysis procedures at the IDC. There are three sets of approaches currently, the single channel analyser curve for particulate, the least squares regression on gamma- and X-rays peaks of xenon isotopes for high resolution spectra and the net count calculation method for beta-gamma coincidence spectra for noble gas, which are based on conventional frequentist statistics. Most daily IMS spectra have low counts close to background level. Decision thresholds by Currie’s definition have been found to tend being underestimated, resulting in false positive detections. Enhancements on current methods could apply optimization regression analyses of standard spectra, 3-D fitting and gross counts, or machine learning which all are consistent with ISO standards on estimation of measurement uncertainty and characterization limits based on Bayesian statistics. Probability distributions of measurands, e.g. activity, concentration and isotopic ratio, could be obtained by the Monte-Carlo method, directly based on distributions of inputs of measurement spectra, calibration data and related parameters, resulting in realistic estimates for measurands, their uncertainties and associated limits of the coverage interval with a given probability. IDC radionuclide analysis reports could be enhanced by reporting not only results and their uncertainties but also associated characterization limits.
Promotional text
The overview on past achievements and potential developments of radionuclide analysis procedures at the IDC reveals the challenge to enhance estimation of measurement uncertainty and characterization limits, improving analysis reliability.
