Seuil de décision et limite de détection : estimation, interprétation et optimisation. 1^repartie : les principes de base

Abstract

RÉSUMÉThe use of decision threshold and detection limit concepts often poses many problems for metrologists in biological analysis laboratories in charge of the management of nuclear plant workers. These problems are frequently related to the understanding of a normalized formula or to the possible choice between several formulas whose numerical results are exceptionally significantly different. The true problems are in fact firstly the significance and the interpretation of these statistical tests, and secondly the numerical values used in these tests. Among those, the background estimate remains the essential and the most delicate value to define. The aim of this first article is to go back to the initial bases of the concepts of the decision threshold and detection limit. In the first step, a distinction will be made between the methods of decision threshold determination based upon definition of a priori first species error risks (i.e. before sample measurement) and the a posteriori interpretation of the sample Article publié par EDP Sciences A. VIVIER et al. 322 RADIOPROTECTION -VOL. 45 -N° 3 (2010) measurement result. In the second step, some examples of decision threshold and detection limit optimization in gamma spectrometry will be described. These examples will show that the essential parameter is the optimization of the estimation of background values, which requires the control of measurement rather than the management of statistical tools.

Detection of Radioactive Gas with Scintillating MOFs

“…On the other side of the concentration range, we have the lowest activity point that cannot be differentiated from the blank measurements. Our linear fit combined with statistical [ 42 ] analysis of our control count rate shows that our system has a decision threshold at 3.03 Bq.cm −3 and a limit of detection at 6.07 Bq.cm −3 . This will be a particular focus in the future, as it will reach a level compatible with the hardest industrial regulation (limit of detection at 1 Bq cm −3 ).…”

Optimisation des programmes de surveillance systématique de l’exposition interne

“…La durée de mesure est de 48 h. Dans la région d'intérêt permettant d'identifier les 239 Pu et 240 Pu qui ne peuvent être séparés par spectrométrie alpha, le bruit de fond moyen est de 2,2 coups par 48 h. Le seuil de décision de 4,2 coups peut en être déduit pour ces isotopes (Vivier et al, 2010). L'efficacité moyenne de détection des particules alphas par le détecteur est de 32,5 %.…”

Decision threshold associated with multiple measurements. Application to the synthesis of the environment monitoring results

“…The identification of the five ROI of barium 133 was carried out under optimal conditions (Vivier et al, 2010). The analysis of the spectrum on each ROI of Barium 133 does not make it possible to detect it, although it was present in the sample.…”