‘TLDetect’ – A New Algorithm for Detecting and Correcting Anomalies in Glow Curves of Thermoluminescence Dosimeters

A pressing issue for External Dosimetry Laboratories (EDLs) is the development of a reproducible and accredited process of automatically evaluating large amounts of glow curves (GCs) of Thermoluminescence Dosimeters (TLDs).

A new algorithm for the detection and correction of anomalies in GCs of TLDs was developed by the EDL of the Radiation Safety Division at Soreq Nuclear Research Center (SNRC). This new algorithm, ‘TLDetect’, can be used by any EDL using any commercial TLD reader for evaluating irregular GCs that would usually need a human review. ‘TLDetect’ identifies anomalies and suggests a correction method, when applicable.

The first stage of TLDetect workflow is performing an initial data processing over the GCs data. This data processing includes some steps: skipping all GCs with doses below EDL reporting level, spikes smoothing and saving all the statistical relevant data for each GC - GC integral, max, min, average and median values.

The second stage of TLDetect is automatically classifying each of the GCs into one of 4 main classes:

Class A – high background of low temperature TL / high temperature TL (LTTL/HTTL), Class B – non-standard GC width (either too narrow or too wide), Class C – main GC peak shifting and Class D –`noisy` GC (too many spikes). This classification is performed using certain mathematical constraints that fit each of the classes. All these constraints are checked against actual GCs in the process of development, using a MATLAB tool that was specially developed for this purpose.

The third stage of TLDetect is suggesting a correction method for each GC, when applicable. At the end of TLDetect process of all selected GCs, the algorithm saves the number of classified GCs into each of the classes.

This tool will dramatically shorten the time and improve the quality of analysis of GCs by the EDL at SNRC.