IMPROVEMENT OF DOSE ESTIMATION PROCESS USING ARTIFICIAL NEURAL NETWORKS.

Area of Science:

• Medical Physics
• Radiation Dosimetry
• Artificial Intelligence in Science

Background:

• Thermoluminescence dosimetry (TLD) is crucial for radiation dose assessment.
• Thermoluminescence dosemeter (TLD) glow curves (GCs) can exhibit anomalies due to various physical parameters.
• Manual review of TLD GCs for anomalies is time-consuming and prone to error.

Purpose of the Study:

• To introduce a novel, fast, and reliable artificial neural network (ANN) algorithm for automatic anomaly detection in TLD GCs.
• To compare the performance of the ANN algorithm against previously developed support vector machine (SVM) classifiers.
• To enable automated classification of TLD GCs into 'anomalous' and 'regular' categories for dosimetry laboratories.

Main Methods:

• Development and implementation of an artificial neural network (ANN) algorithm for TLD glow curve (GC) anomaly detection.
• Comparison of the ANN algorithm's performance with regular and weighted support vector machine (SVM) classifiers.
• Evaluation using three distinct performance metrics to assess classification accuracy.

Main Results:

• The proposed artificial neural network (ANN) algorithm demonstrates a high accuracy rate of 97% in correctly classifying thermoluminescence dosemeter (TLD) glow curves (GCs).
• The ANN algorithm effectively distinguishes between 'anomalous' and 'regular' TLD GCs.
• Performance metrics indicate superior or comparable results to existing support vector machine (SVM) methods.

Conclusions:

• The artificial neural network (ANN) algorithm offers a fast, reliable, and accurate solution for automated anomaly detection in thermoluminescence dosemeter (TLD) glow curves (GCs).
• This automated approach can significantly enhance the efficiency and reliability of dosimetry laboratory workflows.
• The ANN method provides a robust alternative to traditional manual review and existing machine learning classifiers for TLD GC analysis.