Object and person tracking systems to enable extremity dosimetry in nuclear medicine using computational methods

• Published in: Journal of radiological protection Vol. 44; no. 2
• Main Authors: Rondón, Daniel Santiago, Lombardo, Pasquale, Abdelrahman, Mahmoud, Struelens, Lara, Vanhavere, Filip, Bergans, Niki
• Format: Journal Article
• Language: English
• Published: England 01-06-2024
• Subjects: Algorithms; Artificial Intelligence; Hand - radiation effects; Humans; Monte Carlo Method; Nuclear Medicine; Occupational Exposure - analysis; Radiometry - methods; Syringes; radiation protection; computational dosimetry; nuclear medicine; occupational dosimetry; computer vision
• ISSN: 1361-6498
• DOI: 10.1088/1361-6498/ad53d5

Nuclear medicine (NM) professionals are potentially exposed to high doses of ionising radiation, particularly in the skin of the hands. Ring dosimeters are used by the workers to ensure extremity doses are kept below the legal limits. However, ring dosimeters are often susceptible to large uncertainties, so it is difficult to ensure a correct measurement using the traditional occupational monitoring methods. An alternative solution is to calculate the absorbed dose by using Monte Carlo simulations. This method could reduce the uncertainty in dose calculation if the exact positions of the worker and the radiation source are represented in these simulations. In this study we present a set of computer vision and artificial intelligence algorithms that allow us to track the exact position of unshielded syringes and the hands of NM workers. We showcase a possible hardware configuration to acquire the necessary input data for the algorithms. And finally, we assess the tracking confidence of our software. The tracking accuracy achieved for the syringe detection was 57% and for the hand detection 98%.