The staff working in nuclear medicine is frequently exposed to high quantities of ionising radiation, mainly during the preparation and injection of radiopharmaceuticals into patients. Their whole body doses but especially their extremity and skin doses are among the highest of all professionally exposed workers. The EC funded ORAMED project (www.oramed-fp7.eu) showed that there is a clear risk of exceeding the legal dose limits of the skin. With the growing use of nuclear medicine therapy, these doses are going to increase even further, together with the number of exposed personnel. However, unfortunately the monitoring of extremity exposures is not straightforward. The estimation of tissue doses is subject to high uncertainties. This is due to the fact that the highest exposure locations not only are unknown in advance, but they also vary dramatically from one person to another. Futhermore, existing ring dosemeters have far from perfect charactherisitcs. First, they can hinder the work of the medical staff by reducing the mobility of the fingers; secondly, from a dosimetric standpoint they offer low performances, especially with less penetrating particles.
On the other hand, computational methods for dose assessment are evolving very fast. In the past, simplified mathematical phantoms were used, while now very detailed voxel and B-Rep phantoms are available. Thanks to B-Rep, computational phantoms can be designed in all kind of shapes and postures, so that they represent more accurately the real individuals. Thanks to these B-Rep phantoms, it is feasible to perform Monte Carlo simulations so that personalized doses can be calculated. With increasing computational power, we expect that such calculations will go faster and faster, allowing, one day, to deliver accurate dose estimations even in real-time (online).
The objective of this PhD project is to develop a simulation software tool for nuclear medicine in which occupational doses are calculated instead of measured. Thanks to the introduction of GPU computing, Monte Carlo simulations are becoming substantially faster, delivering fairly accurate dose estimations within tens of seconds. By means of smart simulation algorithms, such delays could be even lowered. To allow these dose calculations, the spatial distribution of radiation field, including energy and angular distributions, needs to be known, together with the real movement of the exposed medical staff. To acquire positions and postures within the simulation tool, this PhD project will develop a computer vision algorithm based on depth cameras and neural networks. This algorithm will be integrated in the framework of the simulation tool, and it will be used to define the geometry inputs of the Monte Carlo simulations.
This idea is already being explored for interventional procedures, but applying this principle in nuclear medicine will bring new challenges. The tracking accuracy required in this application will be very demanding. Since often beta and beta+ radiation will be involved and the fingers are always very close to the radiation sources, the exact positioning of the fingers compared to the sources needs to be accurately measured by means of cameras. Also, the position of shielding devices need to be carefully included in the simulation tool. Considering the high rates of contaminations occurring in nuclear medicine, the results of in-vivo contamination monitoring will have to be included in the simulations as well.
This PhD project will make important steps towards the implementation of an ideal simulation tool for calculating doses to the extremities and skin of the worker in nearly real-time. Thanks to the tool, it will even become possible to visualize the radiation fields through Virtual Reality, so that workers can see where the highest and lowest dose rates are found. This will allow to minimize doses and apply the ALARA principle.