ML2C - Machine Learning for robust, real-time dosimetry and MultiLeaf Collimator verification

STFC is heading a collaboration with ICEC and CERN to develop low cost radiotherapy for challenging environments like low income countries. When delivering radiotherapy, acceptance testing and routine (e.g. monthly) quality assurance of LINACs is a time-consuming task involving ancillary equipment such as water-equivalent dosimeters and phantoms, and requiring the presence of a specialist expert (in the UK the Medical Physics Expert). A shortfall in local expertise is a common factor associated with significant downtime of LINAC systems and of erroneous treatments. This is a particularly difficult problem in low income countries where there is a significant shortage of well-trained experts.

We are developing a real-time treatment verification system for radiotherapy that can autonomously perform the Quality Assurance measurements and verify the treatments in real time. This means that the system can successfully be used to deliver treatment without the need for a locally present highly skilled expert. Our current system works well, but our algorithms are not robust for problems that may occur in the device. We have started to use machine learning techniques to develop more robust algorithms. The first results are very promising. We now want to continue the development and furthermore produce algorithms to monitor the health of our detector system and the linac.

This research will form a corner stone of a future GCRF grant in which funds will be requested to produce a first prototype of the low cost radiotherapy machine. If the GCRF bid and the subsequent developments are successful, then low cost radiotherapy can be delivered in many places across the world where it currently is not available. This will enhance the quality of life of millions of people who now do not receive appropriate treatment.