Every year, more than 13,000 men in the Netherlands get prostate cancer. This makes it one of the most common types of cancer in our country. Thanks to the MR-Linac, a combination of an MRI scanner and an radiation device, it is possible to irradiate prostate cancer very precisely. This leads to fewer erection problems afterwards. Scientists at UMC Utrecht are now trying to make the technology even better, for example by using artificial intelligence (AI).
Prostate cancer is increasingly being treated by targeted radiation of the prostate. In many cases, this means that major surgery is no longer necessary. With radiotherapy, it is important that enough radiation reaches the prostate to completely fight the cancer. But surrounding tissues and organs should be spared as much as possible. These include the bladder, rectum, and the nerve and vascular structures responsible for erections.
Radiation area ‘live’ image
Targeted irradiation is already becoming increasingly successful, partly thanks to the MR-Linac: an invention from Utrecht that is now being used worldwide. This is an innovative combination of an MRI scanner and an radiation device, a so-called linear accelerator. This allows doctors to visualize the entire area ‘live’ with MRI images during treatment. This makes it possible to direct the radiation very precisely. Doctors can also immediately correct the direction of the radiation if the prostate shifts a little, for example due to filling of the bladder or due to movements of the intestines.
After making an MRI scan, a radiation oncologist still draws in the relevant structures manually. PhD candidate Inge van den Berg is a technical physician at the Department of Radiotherapy at UMC Utrecht. She is researching the application of artificial intelligence to speed up and improve the drawing of the images.
Draw structures
“Plotting all the structures on the MRI images is very important, because then it becomes clear where the radiation should be concentrated and where the amount of radiation should remain below a certain dose. The doctor now does this manually and that takes a lot of time. Together with computer scientist Mark Savenije, I am experimenting with algorithms that can draw in all important structures very quickly in advance. Ultimately, we also want to use these types of AI techniques during treatment with the MR-Linac, so that the radiation direction is also corrected automatically. With this we want to shorten the treatment time and more treatments can be performed per day with the MR-Linac.”
Faster and more accurate irradiation
“My research focuses specifically on the nerve and vascular structures responsible for erection. It is obviously important for many patients to spare those structures as much as possible. You now see – in addition to the fact that the exact drawing of those structures takes a lot of time – that every radiation oncologist does it a little differently. We think this can be done faster and more consistently. After all, the computer learns from a lot of entries and we pass deep learning please. So if it is consistent then we can further optimize the registration so that it is ultimately both faster and more accurate. Because the better the registration, the better and more targeted the radiation and hopefully fewer erection problems afterwards.”
Questions, comments or tips for the editors?