To be frank, the title 'How chaos is improving proton cancer treatment' is misleading. As such, chaos on its own cannot improve proton cancer treatment.
But treatment plans for this disease can be improved if a mathematical approach called polynomial chaos is applied. It makes it possible to evaluate thousands of scenarios for a proton cancer treatment plan instead of the current 9 or 26 at most.This conclusion led to student Sebastian van der Voort of the TU Delft, being nominated for the IMDI Talent Price 2015-2016 of the ZonMW Programme Innovative Medical Devices Initiative (IMDI).
A more localised dose deposition is a factor that distinguishes proton therapy (soon to be carried out at HollandPTC in Delft) from conventional radiotherapy. This advantage is at the same time a blessing and a challenge. Small deviations in dose delivery, for example when a patient moves during treatment, may result in less dose being delivered to the tumour and more dose to surrounding healthy tissues than was initially planned. This is why during the generation of an individualised treatment plan those deviations are taken into account in order to minimise their dosimetric impact. To do so effectively, computer simulations, mimicing these effects, are required. At present, treatment plans are not evaluated extensively, due to time constraints. It would be advantageous to evaluate more scenario’s, but the applied methodologies carried out so far have been very time-consuming.
When people from different fields meet, interesting things can happen Sebastian van der Voort
Sebastian refers to his research under the guidance of assistant professor Danny Lathouwers of the Faculty of Applied Sciences at TU Delft and Mischa Hoogeman at Erasmus MC. “In contacts between clinicians at Erasmus MC and people from the Delft Faculty of Applied Sciences, the simulation issue in proton therapy came up. This led to the idea to use polynomial chaos to tackle this problem,” says Van der Voort.
Polynomial chaos (PC) is a mathematical approach developed by Norbert Wiener in 1938. It is a method for determining the evolution of uncertainty in a dynamic system, when system parameters are characterised by probabilistic uncertainty. “It had been long forgotten,” says Van der Voort. “But lately it has been making a comeback. The method is currently applied in nuclear reactor simulation, as an example of a situation in which you wish to establish the effect of an uncertain input on the output.”
In the case of proton therapy, the “uncertain input” is the patient’s position during treatment. And “treatment” in reality means a sequence of thirty treatments, in which the patient will need to be repositioned each time. “The present simulations can be restrictive,” says Van der Voort. “Polynomial chaos enables faster simulations, alongside providing statistical information on the delivered dose.”
Thousands per second
With polynomial chaos, a couple of thousand simulations can be made in only one second, instead of one-and-a-half hours for a hundred simulations in current practice. “The new approach that we are suggesting makes it possible to take the thirty individual sessions into account during simulation,” says Van der Voort. “Each treatment leads to a new position with a new deviation. So you have the deviation of each individual session, and you have some systematic deviation. The existing approach just couldn’t cope with that, but polynomial chaos can, by combining these deviations.”
The present question is: Which scenarios do you have to select for simulation in order to achieve a treatment plan that is robust enough for positioning errors? The assumption is, says Van der Voort, that when you know the distribution of uncertainties, you have the information to cover (say) 99% of these. “But in reality this knowledge is not sufficient to guarantee that your treatment plan will be adequate in all of those cases. The assumption that ‘If I cover enough possibilities, the result will be positive’ does not hold mathematically. We reason: ‘If you want a positive outcome, which errors do you need to cover?’ The difference may sound subtle. But we reason from the opposite direction – from the outcome. That makes all the difference.” Ultimately, the work has led to so-called robustness recipes that determine which scenarios should be included in the optimisation stage, taking the magnitude of the positioning errors as input.
As yet, Van der Voort doesn’t know whether polynomial chaos will actually be applied in the clinical practice of treatment plans, but he hopes so. “Since we now have the necessary software, we have been able to make a workable link between the relevant software platforms.” Danny Lathouwers points out that this is the first time this technology has been used in healthcare anywhere in the world. “It is in large part the achievement of Sebastian. He was not the first student to work on this topic, but he took the decisive step to transform it from ‘a nice toy’ to ‘a practically relevant technology’. The notion that you can just take systematic errors and random errors together in this approach was his idea. It’s quite special that, as a student, he has a paper in the International Journal of Radiation Oncology, Biology and Physics to his name.”
For Van der Voort, the project stands out as an example of the added value of co-creation. “In order to properly utilise the technical capabilities of TU Delft, you need to have Erasmus MC’s clinical input and you need to understand the relevant prerequisites. Initially, you have to get accustomed to each other’s totally different professional languages. Once the translation has been made and the gap bridged, we get a better picture of the opportunities that will allow us to take the field further. It’s great to be part of that process.”
Van der Voort’s work will definitely be part of the process, says Hoogeman. “We make corrections to minimise uncertainties during treatment, but there will always remain some unavoidable uncertainty. Sebastian has made a recipe that exactly determines how robust we should make a treatment plan against those uncertainties. This recipe will certainly be applied in our clinical practice at Holland PTC.” The field of application for Polynomial Chaos Expansion as ‘black box’ technology is wide, says Lathouwers. “Together with Zoltan Perko of TU Delft, Sebastian is currently working on an open-source toolbox, so that polynomial chaos can be used in all kinds of medical applications.”
IMDI Talent Prize
The IMDI Talent Prize is awarded to the student at the IMDI Centres of Research Excellence (CoREs) who has produced the best graduate research project on new medical aids.
After pitches of the 3 nominees at the ZonMW Public/Private Partnership Day on March 7th, the prize was finally awarded to Ida Poortinga. Poortinga has developed a mobile application that will enable anyone to develop a simple digital finger splint. The other nominee was Laura van Huizen. She carried out research into a new microscope technique for recognising breast tumours.
Innovative Medical Devices Initiative
The Innovative Medical Devices Initiative (IMDI), set up by ZonMw, brings together science, the care sector and business in order to develop and implement medical technologies. To make these care innovations possible, eight Centres of Research Excellence (CoREs) have been set up to focus on the topics of imaging and image analysis, home and revalidation care, and minimally invasive techniques.
Interview by: Leendert van der Ent