From deep learning research to clinical impact: TU Delft and HollandPTC joining forces

Scientists led by dr. Zoltán Perkó and dr. Danny Lathouwers (Delft University of Technology, Radiation Science and Technology) developed two algorithmic solutions for repeated dose-calculations for proton therapy. With a fast metamodeling approach and a deep learning model, treatment plans based on many tens, hundreds or even thousands of repeated dose-calculations can be fine-tuned faster.

HollandPTC is an independent outpatient center for proton therapy, scientific research and education and was founded in the Medical Delta context by Erasmus MC, LUMC and TU Delft. In addition, HollandPTC Medical Delta scientific program on HTA value proposition.

Accurate dose calculation

Proton therapy can target the desired dose to the tumor more accurately than traditional radiotherapy. Healthy tissues are therefore better preserved and the risk of side effects is reduced. The increased accuracy in targeting the protons also makes the treatment more sensitive to uncertainties. For example, patient positioning, proton range or organ movement play a greater role. These are examples of the uncertainties that must be taken into account when creating a so-called proton plan.

To accurately evaluate how the final plans handle all these inaccuracies, many dozens, hundreds or even thousands of repeated dose-calculations are needed. Ideally, such an accurate quality evaluation would be performed daily, but current methods of dose-calculation were too slow until recently. The algorithms developed by dr. Zoltán Perkó and dr. Danny Lathouwers may be able to change that.

Clinical implementation

The results of thorough verification and testing of the two corresponding open-sourced software packages, have now been published in peer-reviewed literature. The findings show that there is great potential to improve the evaluation of robustness and ultimately the quality of the treatment plan.

To turn that potential into real-world impact, these algorithms must be implemented clinically and integrated with existing systems. This will allow further testing required for actual clinical use. To take this step, TU Delft's Digital Competence Centre (DCC) is partnering with HollandPTC and the RST department in a year-long collaborative project. The project aims to deliver a reliable, streamlined and automated solution for rapid dose calculation in a clinical environment that meets clinical requirements and enables medical professionals to conduct further research and better evaluate robustness.

Better outcomes for patients

By the end of the one-year collaboration, it is expected that there will be two well-documented, tested and versatile software packages that can be seamlessly interfaced with current clinical treatment planning systems to dramatically accelerate dose-calculation times. By adhering to open-source software principles, the project will benefit not only HollandPTC but also other proton clinics, allowing the final codes to be used directly for research and as consultative clinical tools. This could provide more accurate robustness evaluations, which could reduce the current overconservativeness of clinical proton treatments and ultimately lead to better patient outcomes.

Read more here.