17. Artificial intelligence to enhance radiotherapy treatment planning for children with renal tumors.

Renal tumors comprise 6% of all pediatric cancers. Despite the excellent prognosis, more than 30% of these patients will develop severe late effects, with radiotherapy being the principal determinant. To reduce this risk, a consensus guideline on highly-conformal flank target volume delineation has been developed on behalf of the International Society of Pediatric Oncology (SIOP)-Renal Tumor Study Group. This guideline is applicable for advanced image-guided radiotherapy techniques.

However, the widespread implementation of this new approach may be hampered by the time spent by radiation oncologists to delineate all organs/structures-at-risk and generate the postoperative target volume. It also includes the risk of significant delineation variations due to limited experience in the majority of centers.

So, this research project aims to [1] develop a tool for automatically contouring of organs/structures-at-risk and target volumes for flank irradiation and [2] validate these methods in an international setting.

The objectives will be achieved via work packages focusing on [1] the development of a computed tomography (CT)-based deep learning method to delineate the kidney, liver, spleen, pancreas, stomach/bowel, heart, lung, vertebrae and muscles, [2] a magnetic resonance imaging (MRI)-based deep learning method to delineate the primary tumor volume. The output of both work packages will be used to generate postoperative flank target volumes taking the volume shift after resection of the tumor into account. Finally, the deep learning methods need to be developed and validated in a multicenter setting.

It is expected that the introduction of this approach will provide accurate and robust auto-contouring of relevant abdominal/thoracic structures and target volumes, thereby providing harmonized segmentations to allow detailed assessment of tumor/normal tissue volumes. Additionally, the laborious manual tumor segmentations will be alleviated, improving (cost-) effectiveness by reducing staff workload.

This project will take place in collaboration with the ‘compatational imaging group’ (Prof. C.A.T. van den Berg; Dr. M Maspero).

Necessary skills for this position:

• A Master’s degree in computer sciences, physics, mathematics, artificial intelligence, medical imaging ortechnical medicine
• Knowledge on programming skills
• Willingness to learn clinical background
• Prior knowledge of deep learning or any affinity with medical imaging are a plus