Novel automated treatment planning approaches to deal with microscopic disease in radiotherapy

Abstract

The scope of a radiotherapy treatment is to induce damage to tumor cells by irradiating them with ionizing radiation. The radiation should target the tumor as visible in images, also called the macroscopic disease. There is, however, a risk of small, invisible groups of cancer cells in the area surrounding the macroscopic tumor, the microscopic disease. For an optimal clinical outcome of radiotherapy, the microscopic disease needs to be irradiated as well. In current treatment planning, the macroscopic tumor is extended with a margin to better ensure full disease coverage. Because there is uncertainty in the extent of the microscopic disease, also the margin definition is uncertain. Moreover, in current planning, the same margin is used for all patients, and there is no patient-specific exploration of the trade-offs between dose extensions to cover potentially present microscopic disease vs radiation-induced toxicity because of enhanced dose in organs at risk (OARs) surrounding the tumor. This study aimed to explore whether the margin concept could be replaced by more advanced, individualized approaches for irradiation of volumes just outside the macroscopic tumor. All plans were generated automatically to avoid bias by human planners and to reduce workload. Two novel treatment planning approaches were investigated. The first focused on increasing dose coverage of the microscopic disease while at the same time controlling the resulting increased dose to the healthy tissues surrounding the macroscopic tumor, including OARs. Several treatment plans were generated to explore a range of trade-offs between dose in microscopic disease and dose in OARs. The results showed that for optimal microscopic disease irradiation, both low and high OAR doses needed to worsen. However, the most significant increase in microscopic disease coverage could be obtained when accepting higher OAR low doses, that is generally less important for induction of negative side-effects.
In the second approach, an expected Tumor Control Probability (expected TCP) cost function was used to control dose delivery in areas close to the macroscopic tumor. Basis of the expected TCP model was a function describing the probability of finding microscopic disease at a specific distance from the macroscopic tumor. The results again demonstrated opportunities to increase dose to areas close to the tumor, at the cost of enhanced doses in OARs. In conclusion, both approaches had a positive impact on microscopic disease dose coverage. However, improved irradiation of the microscopic disease was always at a price of enhanced dose in OARs. Complementary studies, involving clinicians, need to be carried out to investigate if and how the approaches could be used to replace the current planning with fixed margins.