Jennifer Dhondt
Phd Student
Dynamic Tumor Tracking

In stereotactic body radiotherapy (SBRT), respiratory induced motion, the extent of which is illustrated in Figure 1, has been shown to significantly influence both geometric and dosimetric accuracy in the delivery of therapeutic doses to thoracic and abdominal tumors. As a consequence, the risk of over-dosage to nearby healthy tissue and organs at risk, or a geometrical miss, increases significantly. Various techniques have been developed that incorporate or manage breathing motion, from dedicated treatment margins in the planning stage to real-time tumor tracking (RTTT) during dose delivery. The latter has been reported to significantly reduce the treatment uncertainties brought up by respiratory induced motion, while still allowing the patient to breath freely and maintaining acceptable treatment times.

However, the only two systems commercially available that are capable of performing this type of RTTT, the Cyberknife Robotic Radiosurgery system and the Vero SBRT system, both rely on implanted fiducials for robust and accurate target localization on kV images. Studies evaluating the transthoracic implantation of these markers have reported severe risk for complications such as pneumothorax, and the presence of possible marker migration with respect to the tumor. The Cyberknife system has therefore been implemented with the XSight lung tracking protocol for markerless tumor tracking, relying on 2D/3D intensity based registration. However, for this method to be successful, the tumor has to comply to a certain degree of visibility, limiting the applicability of the protocol to only a small subgroup of patients which require tracking. On the other hand, the Vero SBRT system is not yet equipped with a markerless solution, requiring an implanted fiducial for all tracking patients.

The aim of this PhD project is to optimize the dynamic tumor tracking workflow. Firstly, by analyzing breathing-induced tumor motion through different imaging modalities such as MR, 4D-CT and X-ray fluoroscopy and evaluating the accuracy of these modalities. Secondly, by evaluating if a treatment decision system can be created through texture analysis, the breathing motion captured during treatment by either X-ray fluoroscopy or cine-MR, and machine learning for modelling. And lastly, by developing a markerless tumor tracking methodology, using the Vero SBRT system as a testing platform and the currently implanted fiducials as a benchmark, while ensuring that the methodology can be translated to other less dedicated treatment modalities.