Radiography driven deformable image registration in adaptive proton therapy

• Applicants: Dr.-Ing. Prasannakumar Palaniappan; Professor Dr.-Ing. Marco Riboldi
• Subject Area: Medical Physics, Biomedical Technology
• Term: from 2021 to 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 455550444

In this project, we aim to account for anatomical changes in proton therapy by using in-room proton radiographies through a novel deformable image registration (DIR) algorithm. In clinical practice, early diagnosis and precise dose delivery are strategical to maximize therapeutic outcome and minimize side effects. In adaptive radiation therapy, the treatment plan is updated relying on three dimensional (3D) in-room tomographic imaging methodologies such as cone beam computed tomography (CBCT). CBCT images suffer from low image quality due to scatter contamination and image noise at the projection level. Alternative to CBCT, proton therapy facilities can provide proton imaging capabilities. However, clinical and geometrical constraints can prevent 3D in-room tomographic imaging. Based on the 2D in-room proton radiographies (pRads) a DIR algorithm is developed to update the treatment planning CT image of the patient, thus enabling lower imaging dose compared to traditional CBCT. A dedicated methodological development of 2D-3D DIR will be carried out to compensate anatomical mismatches. The proposed DIR algorithm is aimed to be clinically feasible in terms of flexibility, automation and high accuracy. In addition, with GPU enhancement the 2D-3D DIR will be designed in the perspective of a realistic clinical application. In order to assess the feasibility of the developed DIR algorithm, dedicated simulation studies considering both rigid and non-rigid compensation are exploited. Such simulations will rely on phantoms and clinical images at variable complexity and realism. Monte Carlo (MC) simulations for ideal detectors and a detailed Monte Carlo emulator of a prototype proton CT scanner will be pursued for validation. In addition, a deformable PMMA phantom modeling head and neck region and a dedicated 3D printed motion phantom are developed for experimental measurements. For realistic anatomical mismatches, clinical CBCT images taken over the fractionated treatment are considered and a FLUKA MC simulations for different (ideal) list-mode and integration-mode detectors in pencil beam scanning are adopted. The 2D-3D DIR is quantified using dedicated contours and patient-specific measures for accuracy evaluation. A comparative study based on 2D-3D DIR using pRads and X-ray projections of CBCT is carried out to explore current research perspectives in ideal low dose imaging for treatment plan adaptation in proton therapy.

• DFG Programme: Research Grants