One of the key challenges in modern prostate radiotherapy is the anatomical variation of the patient between treatment fractions, which leads to discrepancies between the planned and delivered dose and reduces treatment accuracy. Adaptive radiotherapy and daily CBCT imaging can mitigate these issues, but the effectiveness of the process strongly depends on the quality of the registration between planning CT and CBCT scans.
In clinical practice, alignment is typically performed by manually adjusting the correspondence of fiducial markers, a time-consuming procedure with an unknown and operator-dependent error. In this work, I developed a method that automatically detects fiducial markers in both imaging modalities, correctly pairs them, and computes the optimal rigid transformation using geometric criteria. Particular emphasis was placed on assessing registration accuracy using metrics derived from the geometry of the marker triangle.
The results show that this approach provides a consistent and quantitative estimate of registration error and offers a basis for automated clinical alignment in adaptive radiotherapy. Such a tool could reduce the workload of radiation therapists and improve the reliability of daily treatment adaptations.
|
