Marker-free lung tumor trajectory estimation from a cone beam CT sinogram

Area of Science:

• Medical Physics
• Radiotherapy Technology
• Image-Guided Interventions

Background:

• Accurate 3D tumor localization is critical for effective lung cancer radiotherapy.
• Cone beam CT (CBCT) offers intra-fractional imaging but requires precise tumor position estimation.
• Respiration-induced tumor motion poses a significant challenge in lung cancer treatment planning and delivery.

Purpose of the Study:

• To develop and validate an algorithm for estimating 3D lung tumor position from CBCT projection data.
• To compare the performance of various template registration algorithms for tumor localization.
• To assess the accuracy and robustness of the developed method in phantom and clinical studies.

Main Methods:

• An algorithm utilizing CBCT sinogram data and template registration was developed.
• Templates were generated from pre-existing respiration-correlated CT images.
• Robust registration metrics (block correlation, robust correlation coefficient, robust gradient correlation) were evaluated and compared.
• Registration search regions were constrained using mean tumor position data.

Main Results:

• Robust registration metrics demonstrated reduced sensitivity to occlusions from overlying tissue and the treatment couch.
• The algorithm achieved a mean accuracy of 1.4 mm in phantom studies using a robust registration method.
• In two patients with peripheral lung tumors, mean position and excursion were estimated within 2.0 mm accuracy compared to 4D CT registration.

Conclusions:

• The developed template registration algorithm provides accurate and robust 3D lung tumor position estimation from CBCT data.
• Robust registration methods are advantageous in overcoming imaging artifacts and occlusions common in CBCT.
• This technique shows promise for improving image-guided lung cancer radiotherapy by enhancing tumor localization accuracy.