Robust breathing signal extraction from cone beam CT projections based on adaptive and global optimization techniques

Area of Science:

• Medical Physics
• Radiotherapy Imaging

Background:

• Accurate respiratory motion management is crucial in radiotherapy, especially for thoracic and abdominal tumors.
• Current methods for respiratory signal extraction from imaging data have limitations.

Purpose of the Study:

• To develop and validate a markerless method for extracting respiratory signals from cone beam computed tomography (CBCT) projections.
• To enhance the Amsterdam Shroud (AS) technique for improved respiratory signal extraction.

Main Methods:

• CBCT projections were preprocessed to create enhanced attenuation images using adaptive robust z-normalization filtering.
• A two-step optimization approach was used to extract respiratory signals from the enhanced AS images.
• The proposed method was evaluated using CBCT data from five patients and compared against a reference air bellows belt system.

Main Results:

• Stable and reliable respiratory signals were extracted with an average error of -0.07 bpm and a standard deviation of 1.58 bpm compared to reference waveforms.
• The new algorithm demonstrated an improvement of 8.5% to 30% over the original AS technique.
• Gantry rotation was found to have a minimal impact on the extracted breathing signal frequency.

Conclusions:

• The proposed technique offers a practical, markerless solution for extracting respiratory signals from CBCT projections.
• This method has the potential to improve the accuracy and efficiency of radiotherapy for thoracic and abdominal cancer patients.