Brachyct: Real-Time Anatomical Reconstruction and Verification Using Sparse Ir-192 Gamma-Ray Emissions In HDR Brachytherapy
Abstract
Purpose
Accurate source localization and anatomical verification are critical for high dose rate (HDR) brachytherapy dose monitoring due to steep dose gradients. Our previous work only triangulated markers for source localization without fully utilizing the projection information for real-time anatomical imaging. This study investigates the feasibility of tomographic reconstruction using sparse gamma-ray emissions from the Ir-192 source.
Methods
A total of 32 brachytherapy dwell positions were sampled within the prostate planning target volume. For each position, projection data were collected using four detector panels positioned anterior/posterior and laterally to the patient, yielding a total of 128 views. Forward and backward projections were performed using a separable-footprint system model. The ground-truth image simulating the treatment anatomy contained a new air bubble within the rectum that was not present in the pre-treatment anatomy. Reconstruction was performed using a prior image constrained compressed sensing (PICCS) framework incorporating data fidelity, a quadratic prior constraint, and Total Variation (TV) regularization. The optimization is solved with a Fast Iterative Shrinkage-Thresholding Algorithm (FISTA).
Results
The air cavity was visible on the raw gamma-ray projections. With our PICCS framework, the reconstructed volumes successfully recovered the 3D rectal air cavity. Quantitative HU analysis of the air cavity ROI confirms that the reconstructed volume (-627.6 ± 76.4 HU) is substantially closer to the expected air attenuation value (−958.8 HU) compared with the prior CT (35.8 ± 26.8 HU).
Conclusion
Utilizing gamma-ray emissions from various HDR dwell positions, the separable-footprint-based PICCS framework facilitates the detection of anatomical changes during treatment, paving the way for real-time 3D imaging, dosimetry, and adaptive strategies.