Click here to

Session: Multi-Disciplinary General ePoster Viewing [Return to Session]

Artificial Intelligence-Aided Morphological Design for PET-Guided Radiotherapy

K Wang*, M Huq, UPMC Hillman Cancer Center and University of Pittsburgh School of Medicine, Pittsburgh, PA

Presentations

PO-GePV-M-319 (Sunday, 7/10/2022)   [Eastern Time (GMT-4)]

ePoster Forums

Purpose: PET-guided radiotherapy became available in recent years, however, the PET unit for this purpose needs more renovations to fulfil its capabilities. Like CBCT vs CT, the design of PET unit for IGRT should differ from its ancestor which was for diagnostic usage because the clinical roles migrated. In the study, we propose a novel morphological design of PET with artificial intelligence (AI) for IGRT, focusing on treatment alone.

Methods: In PET-guided radiotherapy, the target position is already known, thus information other than the target is less important for treatment hence could be ignored, and this is the major difference from diagnostic PET. With helps of AI, the information for known target is pre-loaded to the IGRT PET unit prior to treatment. Unlike diagnostic PET which tends to generate a noisy scan of “entire” information with fixed and low collimation for detectors, all detectors in our morphological PET scanner could adjust their orientations hence point to the known target position. In addition, the collimation heights for each detector are adjustable to filter out noises and to optimize the entire collecting efficiency, based on the pre-loaded target information. During the first 1-2 minute of imaging prior to the treatment, the PET unit could fine tune its state based on the real-time imaging information.

Results: In this morphological design, PET image for IGRT purpose could be achieved, focusing on the target only. This allows much higher contrast and much less noises in the IGRT PET scan. Due to high resolution of the scan, inter-fractional volume variation of the target could be precisely acquired, hence an adaptive treatment plan could be quickly generated if needed.

Conclusion: We proposed a novel AI-aided morphological design for IGRT PET, focusing on target tracking alone. We also recommended a procedure for adapted RT based on this design.

Keywords

PET, Image Guidance

Taxonomy

Not Applicable / None Entered.

Contact Email

Share: