MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Deep Spatial Pyramid Convolutional Framework-Based MR-Only Abdominal Adaptive Radiation Therapy: A Preliminary Feasibility Study

S Olberg1*, J Chun2 , B Cai1 , D Yang1 , M Roach1 , H Kim1 , J Shin2 , J Kim2 , S Mutic1 , O Green1 , J Park1 , (1) Department of Radiation Oncology, Washington University in St. Louis, St Louis, MO 63110, (2) 2Department of Radiation Oncology, Yonsei Cancer Center, Yonsei University College of Medicine, Seoul, South Korea

Presentations

(Sunday, 7/14/2019) 5:00 PM - 6:00 PM

Room: Stars at Night Ballroom 2-3

Purpose: MRI-guided adaptive radiation therapy (MR-IG-ART), enabled by using MRI for planning and staging, aims to improve outcomes by optimizing treatment plans based on a patient’s daily anatomy that changes fractionally. Accurate dose calculations are critical for ART for abdominal cases. Intestinal gas in the bowel changes size and position relative to the target frequently, significantly impacting dose delivery to the target and surrounding OARs due to changes in attenuation in air. We present a preliminary study of a hybrid deep learning method for synthetic CT (sCT) reconstruction, aiming to enable accurate MR-only ART in this challenging region.

Methods: Viewing sCT reconstruction as an image-to-image translation task, we generate sCT data based on a patient’s MRI data using a novel deep spatial pyramid convolutional framework. To support accurate dose calculation around intestinal gas, we bolster the reconstruction with an automated segmentation process that identifies and propagates gas pockets from MR images to the generated sCT images.

Results: The preliminary results illustrate that the clinical plan calculated based on the simulation CT overestimates the target and OAR doses compared to the plan calculated based on our sCT with propagated gas pockets. The percentage of the PTV covered by 95% of the prescribed dose is 17% higher in the clinical plan. Similarly, max doses to the liver, stomach, large bowel, and duodenum are 6%, 6%, 3%, and 5% higher respectively in the clinical plan compared to the proposed plan.

Conclusion: The proposed sCT reconstruction method shows promise for MR-only ART for the abdomen, allowing for more accurate dose calculations when the day's anatomy differs greatly from that represented at simulation. Improved accuracy affords the opportunity to escalate the target dose when possible, which is critical for sites for which adequate coverage is a challenge to achieve.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email