MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Development and In-Patient Validation of Inter-Fraction Fidelity of a Surface Photogrammetry+CT-Based Volumetric Motion Model for Lung Radiotherapy

M Ranjbar1*, P Sabouri2 , S Mossahebi3 , D Leiser4 , G Lasio5 , J Hung6 , A Sawant7 , J Zhou8 , (1) UMBC/UMB, Baltimore, MD, (2) University of Maryland School of Medicine, Baltimore, MD, (3) University of Maryland School of Medicine, Baltimore, MARYLAND, (4) U maryland, Baltimore, (5) University of Maryland School of Medicine, Bel Air, MD, (6) U maryland, Baltimore, ,(7) University of Maryland School of Medicine, Baltimore, MD, (8) University of Maryland School of Medicine, Bel Air, MD

Presentations

(Sunday, 7/14/2019)  

Room: ePoster Forums

Purpose: Current limitations in accounting for respiration-induced complex volumetric changes hinder accurate geometric and dosimetric targeting in lung radiotherapy (RT). While numerous volumetric motion models have been proposed, , to our knowledge, no methodology exists for validating the inter- and intra-fraction accuracy of these models. Here, we develop a motion model that uses surface photogrammetry and 4DCT to generate a multi-cycle, time-varying, anatomical volume. Subsequently, we develop and empirically demonstrate an in-patient validation strategy based on comparing 2D+time digital fluoroscopy generated from the motion model with in-room kV fluoroscopy.

Methods: Following IRB approval, a consented lung-cancer patient underwent 4DCT scan concurrently with surface photogrammetry using a research VisionRT (VRT) system. Using end-exhale as reference, deformation vector fields (DVFs) were computed from each other phase to reference. A learning model was trained using the 4DCT-DVFs and the acquired VRT surfaces. For each surface instance, the model generated a new 3DCT. Inter-fraction accuracy and fidelity of the model were evaluated as follows: during first and third fractions, immediately prior and following treatment, fluoroscopic images (30sec;7fps;100kVp;7mAs) with concurrent surface monitoring (~15 Hz). Using the surfaces as model input, a model-estimated CT that corresponds to each fluoro time sequence was created. A ray-tracing algorithm was used to generate digitally reconstructed fluorographs (DRFs) and contours of the liver, and positions of three patches on the liver as observed on the DRFs were compared with fluoroscopy.

Results: For three of the four fluoroscopic acquisition sessions, the model out-performed 4DCT in estimating liver contours (Dice-model: 0.9-0.97, Dice-4DCT: 0.88-0.96). Model- vs 4DCT-measured position-errors ranged from 5 to -15mm vs 5 to -22.4mm, respectively.

Conclusion: These early results indicate that our model accurately estimates patient anatomy over multiple respiratory cycles/fractions. The DRF-based validation developed here is model-agnostic and, therefore, a powerful tool for in-patient validation of other volumetric models.

Funding Support, Disclosures, and Conflict of Interest: This study was partly supported by R01 CA169102 and Varian Medical Systems.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email