A Surface Imaging-Guided Framework for Real-Time 3D Abdominal Tumor Motion Tracking Using Patient-Specific 4DCT Models
Abstract
Purpose
Effective radiotherapy for upper abdominal tumors requires dose escalation but is limited by respiratory motion and the low gastrointestinal radiation tolerance. Current clinical motion management on conventional Linacs relies on simple external surrogates that correlate poorly with internal organ motion. This work aims to develop a clinically feasible, non-invasive approach that uses continuous patient surface depth imaging to accurately predict internal abdominal tumor motion for real-time motion management. .
Methods
First, compact and patient-specific 4D respiratory motion models were derived from 4DCTs using deformable image registration (DIR) followed by principal component analysis (PCA). To emulate realistic free-breathing motion beyond phase-binned 4DCTs, respiratory signals from additional patients were applied to the PCA model to synthesize continuous motion. Corresponding time-resolved abdominal surface maps were then simulated directly from the motion-deformed anatomy. A lightweight CNN was trained on these synthesized surface maps to predict PCA coefficients in real-time. During treatment delivery, the surface maps can be acquired using a depth imaging camera to predict PCA motion coefficients and reconstruct continuous 3D internal motion for beam guidance. Performance of the motion tracking framework was evaluated in five pancreatic cancer patients using eight independent real patient breathing patterns per case.
Results
The 4DCT motion models accurately represented respiratory deformation, with three PCA components approximating dense motion at 0.1 ± 0.2 mm error. The proposed prediction framework reliably recovered continuous 3D motion across diverse breathing scenarios, achieving less than 0.5% relative errors on PCA coefficient prediction and a mean motion prediction error of 0.30 ± 0.06 mm.
Conclusion
This work represents a novel and systematic investigation of internal abdominal tumor motion prediction using skin surface depth imaging. The achieved accuracy and computational efficiency support clinical feasibility and offer a practical pathway to margin reduction, improved motion management, and enhanced treatment precision for real-time radiotherapy guidance.