Performance Evaluation of Deformable Image Registration Systems - SmartAdapt^® and Velocity™

Area of Science:

• Medical Physics
• Radiotherapy
• Image Analysis

Background:

• Deformable image registration (DIR) is crucial for adaptive radiotherapy, enabling accurate target delineation and dose escalation.
• Commercial DIR systems require rigorous validation against established benchmarks like Task Group 132 (TG-132) guidelines.

Purpose of the Study:

• To commission and validate two commercial DIR systems, SmartAdapt® and Velocity™, using TG-132 digital phantom datasets.
• To compare the DIR algorithm performance of SmartAdapt® and Velocity™.

Main Methods:

• TG-132 digital phantoms were used for commissioning and validation of SmartAdapt® and Velocity™.
• Registration accuracy was assessed using rigid and deformable registrations, comparing results against known shifts.
• Dice Similarity Coefficient (DSC), Mean Distance to Agreement (MDA), and Jacobian determinant were used for evaluation.

Main Results:

• Both systems met TG-132 criteria for translation-only registration, except for CT-PET. Translational and rotational registration failed TG-132 criteria for all modalities.
• Velocity™ showed slightly better performance in soft tissue registration (parotids, bladder, rectum, prostate) compared to SmartAdapt®.
• Both systems demonstrated comparable results for rigid structures like the mandible, spinal cord, and femoral heads.

Conclusions:

• SmartAdapt® and Velocity™ DIR systems are suitable for clinical applications.
• While minor differences exist, Velocity™ demonstrated a slight advantage in soft tissue registration accuracy.
• Further validation with clinical data is recommended for comprehensive assessment.