Anatomically-adaptive multi-modal image registration for image-guided external-beam radiotherapy.

Area of Science:

• Medical physics
• Radiotherapy
• Medical imaging

Background:

• Image-guided radiotherapy (IGRT) uses imaging to monitor tumors and organs-at-risk (OAR) during treatment, aiding in reducing uncertainties.
• Manual assessment of tissue displacement is time-consuming and prone to errors, especially with multiple imaging modalities.
• Current deformable image registration (DIR) algorithms often assume uniform deformation across all structures, leading to potential anatomical implausibility.

Purpose of the Study:

• To develop and evaluate an anatomically-adaptive variational multi-modal DIR algorithm for improved accuracy and anatomical plausibility.
• To address the limitations of global deformation assumptions in existing DIR methods.
• To create a DIR algorithm suitable for online adaptive IGRT applications.

Main Methods:

• Proposed an anatomically-adaptive variational multi-modal DIR algorithm utilizing a regionalized registration model.
• Compared the proposed algorithm against two existing DIR methods with global deformation assumptions.
• Evaluated the algorithm's performance in terms of anatomical plausibility, accuracy, and computational speed.

Main Results:

• The proposed anatomically-adaptive DIR algorithm demonstrated improved anatomical plausibility of estimated deformations across the field-of-view.
• Achieved higher overall accuracy compared to existing global DIR approaches.
• The algorithm is computationally fast, making it suitable for online adaptive radiotherapy scenarios.

Conclusions:

• Anatomically-adaptive DIR models offer superior performance over global models for image-guided radiotherapy.
• The proposed method enhances precision in contour propagation and dose accumulation for adaptive IGRT.
• This approach holds significant potential for future adaptive IGRT workflows requiring on-the-fly adjustments.