Patient positioning with X-ray detector self-calibration for image guided therapy

Area of Science:

• Medical Imaging
• Image-Guided Therapy
• Radiological Physics

Background:

• Accurate patient alignment is critical in image-guided therapy, but misaligned X-ray cameras introduce significant inaccuracies.
• Existing calibration methods often rely on optical cameras, calibration bodies, or detectable features, which are frequently infeasible or undesirable in X-ray-based image-guided therapy.
• A method for X-ray camera self-calibration, usable during treatment and not requiring external features, is needed to enhance patient alignment precision.

Purpose of the Study:

• To develop and evaluate a novel approach for automatic X-ray camera self-calibration and 3D pose recovery.
• To improve the accuracy of patient alignment in image-guided therapy by compensating for camera misalignments.
• To demonstrate the robustness and effectiveness of the proposed method using phantom and real-patient data.

Main Methods:

• A stereoscopic X-ray camera model was proposed, integrating intensity-based calibration with 3D pose recovery.
• Relevant camera parameters were automatically computed by comparing X-ray images with CT images.
• The method was tested on an anatomic phantom with induced misalignments and real patient data from a particle therapy machine.

Main Results:

• The proposed approach successfully compensated for camera misalignments, reducing errors from over 5 mm to below 0.2 mm.
• The method demonstrated robustness against image degradation, tolerating 2-5% artificial noise.
• Rigid body alignment accuracy was significantly improved, confirming the feasibility of self-calibration without feature detection.

Conclusions:

• X-ray camera self-calibration is a viable and effective strategy for improving patient alignment accuracy in image-guided therapy.
• The developed method enhances precision by automatically correcting for camera misalignments, even when feature detection is not possible.
• This technique offers a significant advancement for image-guided interventions requiring precise patient positioning.