Virtual cone-beam computed tomography simulator with human phantom library and its application to the elemental...

Area of Science:

• Medical Physics
• Radiology
• Computer Vision

Background:

• Cone-beam CT (CBCT) is crucial for radiation therapy planning.
• Quantitative imaging requires accurate material decomposition.
• Existing methods face challenges with scatter and noise.

Purpose of the Study:

• Develop a virtual CBCT simulator with a head and neck (HN) human phantom library.
• Demonstrate the feasibility of elemental material decomposition (EMD) for quantitative CBCT imaging.
• Validate the simulator's performance using real patient data.

Main Methods:

• Created a library of 36 HN phantoms based on anthropometric statistics.
• Developed a virtual CBCT simulator using ray-tracing and deep-learning (DL) for X-ray simulation.
• Incorporated Gaussian noise and evaluated against a real CBCT system.
• Applied a DL-based EMD model to virtual and real patient data.

Main Results:

• The virtual simulator generated diverse CBCT images from the phantom library.
• Successful EMD prediction was achieved using the virtual system's CBCT database.
• Image degradation from scatter and noise minimally impacted EMD prediction accuracy.
• Elemental distribution was predictable from real CBCT images.

Conclusions:

• The virtual CBCT simulator facilitates medical data preparation and analysis.
• Computer vision holds significant potential for enhancing quantitative imaging.
• This technology can improve patient outcomes, particularly in adaptive radiation therapy.