Gain correction for variable imaging geometry.

Area of Science:

• Medical Imaging
• Image Processing
• X-ray Technology

Background:

• Gain correction is crucial for X-ray imaging quality, compensating for detector variations.
• Conventional methods require geometry-matched references, which are impractical for variable-geometry systems.
• Variable geometry introduces challenges like inverse-square dilution and anode heel shading.

Purpose of the Study:

• To develop and evaluate a geometry-aware gain correction method.
• To synthesize geometry-matched reference fields from a compact calibration dataset.
• To enable accurate gain correction in variable geometry X-ray systems without specific reference acquisitions.

Main Methods:

• Modeled X-ray signal considering dark current, detector response, geometry-dependent dilution, and source fluence map.
• Utilized rigid transformations for mapping between detector and source domains.
• Developed a calibration process using dark acquisition, large SDD air acquisition, and moderate SDD air images to build a source fluence map.

Main Results:

• Synthetic reference images closely matched measured air references.
• Residual non-uniformity after correction was below 0.5% in air-only images.
• Chest phantom images showed high agreement between synthetic and matched reference corrections (SSIM=0.995, CCC=0.999).

Conclusions:

• The synthetic-reference approach provides accurate gain correction for variable imaging geometry.
• Performance is comparable to conventional matched-reference methods.
• Eliminates the need for geometry-specific air reference acquisitions in variable geometry systems.