Improving model-data mismatch for photon-counting detector model using global and local model parameters.

Area of Science:

• Medical Imaging Physics
• Detector Technology
• Computational Modeling

Background:

• Photon counting detectors (PCDs) offer clinical advantages but their energy-discriminating capability is degraded by charge sharing and pulse pileup, distorting X-ray spectra.
• Existing PCD output models show agreement but suffer from global model-data mismatch and pixel-to-pixel variations.

Purpose of the Study:

• To improve PCD modeling by incorporating count-rate-dependent parameters to address model-data mismatch and pixel variations.
• To evaluate the accuracy of the enhanced PCD model against physical detectors.

Main Methods:

• A cascaded model was adapted with condition-dependent and pixel-specific parameters to mitigate discrepancies.
• Model parameters were estimated using data from varying phantom thicknesses and X-ray tube currents.
• Performance was validated using CdZnTe-based PCDs, comparing models with default, global, and global-and-local parameters via Mean Absolute Percentage Errors (MAPEs).

Main Results:

• The proposed model demonstrated strong visual agreement with PCD measurements across various count rates.
• MAPEs were significantly reduced: 49.2-51.1% (default), 8.0-9.8% (global), and 1.2-2.7% (global-and-local parameters).

Conclusions:

• The enhanced PCD model accurately estimates detector outputs from low to high count rates.
• This model is expected to be valuable in applications requiring precise pixel-by-pixel PCD characterization.