A Physics-Driven Method for Photon-Counting Detector Calibration
Abstract
Purpose
Accurate calibration in photon-counting detectors (PCDs) for inhomogeneous pixel responses caused by charge sharing, pulse pile-up, and energy-dependent efficiency is the most critical and challenging task for PCD-based imaging applications. Conventional gain-based calibration method directly measured counts in low- and high- energy threshold channels (LE/HE) to corresponding expected counts. With a single input, it cannot disentangle or correct effects caused by both fluence and beam spectrum variation. We present a response calibration method that addresses both effects simultaneously.
Methods
Based on expressions of counts in LE/HE channels under given aluminum thickness and fluence, a new variable t, count ratio between the two channels, was found to only depend on aluminum thickness, and hence the spectrum. Raw counts at LE/HE channels (energy thresholds 26 and 70 keV) were acquired using a well-characterized 110 kV x-ray spectrum under a range of aluminum thickness and fluence levels controlled by current 0-200 µA. Measured t value and expected LE/HE counts were calculated. A bivariate polynomial fitting was employed to model the relationship between measured variables and expected LE/HE counts. The calibration model was integrated into a PCD-CT reconstruction pipeline and was evaluated through phantom studies.
Results
Compared with conventional gain-based calibration, the proposed method reduced the average relative error from 13% to 0.6% in the LE channel and from 1.7% to 0.7% in the HE channel. Phantom studies demonstrated that the ring artifacts can be greatly reduced compared to the conventional gain-based calibration methods. Quantitatively, the one-side uniformity index improved 3.88% over the gain-based calibration.
Conclusion
The proposed calibration method substantially improved the accuracy of calibration results compared with conventional gain-based calibration. It has the potential to enhance spectral imaging quantitative performance in PCD-CT applications.