Air-Scan-Based Point-Spread-Function Estimation for Megavoltage CT with Flat-Panel Detectors

Accurate spatial-resolution characterization of flat-panel detectors at megavoltage energies is difficult due to long-tailed cross-talk. Conventional measurements at high energy using line-pair gauges or knife-edge phantoms require expensive, carefully fabricated objects and precise alignment, and despite these, scatter from the phantom, detector, and imaging room can further confound the measurement. We present a method for estimating the detector point spread function (PSF) from the mean noise power spectrum (NPS) calculated from repeated radiographs in air. The approach is mathematically equivalent to an autocovariance-based method but is more practical for long-tailed PSFs. We validate the method using simulated air data and Monte Carlo N-particle (MCNP6) calculations of energy deposition in a flat-panel detector. We show that deconvolving the experimental PSF increases the accuracy of transmission radiographs of step wedges. The method requires no phantom or alignments and is straightforward to reproduce.