Neural-networks-based Photon-Counting Data Correction: Pulse Pileup Effect

TL;DR

This paper presents a neural network-based method to correct pulse pileup effects in photon-counting detectors, significantly improving data fidelity in medical imaging applications.

Contribution

The authors develop a model-free, end-to-end neural network approach that effectively compensates for pulse pileup effects in photon-counting detectors, a novel solution in this context.

Findings

Neural network reduces count error by an order of magnitude.

Approach maintains low variance across different attenuation paths.

Simulation results validate the effectiveness of the correction method.

Abstract

Compared with the start-of-art energy integration detectors (EIDs), photon-counting detectors (PCDs) with energy discrimination capabilities have demonstrated great potentials in various applications of medical x-ray radiography and computed tomography (CT). However, the advantages of PCDs may be compromised by various hardware-related degradation factors. For example, at high count rate, quasi-coincident photons may be piled up and result in not only lost counts but also distorted spectrums, which is called the pulse pileup effects. Considering the relative slow detection speed of current PCDs and high x-ray flux for clinical imaging, it is vital to develop an effective approach to compensate or correct for the pileup effect. The aim of this paper was to develop a model-free, end-to-end, and data-driven approach in the neural network framework. We first introduce the trigger threshold concept to regularize the correction process. Then, we design a fully-connected cascade artificial neural network and train it with measurements and true counts. After training, the network is used to recover raw photon-counting data for higher fidelity. To evaluate the proposed approach, Monte Carlo simulations are conducted. We simulate four types of PCDs combining the unipolar or bipolar pulse shape with the nonparalyzable or paralyzable detection mode. The simulation results demonstrate that the trained neural network decreases the relative error of count averages of raw data by an order of magnitude while maintaining a reasonable variance over a wide range of attenuation paths.