Optimal parameters estimation for K-edge subtraction imaging using PixiRad-2/PixieIII photon counting detector on a conventional laboratory X-ray micro-tomograph

Photon Counting Detectors (PCDs) open new opportunities in X-ray imaging. Pixie III is a PCD using simultaneously two energy thresholds. This enables to acquire images using two distinct energy bins in a single exposure and allows to perform K-Edge Subtraction (KES) imaging with laboratory sources....

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2023-01
Main Authors: Granger, Rémi, Salvo, Luc, Sabine Rolland du Roscoat, Lhuissier, Pierre
Format: Article
Language:English
Subjects:
Barium
Bins
Energy
Image acquisition
Iodine
Laboratories
Mathematical models
Optimization
Parameters
Photons
Subtraction
Thresholds
X ray imagery
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Photon Counting Detectors (PCDs) open new opportunities in X-ray imaging. Pixie III is a PCD using simultaneously two energy thresholds. This enables to acquire images using two distinct energy bins in a single exposure and allows to perform K-Edge Subtraction (KES) imaging with laboratory sources. In that context, one has however to deal with an energy bin optimization: narrow energy bins lead to high KES signal at the expense of higher noise, while wider energy bins lead to poor KES signal but better statistics. This work presents a model that aims at finding the optimal energy thresholds and source voltage in order to retrieve the best Contrast to Noise Ratio (CNR) for a given sample. The model also optimizes the parameters for conventional absorption modality and compares both modalities. Since the input flux and the energy difference between the thresholds influence image noise, this is included in the model using phenomenological laws. The model is then compared to empirical optimization by experimental screening of the parameters using model materials composed of barium, iodine and water. Finally, it is explained how to model the influence of sample composition on the predicted CNR values.
ISSN:2331-8422
DOI:10.48550/arxiv.2207.14673