Premium
Effects of calibration methods on quantitative material decomposition in photon‐counting spectral computed tomography using a maximum a posteriori estimator
Author(s) -
Curtis Tyler E.,
Roeder Ryan K.
Publication year - 2017
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1002/mp.12457
Subject(s) - imaging phantom , calibration , photon counting , estimator , materials science , medical imaging , attenuation , optics , photon , mathematics , physics , statistics , computer science , artificial intelligence
Purpose Advances in photon‐counting detectors have enabled quantitative material decomposition using multi‐energy or spectral computed tomography ( CT ). Supervised methods for material decomposition utilize an estimated attenuation for each material of interest at each photon energy level, which must be calibrated based upon calculated or measured values for known compositions. Measurements using a calibration phantom can advantageously account for system‐specific noise, but the effect of calibration methods on the material basis matrix and subsequent quantitative material decomposition has not been experimentally investigated. Therefore, the objective of this study was to investigate the influence of the range and number of contrast agent concentrations within a modular calibration phantom on the accuracy of quantitative material decomposition in the image domain. Methods Gadolinium was chosen as a model contrast agent in imaging phantoms, which also contained bone tissue and water as negative controls. The maximum gadolinium concentration (30, 60, and 90 mM ) and total number of concentrations (2, 4, and 7) were independently varied to systematically investigate effects of the material basis matrix and scaling factor calibration on the quantitative (root mean squared error, RMSE ) and spatial (sensitivity and specificity) accuracy of material decomposition. Images of calibration and sample phantoms were acquired using a commercially available photon‐counting spectral micro‐ CT system with five energy bins selected to normalize photon counts and leverage the contrast agent k ‐edge. Material decomposition of gadolinium, calcium, and water was performed for each calibration method using a maximum a posteriori estimator. Results Both the quantitative and spatial accuracy of material decomposition were most improved by using an increased maximum gadolinium concentration (range) in the basis matrix calibration; the effects of using a greater number of concentrations were relatively small in magnitude by comparison. The material basis matrix calibration was more sensitive to changes in the calibration methods than the scaling factor calibration. The material basis matrix calibration significantly influenced both the quantitative and spatial accuracy of material decomposition, while the scaling factor calibration influenced quantitative but not spatial accuracy. Importantly, the median RMSE of material decomposition was as low as ~1.5 mM (~0.24 mg/ mL gadolinium), which was similar in magnitude to that measured by optical spectroscopy on the same samples. Conclusion The accuracy of quantitative material decomposition in photon‐counting spectral CT was significantly influenced by calibration methods which must therefore be carefully considered for the intended diagnostic imaging application.