Techniques to Improve the Accuracy of Presampling MTF Measurement in Digital X-ray Imaging Based on Constrained Spline Regression

• Published in: IEEE transactions on biomedical engineering Vol. 61; no. 4; pp. 1339 - 1349
• Main Authors: Zhou, Zhongxing, Zhu, Qingzhen, Zhao, Huijuan, Zhang, Lixin, Ma, Wenjuan, Gao, Feng
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.04.2014; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Algorithms; Biomedical measurement; C-spline regression; Computer simulation; Curve fitting; Digital; digital X-ray imaging; edge spread function; Humans; Image edge detection; Imaging; Measurement uncertainty; Models, Statistical; modulation transfer function; oversampling techniques; Radiographic Image Enhancement - methods; Radiographic Image Interpretation, Computer-Assisted - methods; Radiography; Regression; Regression Analysis; Smoothing methods; Splines (mathematics); Uncertainty
• ISSN: 0018-9294; 1558-2531; 1558-2531
• DOI: 10.1109/TBME.2014.2304955

To develop an effective curve-fitting algorithm with a regularization term for measuring the modulation transfer function (MTF) of digital radiographic imaging systems, in comparison with representative prior methods, a C-spline regression technique based upon the monotonicity and convex/concave shape restrictions of the edge spread function (ESF) was proposed for ESF estimation in this study. Two types of oversampling techniques and following four curve-fitting algorithms including the C-spline regression technique were considered for ESF estimation. A simulated edge image with a known MTF was used for accuracy determination of algorithms. Experimental edge images from two digital radiography systems were used for statistical evaluation of each curve-fitting algorithm on MTF measurements uncertainties. The simulation results show that the C-spline regression algorithm obtained a minimum MTF measurement error (an average error of 0.12% ± 0.11% and 0.18% ± 0.17% corresponding to two types of oversampling techniques, respectively, up to the cutoff frequency) among all curve-fitting algorithms. In the case of experimental edge images, the C-spline regression algorithm obtained the best uncertainty performance of MTF measurement among four curve-fitting algorithms for both the Pixarray-100 digital specimen radiography system and Hologic full-field digital mammography system. Comparisons among MTF estimates using four curve-fitting algorithms revealed that the proposed C-spline regression technique outperformed other algorithms on MTF measurements accuracy and uncertainty performance.