Intercomparison of methods for image quality characterization. II. Noise power spectrum

• Published in: Medical physics (Lancaster) 2006-05, Vol.33 (5), p.1466-1475
• Main Authors: Dobbins, James T., Samei, Ehsan, Ranger, Nicole T., Chen, Ying
• Format: Article
• Language: English
• Subjects: Algorithms; digital radiography; IEC 62220-1; IMAGE PROCESSING; image quality; linear systems analysis; NOISE; Noise power spectrum (NPS); Quality Assurance, Health Care - methods; QUANTUM EFFICIENCY; Radiographic Image Enhancement - methods; Radiographic Image Interpretation, Computer-Assisted - methods; RADIOLOGY AND NUCLEAR MEDICINE; Reproducibility of Results; Sensitivity and Specificity; SPATIAL RESOLUTION; Stochastic Processes; SYSTEMS ANALYSIS
• ISSN: 0094-2405; 2473-4209
• DOI: 10.1118/1.2188819

Second in a two-part series comparing measurement techniques for the assessment of basic image quality metrics in digital radiography, in this paper we focus on the measurement of the image noise power spectrum (NPS). Three methods were considered: (1) a method published by Dobbins [ Med. Phys. 22 , 1581-1593 ( 1995 )] , (2) a method published by Samei [ Med. Phys. 30 , 608-622 ( 2003 )] , and (3) a new method sanctioned by the International Electrotechnical Commission (IEC 62220-1, 2003), developed as part of an international standard for the measurement of detective quantum efficiency. In addition to an overall comparison of the estimated NPS between the three techniques, the following factors were also evaluated for their effect on the measured NPS: horizontal versus vertical directional dependence, the use of beam-limiting apertures, beam spectrum, and computational methods of NPS analysis, including the region-of-interest (ROI) size and the method of ROI normalization. Of these factors, none was found to demonstrate a substantial impact on the amplitude of the NPS estimates ( ⩽ 3.1 % relative difference in NPS averaged over frequency, for each factor considered separately). Overall, the three methods agreed to within 1.6 % ± 0.8 % when averaged over frequencies > 0.15 mm − 1 .