Sensakovic WF, O'Dell MC, Letter H, Kohler N, Rop B, Cook J, Logsdon G, Varich L. Image quality and dose differences caused by vendor-specific image processing of neonatal radiographs.
Pediatr Radiol 2016;
46:1606-13. [PMID:
27488507 DOI:
10.1007/s00247-016-3663-2]
[Citation(s) in RCA: 10] [Impact Index Per Article: 1.3] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 11/17/2015] [Revised: 04/14/2016] [Accepted: 06/23/2016] [Indexed: 12/28/2022]
Abstract
BACKGROUND
Image processing plays an important role in optimizing image quality and radiation dose in projection radiography. Unfortunately commercial algorithms are black boxes that are often left at or near vendor default settings rather than being optimized.
OBJECTIVE
We hypothesize that different commercial image-processing systems, when left at or near default settings, create significant differences in image quality. We further hypothesize that image-quality differences can be exploited to produce images of equivalent quality but lower radiation dose.
MATERIALS AND METHODS
We used a portable radiography system to acquire images on a neonatal chest phantom and recorded the entrance surface air kerma (ESAK). We applied two image-processing systems (Optima XR220amx, by GE Healthcare, Waukesha, WI; and MUSICA(2) by Agfa HealthCare, Mortsel, Belgium) to the images. Seven observers (attending pediatric radiologists and radiology residents) independently assessed image quality using two methods: rating and matching. Image-quality ratings were independently assessed by each observer on a 10-point scale. Matching consisted of each observer matching GE-processed images and Agfa-processed images with equivalent image quality. A total of 210 rating tasks and 42 matching tasks were performed and effective dose was estimated.
RESULTS
Median Agfa-processed image-quality ratings were higher than GE-processed ratings. Non-diagnostic ratings were seen over a wider range of doses for GE-processed images than for Agfa-processed images. During matching tasks, observers matched image quality between GE-processed images and Agfa-processed images acquired at a lower effective dose (11 ± 9 μSv; P < 0.0001).
CONCLUSION
Image-processing methods significantly impact perceived image quality. These image-quality differences can be exploited to alter protocols and produce images of equivalent image quality but lower doses. Those purchasing projection radiography systems or third-party image-processing software should be aware that image processing can significantly impact image quality when settings are left near default values.
Collapse