Observer Performance Studies

Reduction of Patient Exposure in Pediatric Radiology

The goal of this study was to determine if presently used exposure levels in pediatric imaging can be reduced without loss of information or decrease in diagnostic radiology. An ROC study was conducted as well as physical measurements of the images.[8].



  • 32 computed radiography neonate chest images
  • 16 lesion-free
  • 16 with hyaline membrane disease (HMD)
  • A method of multiple stacked detectors was used to acquire the exact same image of the same patient at 4 different dose levels
  • The final test set had 128 images
  • Each successive plate in the stack represents a 50% exposure reduction
  • The multiple stack method involves placing 4 imaging plates on top of each other in the same CR cassette


  • Standard viewbox


  • 7 staff radiologists read all the images in a randomized order


  1. Report HMD present or absent
  2. Report confidence in decision using 6-level scale
  3. Rate image quality as poor, fair, good or excellent


ROC Az Results

  No Dose Reduction 50% Reduction 75% Reduction 90% Reduction
AZ Mean .920 .905 .903 .871

ANOVA analysis revealed that Exposure 4 (90% dose reduction) resulted in significantly lower Az performance than the other exposure levels (df = 3, F = 4.563, p = .01) Az for exposure levels 1, 2 & 3 did not differ significantly.

Mean TP & FP Rates

Exposure % TPs %FPs
No Reduction 83 18
50% 83 21
75% 80 17
90% 69 14

Image Quality

Exposure Mean Rating (4 = excellent, 1 = poor)
No Reduction 3.29
50% 2.53
75% 1.86
90% 1.25

ANOVA revealed that all quality ratings differed significantly from each other (df = 3, F = 516.56, p = .0001).


  • Decreasing exposure levels to about 75% of current levels may be acceptable in some clinical situation where dose is a concern, such as a in pediatric imaging
  • Diagnostic performance (ROC Az) did not decrease significantly until exposure was reduced by 90%
  • At 90% reduction the TP & FP rates decreased
  • Image quality decreased significantly with each drop in exposure, but it did not affect diagnostic performance until the 90% reduction level
  • Perhaps full dose images can be acquired when a pediatric patient is first admitted, then lower dose images can be acquired for follow-up exams

Return to Krupinski page