Receiver operating characteristic analysis of fracture and pneumonia detection: Comparison of laser-digitized workstation images and conventional analog radiographs

Stacey J. Ackerman, Joseph N. Gitlin, Robert W. Gayler, Charles D. Flagle, R. Nicholas Bryan

Research output: Contribution to journalReview articlepeer-review

53 Scopus citations

Abstract

Forty pneumonia and 40 fracture cases were matched by patient age and gender with normal cases. Pediatric cases constituted 25% of this sample. All 160 examinations were laser-digitized at a spot size of 210 μm, a contrast resolution of 12 bits, and a spatial resolution of 2.35 line pairs per millimeter. The digitized images were transmitted 5 miles over a dedicated telephone line at a T1 rate from Francis Scott Key Hospital to Johns Hopkins Hospital (both in Baltimore, Md). At Johns Hopkins Hospital, eight radiologists interpreted the original analog radiograph and corresponding soft-copy display (1,280 × 1,024 pixels). Findings, confidence ratings, and image quality were reported for each reading. A receiver operating characteristic (ROC) analysis was conducted to compare observer performance under the analog and digital reading conditions. The overall sensitivity of the analog method was 89% versus 78% for the digital method (P < .001), while the specificity values were 96% and 92%, respectively (P > .20). ROC analysis similarly indicated a statistically significant difference in favor of analog radiographs (P = .028).

Original languageEnglish (US)
Pages (from-to)263-268
Number of pages6
JournalRADIOLOGY
Volume186
Issue number1
DOIs
StatePublished - Jan 1993

Keywords

  • Diagnostic radiology, observer performance
  • Picture archiving and communication system (PACS)
  • Receiver operating characteristic curve (ROC)
  • Video systems

ASJC Scopus subject areas

  • Radiology Nuclear Medicine and imaging

Fingerprint

Dive into the research topics of 'Receiver operating characteristic analysis of fracture and pneumonia detection: Comparison of laser-digitized workstation images and conventional analog radiographs'. Together they form a unique fingerprint.

Cite this